text
stringlengths 87
880k
| pmid
stringlengths 1
8
| accession_id
stringlengths 9
10
| license
stringclasses 2
values | last_updated
stringlengths 19
19
| retracted
stringclasses 2
values | citation
stringlengths 22
94
| decoded_as
stringclasses 2
values | journal
stringlengths 3
48
| year
int32 1.95k
2.02k
| doi
stringlengths 3
61
| oa_subset
stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1610483310.1371/journal.pmed.0020250Policy ForumHealth EconomicsHealth PolicyPediatricsPediatricsHealth PolicyResource allocation and rationingSocioeconomic determinants of healthPublic HealthWhat Can We Do to Improve Child Health in Southern Italy? Policy ForumBonati Maurizio *Campi Rita *To whom correspondence should be addressed. E-mail: [email protected] Bonati is Head of, and Rita Campi is Senior Researcher at, the Mother and Child Health Laboratory at the Mario Negri Institute of Pharmacological Research, Milan, Italy.
Competing Interests: The authors declare that no competing interests exist.
9 2005 23 8 2005 2 9 e250Copyright: © 2005 Bonati and Campi.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Southern Italy has one of the highest rates of poverty in Europe, and children's health status in this region is alarming. Bonati and Campi outline the crucial policies that would address this crisis.
==== Body
Two-thirds of all poor Italian families are in the south. Of the 20.7 million people (36.1% of Italians) living in southern Italy, 7.3 million (35.4%) are poor, living on less than €521 per month. Some 4.6 million of these people (63.3%) are extremely poor, living on less than €435 per month [1,2]. With such an economic profile, if southern Italy was thought of as an independent European country, it would be the European country with the highest poverty rate, weighted for national income. In this article, we discuss health inequalities in Italy, with a particular focus on the very poor health status of children in southern Italy, and we ponder possible health policies to address these inequalities [3].
How does poverty in Italy compare with poverty in the rest of Europe? Within the 25 European Union countries, in 2001 about 68 million people lived on less than 60% of their respective national median income [4]. Of these 68 million, 3.6 million were children under five years old [4]. The proportion of people at risk of poverty, redistributed based on a value of one as the EU average, ranges from a minimum of 0.3 in Slovakia and a maximum of 1.4 in Ireland. Italy's average value is 0.6, but is 0.4 in the north, 0.9 in the center, and 2.3 in the south.
Southern Italy can, therefore, be considered the European country with the highest risk of poverty. The proportion of people who are under five years old in southern Italy is 6.4%—thus, southern Italy has the third highest percentage of children under five, after Ireland and Cyprus.
The Crisis of Poor Child Health in Southern Italy
The monitoring and planning of child welfare is crucial for the health of a community and for a nation's public health in general. Social and economic factors are determinants of child health inequalities [5,6]. Inequalities within countries, including developed ones, are well-known to affect the health of populations, in particular that of minorities such as children [7,8].
Two-thirds of cases accounting for the national infant mortality rate (3.3 per 1,000 live births) involve neonatal death, especially within the first week of life (early neonatal mortality). Neonatal mortality varies widely between the different Italian regions, with rates four times higher in the south (Sicily and Basilicata's rates are 5.7 per 1,000 live births) than in the north (Friuli Venezia Giulia's rate is 1.3 per 1,000 live births) (Figure 1 shows the different regions of Italy) [9].
Figure 1 The Different Regions of Italy
Although birth weight is one of the recognized factors that contribute to infant mortality, there are no regional variations in the distribution of low and very low birth weight infants. However, the risk of early neonatal death for low birth weight infants born, for example, in Sicily or Abruzzo is more than nine times higher than the risk for newborns in the Aosta Valley (rates of 91.7 and 101.7 per 1,000 live births versus 11.4 per 1,000 live births, respectively). One possible explanation for this could be a wide gap in perinatal care quality (health-care structures and quality of health assistance) associated with latitude in Italy [10].
Hospitalization rates during childhood (age 14 years or younger) are similar between regions (average 151.8 per 1,000 inhabitants), and are higher in infants, in male children, and for respiratory system diseases. However, more than 22% of hospitalized children from the Basilicata and Molise regions and more than 13% of hospitalized children from the Calabria and Abruzzo regions are treated in hospitals in northern or central Italy, also suggesting a lack of adequate pediatric services in the south, both in terms of quality and quantity [9]. Because of unequal distribution of services in the south, in many circumstances people would need to move across the region to access health services, and traveling to the north may be a simpler option since, according to cultural beliefs, the quality of public services are thought to be better in the north [11].
The Italian National Health Service
The Italian National Health Service (NHS), introduced in 1978, provides universal coverage and comprehensive health care, free of cost or at a nominal charge upon delivery. The NHS is structured in three different levels of public authority: the central government, 20 regions, and 196 local health units covering an average of 290,000 persons each. Despite a strong, recent devolution policy, shifting power to the regions, the Italian government provides most of the funding for the NHS and is responsible for ensuring the NHS provides uniform, essential levels of health-care services across the country.
One of the most significant features of the public system is the gate-keeping function of the family practitioner (or family pediatrician for children under six years). Every Italian resident is required to register with a family practitioner, who is responsible for prescribing pharmaceuticals and diagnostic procedures and for referring patients to specialists and hospitals.
Although the World Health Report 2000 ranked the Italian health-care system second among 191 countries (France was the first) with respect to health status, fairness in financial contribution, and responsiveness to people's expectations of a health-care system [35], the dissatisfaction of Italians with respect to the efficiency and quality of their NHS is the highest in Europe [36]. Moreover, there is considerable disparity between the southern and northern Italian regions, with respect to satisfaction with NHS performance: for example, 19% in Sicily versus 53% in Emilia Romagna [37]. Tight budgets and the need to restrain rising health-care expenditures have led the NHS to undertake several cost-containment measures to encourage cost-conscious behavior by consumers and providers, accentuating economic and social interregional disparities [38].
The inequalities across Italy in children's health status are not only related to inequalities in treatment, but also in health prevention. The vaccine uptake rate for measles by a child's second birthday, for example, ranges from 54.9% in Calabria to 89.6% in Tuscany [12]. Italy continues to have the lowest coverage rate for measles among European countries; a national campaign was recently launched to increase the coverage to an expected rate of 90%–95% [13].
Geographical Variations in Social and Educational Opportunities
Regional inequalities in the provision of social and educational services can have a profound effect on the welfare of Italian children. There is evidence to suggest that the south of Italy has inadequate public services for its needs. For example, Campania and Sicily, the regions in the south with the highest birth rates in the country (11.5 and 10.4 births per 1,000 inhabitants, respectively), have the lowest national rate of access to nursery school (2.2% and 4.7%, for 0- to 2-year-old children). Both of these regions, along with Calabria and Puglia (all in southern Italy), show a rate of primary school abandonment (i.e., children abandon school completely) that is 2.5 times higher than in the Friuli Venezia Giulia region (about 24% versus 9%) [14].
There are also regional variations in the youth unemployment rate (i.e., the rate for those aged 15–19 years). The rate ranges from 65.2% and 61.4% for Calabria and Sicily, respectively, to 7.1% for Trentino Alto Adige.
The Italian Welfare State
The Italian welfare state is a mix of (1) occupationally fragmented social security systems, especially with regard to pensions (such as the German-style democratic social policy) and (2) social security systems that are based on universal coverage, as in the field of health-care provision (such as the British-style liberal social policy, based on universal social insurance schemes) [39]. Italy, thus, has a mixed model of social and occupational provision. The Italian health-care and social service sectors are, moreover, highly reliant on the support of nonprofit organizations.
The state is also charged with providing financial support for low-income workers and different forms of housing assistance. Unemployment benefits are provided by the government in the form of cash transfers based on contributions. Old-age pension is insurance-based and proportional to contributions. Education for a child age 10–15 years old is free but vocational training for those who are older is not free. The state is responsible for secondary schooling and university education [40]. The rapid increase in the elderly population, with the decrease of the overall adult population that provides health care, shows the growing need for a reorganization of the whole economic and welfare systems [41].
Over the past 6–7 years, Italy has been steadily sliding down the ranking in the United Nations Development Programme's Human Development Index (21st place in 2002), whereas it remains in 11th place in the Human Poverty Index. With the Berlusconi government (the head of which is the richest Italian, who also controls around 75% of the media and 90% of national broadcasting), decades of social achievements have been lost because of structural reforms and privatizations. These reforms benefit only a small proportion of the population, with most Italians seeing a decline in their quality of life, an increase in inequalities, and re-establishment of privileges for the few [42].
Children in southern Italy, therefore, face a constellation of risks—a high rate of school dropout, a low youth employment rate, and a higher likelihood of living in difficult family circumstances (involving factors such as low family educational level, family poverty, and families doing illegal work). Children in the north are much less likely to experience this set of risks [15].
The Ethics of Child Poverty
Inequalities in society raise fundamental ethical questions. In particular, such inequalities challenge us to take actions to improve health, based upon an ethical framework in which human rights and dignity are taken into consideration [16]. Inequality is based on the deprivation of rights such as education, work, and access to social services; inequality also means loss of human dignity, which is in itself linked to poverty [17]. Well-being is not only affected by money and economic status (i.e., gross domestic product and income), but also by social- and health-related rights and opportunities [18]. Children's rights to health must always be a priority, not only between but also within countries, and not only in developing but also in developed countries, until inequalities are overcome [19].
Policies to Improve Child Health in Southern Italy
Even though its effects have been questioned by the government, income inequality undoubtedly affects health outcomes, even in Italy, where health care and education are mostly public and free, and the unemployed receive social benefits [20]. The reduction of income inequality not only requires policies promoting economic development but also policies minimizing material deprivation (education, health services, transportation, environmental controls, availability of good quality food, quality of housing, and occupational health regulations). In addition, social inequalities may cause negative feelings (stress, shame, and distrust), which translates into poorer health by way of psycho-neuro-endocrine-immunological mechanisms and stress-induced behaviors such as smoking and overeating [21–23].
Initiatives to meet the identified needs should be carefully planned and overseen to ensure their successful and sustainable implementation. Tackling health inequalities requires a working partnership between government, social administration and organizations, the private sector, and civil society, at the national, regional, and local levels [24]. Equity is a permanent ethical and socioeconomic goal that requires adequate capacity, sound policies for improvement, and enough financial investment to ensure (or at least try to ensure) success. Intention to change is obviously not enough if egalitarian values are not widespread and felt by an entire nation, and if the social, cultural, and political causes of health inequalities are not addressed [25]. If dignity is not central to policy change, and rights are neglected, children will remain one of the most marginalized groups in society [26,27].
What are the particular prerequisites for addressing inequities in children's health? Strong and effective programs for improving child equity must be based on adequate, targeted financial, human, and service resource allocations. A starting point for planning such programs must be the systematic, ongoing evaluation of child health. This evaluation should not only take into consideration national averages, but also analyses stratified by socioeconomic and geographic categories, according to worldwide recognized indicators and methods [5]. Such analyses must be performed, and the data used, by national and regional governments to establish and monitor objectives related to health status and health-service use in disadvantaged population groups.
Another prerequisite is the creation of information, which is a powerful and influential tool and is, therefore, essential for improving accountability in a society. Moreover, reduction of child health inequalities can be pursued by promoting useful and adequate information, aimed not only at policymakers and program managers, but also at regional and local communities [28].
Unfortunately, the latest Italian National Health Programme (2003–2005; http://www.ministerosalute.it/psn/psnHome.jsp) ignores the unacceptably wide regional differences in child health care. There is no mention of improving the availability of health-related structures or the quality of assistance in perinatal care or childhood hospitalization. Similarly, the new public-school reform (http://www.istruzione.it/normativa/2004/schemadec210504.shtml) only mentions general principles concerning rights and duties, and ignores the dramatic school abandonment rate in southern Italy. It is, therefore, apparent that the “business as usual” approach, based on national averages, instead of regional and local needs, and supported by inadequate economic resources, can only increase the inequity gap instead of closing it.
Conclusion
The European experience of the development, implementation, and evaluation of policies and interventions to reduce health inequalities has been well-described [29–34]. For example, the European Network on Interventions and Policies to Reduce Inequalities in Health has recently analyzed policy developments concerning health inequalities in different European countries between 1990 and 2001 [29]. Compared to the rest of Europe, Italy's approach to ensuring the future health of its children is not encouraging. In a country with an aging population and an increasing gap between regions, the implementation of public-health programs focusing on promoting, monitoring, and improving children's well-being should be taken on as a recognized challenge and should be a political commitment. Furthermore, international and intracountry exchanges focused on child health can help enhance and speed up learning.
Citation: Bonati M, Campi R (2005) What can we do to improve child health in southern Italy? PLoS Med 2(9): e250.
Abbreviation
NHSItalian National Health Service
==== Refs
References
Italiana C Zancan FE Vuoti a perdere Rapporto 2004 su esclusione sociale e cittadinanza incompiuta 2004 Milano Feltrinelli 322
Istituto Nazionale di Statistica La povertà relativa in Italia nel 2003 2005 Available: http://www.istat.it . Accessed 8 July 2005
Bonati M Campi R Nascere e crescere oggi in Italia 2005 Roma Il Pensiero Scientifico Editore 191
Eurostat The social situation in the European Union 2004 Available: http://europa.eu.int/comm/employment_social/news/2004/oct/socsit_2004_en.pdf . Accessed 8 July 2005
United Nations International Children's Emergency Fund Child poverty in rich nations: Innocenti report card no. 1 2000 Florence United Nations International Children's Emergency Fund 28
United Nations International Children's Emergency Fund Child poverty in rich countries: Innocenti report card no. 6 2005 Florence United Nations International Children's Emergency Fund 36
Macinko JA Starfield B Annotated bibliography on equity in health, 1980–2001 Int J Equity Health 2002 1 1 12234390
Sachs JD McArthur JW The Millenium Project: A plan for meeting the Millenium Development Goals Lancet 2005 365 347 353 15664232
Bonati M Bosetti C Ferraro R Miglio D Sanvito E The southern Italy neonatal intensive care units network: Very low birth-weight outcomes for 1994 J Epidemiol Biostat 1997 2 171 178
Campi R Bonati M Diseguaglianza di mortalità infantile in Italia Prospettive Sociali Sanitarie 2000 219 6 9
Ministero della Salute Le caratteristiche dell ospedalizzazione pediatrica in Italia 2001 Available: http://www.ministerosalute.it/resources/static/primopiano/206/documento.pdf . Accessed 15 July 2005
Ciofi Degli Atti ML Rota MC Bella A Salmaso S ICONA Study Group Do changes in policy affect vaccine coverage levels? Results of a national study to evaluate childhood vaccination coverage and reasons for missed vaccination in Italy Vaccine 2004 22 4351 4357 15474728
Ministero della Salute Campagna straordinaria MRP 2005 Available: http://www.ministerosalute.it/servizio/campagna.jsp?idarc=9 . Accessed 15 July 2005
Istituto Nazionale di Statistica Rapporto annuale. 2004: Dati statistici 2004 Rome Istituto Nazionale di Statistica Available: http://www.istat.it/dati/catalogo/rapporto2004/dati.html . Accessed 15 July 2005
Ministero dell'Istruzione dell'Università e della Ricerca Indagine campionaria sulla dispersione scolastica nelle scuole statali elementari e medie Anno scolastico 2002–2003 2004 Available: http://www.istruzione.it/mpi/pubblicazioni/2004/dispersione04.shtml . Accessed 15 July 2005
Horton R Rediscovering human dignity Lancet 2004 364 1081 1085 15380969
Schuftan C Poverty and inequity in the era of globalization: Our need to change and to re-conceptualize Int J Equity Health 2003 2 1 7 12605720
Ruger JP Ethics of the social determinants of health Lancet 2004 364 1092 1097 15380971
Waterston T Inequity in child health as a global issue Pediatrics 2003 112 739 741 12949338
Caiazzo A Cardano M Cois E Costa G Marinacci C [Inequalities in health in Italy] Epidemiol Prev 2004 28 Suppl 3 1 161
De Vogli R Ministry R Gnesotto R Cornia GA Has the relation between income inequality and life expectancy disappeared? Evidence from Italy and top industrialised countries J Epidemiol Community Health 2005 59 158 162 15650149
Lynch JW Smith GD Kaplan GA House JS Income inequality and mortality: Importance to health of individual income, psychosocial environment, or material conditions BMJ 2000 320 1200 1204 10784551
Costello A White H Reducing global inequalities in child health Arch Dis Child 2001 84 98 102 11159279
Oliver A Healey A Le Grand J Addressing health inequalities Lancet 2002 360 565 567 12241675
Berlinguer G Bioethics, health, and inequality Lancet 2004 364 1086 1091 15380970
Spencer N Social, economic, and political determinants of child health Pediatrics 2003 112 704 706 12949325
Wagstaff A Child health on a dollar a day: Some tentative cross-country comparisons Soc Sci Med 2003 57 1529 1538 12948564
Zurriaga O Martinez-Beneito MA Abellan JJ Carda C Assessing the social class of children from parental information to study possible social inequalities in health outcomes Ann Epidemiol 2004 14 378 384 15246325
Mackenbach JP Bakker MJ Tackling socioeconomic inequalities in health: Analysis of European experiences Lancet 2003 362 1409 1414 European Network on Interventions and Policies to Reduce Inequalities in Health 14585645
Doran T Drever F Whitehead M Is there a north-south divide in social class inequalities in health in Great Britain? Cross sectional study using data the 2001 census BMJ 2005 328 1043 1045
Graham H Power C Childhood disadvantage and health inequalities: A framework for policy based on life course research Child Care Health Dev 2004 30 671 678 15527477
Waterston T Alperstein G Stewart Brown S Social capital: A key factor in child health inequalities Arch Dis Child 2004 89 456 459 15102639
West P Sweeting H Evidence on equalisation in health in youth from the West of Scotland Soc Sci Med 2004 59 13 27 15087139
Garcia-Gil C Cruz-Rojo C Alvarez-Giron M Solano-Pares A Health inequalities in Seville, Spain: Use of indicators of social deprivation and mortality in small areas Public Health 2004 118 11 20 14643623
World Health Organization Statistics World Health Report 2000 2000 Available: http://www.who.int/whr/2000/en/index.html . Accessed 15 July 2005
Mossialos E Citizens' views on health care systems in the 15 Member States of the European Union Health Econ 1997 6 109 116 9158964
Donatini A Rico A Giuseppina D'Ambrosio M Lo Scalzo A Orzella L Health care systems in transition: Italy 2001 Copenhagen European Observatory on Health Care Systems Available: http://www.euro.who.int/document/e73096.pdf . Accessed 15 July 2005
Maio V Manzoli L The Italian health care system: WHO ranking versus public perception Pharmacol Ther 2002 27 301 308
Aspalter C Different worlds of welfare capitalism: Australia, the United States, the United Kingdom, Sweden, Germany, Italy, Hong Kong and Singapore 2001 Canberra (Australia) Australian National University Asia Pacific School of Economics and Government Available: http://apseg.anu.edu.au/degrees/pogo/discussion_papers_previous/No80Aspalter.pdf . Accessed 8 July 2005
Niero M George V Taylor-Gooby P Italy: Right turn for the welfare state 1996 London MacMillan 117 135 European welfare policy: Squaring the welfare circle
Golini A Basso S Reynaud C Population aging in Italy: A challenge for our country, a workshop for the world G Gerontol 2003 51 528 544
Messina A Siniscalchi S Nardi J Sliding into insecurity 2004 Montevideo (Uruguay) Social Watch 170 171 In: Social Watch annual report. 2004: Fear and want—Obstacles to human security. Available http://www.socialwatch.org/en/informeImpreso/pdfs/italy2004_eng.pdf . Accessed 15 July 2005
|
16104833
|
PMC1188255
|
CC BY
|
2021-01-05 11:13:39
|
no
|
PLoS Med. 2005 Sep 23; 2(9):e250
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020250
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 1610483410.1371/journal.pmed.0020301EssayOtherEpidemiology/Public HealthToxicology/Environmental HealthMedical consequences of war/conflictPublic HealthNuclear Weapons 60 Years On: Still a Global Public Health Threat EssayMacDonald Rhona Rhona MacDonald is a medical doctor, freelance medical editor and journalist, and charity worker. She is currently the volunteer Section Editor of PLoS Medicine's Student Forum and works as a volunteer with many different charities including Oxfam, Médecins Sans Frontières, the Drugs for Neglected Diseases Initiative, and Doctors For Iraq. E-mail: [email protected]
Competing Interests: The author has declared that no competing interests exist.
11 2005 23 8 2005 2 11 e301Copyright: © 2005 Rhona MacDonald.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.This year marks the 60-year anniversary of the bombings of Hiroshima and Nagasaki. Has the world learned anything about the threat of nuclear weapons? What role can the health community play in reducing it?
This year marks the 60-year anniversary of the bombings of Hiroshima and Nagasaki. Has the world learned anything about the threat of nuclear weapons? What role can the health community play in reducing it?
==== Body
The United States carried out the world's first nuclear test, codenamed “Trinity,” on 16 July 1945 in the desert of New Mexico. Just three weeks later, on 6 August, the US exploded a uranium device called “Little Boy” 2,000 feet above the Japanese city of Hiroshima, killing around 150,000 people. Three days later, the US deployed a second nuclear bomb, a plutonium device called “Fat Man,” that exploded above the Japanese city of Nagasaki, resulting in at least 74,000 deaths. These two terrible acts heralded the start of the nuclear age, which reached its peak during the Cold War.
Sixteen years after the fall of the Berlin wall, symbolising the end of the Cold War, it is easy to forget the terror that gripped the world over the threat of a nuclear war that could destroy the planet. But the state of the world at the 60-year anniversary of the bombing of Hiroshima should be enough to shake us from our complacency, especially considering the number of countries that have nuclear weapons and haven't signed or ratified the relevant treaties, and the outcome of the latest review of the Non-Proliferation Treaty (NPT) earlier this year (see below).
Who Has Nuclear Weapons?
Russia, the US, the United Kingdom, France, and China are the only “declared” nuclear states; that is, they have declared that they have nuclear weapons in the NPT. But there are also “undeclared” states. Table 1 shows the countries that have nuclear weapons and how many they have. Box 1 shows the current status of these countries and in what situation they have indicated they would be prepared to use their nuclear weapons.
Box 1. Current Status of Nuclear States
UK
One of the four Trident submarines is on patrol at all times.
The missiles are not targeted (that is they are not aimed at a specific target) and are normally at several days notice to fire.
Accepts a first-use policy. That means that, in certain circumstances, it is prepared to use nuclear weapons first. Under the Conservative Government there was a policy of using nuclear weapons to protect Britain's ‘vital interests’. That policy has never been changed. In the build-up to the Iraq war in 2003, Defence Secretary Geoff Hoon said that the UK would use nuclear weapons if its troops were attacked with chemical or biological weapons.
The UK opposes a time-bound framework for disarmament. ‘Time-bound’ means putting deadlines on when agreed things have to be achieved. It also voted against multilateral negotiations proposed at the UN General Assembly.
France
One submarine is on patrol at all times.
Although policy is vague, France has never supported ‘no first-use’ and has said that it would use nuclear weapons to defend its ‘vital interests’.
Opposes a time-bound framework and a multilaterally negotiated nuclear disarmament convention.
US
At least ten submarines are on patrol at all times.
The Nuclear Posture Review of 2002 gave examples of when the US would use nuclear weapons first.
Opposes multilateral negotiations
Opposes including nuclear weapons on the International Court of Justice list of prohibited weapons.
Russia
Estimated that at least two submarines on patrol at all times.
Russia has a ‘no first-use’ policy.
Supports multilateral negotiations.
China
Has a ‘no first-use’ policy and has also said that it would not use nuclear weapons against a non-nuclear weapon state.
Supports a time-bound framework and has called for a convention banning nuclear weapons.
India
Has a ‘no first-use’ policy and has said that it would not use nuclear weapons against a non-nuclear weapon state.
Supports a nuclear weapons convention and sponsors a resolution at the UN on de-alerting nuclear weapons.
Pakistan
Has said that it would use nuclear weapons first in a conflict.
Israel
Will not confirm or deny having nuclear weapons.
Says that it will not be the first to introduce nuclear weapons into the Middle East but will not explain exactly what it means by that.
Opposes a time-bound framework.
Box text quoted from [5].
Table 1 Countries That Have Nuclear Weapons
Source: [7].
The Crucial Treaties
The NPT
The main objective of the NPT is to stop the spread, or “proliferation,” of nuclear weapons. The declared nuclear states had to agree not to pass on to other countries any nuclear weapons technology and, under Article VI, they also have to “pursue negotiations in good faith on effective measures relating to the cessation of the nuclear arms race at an early date and to nuclear disarmament….” Non-nuclear-weapon states had to promise not to make any attempt to acquire nuclear weapons. If they complied, in return they could get help to develop a nuclear power programme. Box 2 shows who has not signed the NPT.
Box 2. The NPT
The treaty opened for signature in 1968.
It entered into force in 1970.
A total of 188 countries have signed.
India, Pakistan, Israel, Cook Islands, and Niue have not signed.
The Democratic People's Republic of Korea (North Korea) announced its withdrawal from the NPT in January 2003.
Source: [2].
The US recently granted India access to its civilian nuclear knowledge in exchange for a “global partnership.” India is not a signatory of the NPT, so is not bound by its provisions, and it has always been American foreign policy, upheld by law, that only countries that are NPT members should share any benefits of American civilian nuclear expertise, so this is a worrying development [1].
Comprehensive Test Ban Treaty
After years of negotiations, the Comprehensive Test Ban Treaty was overwhelmingly endorsed in 1996 at the United Nations in New York. To date, it has been signed by 167 countries and ratified by 99. The Comprehensive Test Ban Treaty outlaws nuclear testing of any kind and must be signed and ratified by the 44 countries identified as having nuclear power plants or research reactors. Ten of those countries have signed but not ratified: Algeria, China, Colombia, Democratic Republic of Congo, Egypt, Indonesia, Iran, Israel, the US, and Viet Nam. Three have not signed it at all: India, Pakistan, and North Korea [2].
The Seventh Review of the NPT
There was some optimism after the sixth NPT review in May 2000, with the media reporting that a nuclear-free world was in sight. There were 13 points that all participant countries agreed to adhere to in time for the next convention (see Box 3). When the UK's Minister of Defence, Geoff Hoon, said a few months later, “The NPT agreement is an aspiration; it is not likely to produce results in the short term” [3], it was a sign of things to come.
Box 3. The 2000 NPT Review Conference Final Document
A ‘Programme of Action’ (often referred to as the ‘13 practical steps towards global nuclear disarmament’) became part of the Final Document. They are summarised as:
(1) Progress needs to be made on entry-into-force of the Comprehensive Test Ban Treaty (CTBT).
(2) The moratorium on nuclear weapon test explosions must be maintained.
(3) The Conference on Disarmament (CD) must move forward in establishing a Fissile Material Cut-Off Treaty (FMCT).
(4) A subsidiary body with a mandate to deal with nuclear disarmament is needed.
(5) The principle of irreversibility on arms control and reduction agreements must be applied to nuclear disarmament measures.
(6) Progress on nuclear disarmament (implementation of Article VI) is required.
(7) Implementation of arms reduction agreements and pursuit of binding agreements on further irreversible reductions must be instituted.
(8) Greater emphasis must be attached to the implementation of the Trilateral Initiative and greater support must be forthcoming for the International Atomic Energy Agency (IAEA).
(9) Confidence building measures and progressive steps to lower the nuclear threshold must be offered. [These include increased effort by the NWS to reduce their nuclear arsenals unilaterally; increased transparency by the NWS about their nuclear weapons capability; further reductions of non-strategic nuclear weapons; a reduction in the operational status of nuclear weapons (de-alerting); a diminished role for nuclear weapons in security policies (doctrines); and the engagement of all NWS in facilitating the elimination of nuclear weapons.]
(10) Further fissile material stocks must be put under IAEA Safeguards.
(11) The ultimate objective of complete nuclear disarmament must be reaffirmed.
(12) The formal reporting back by States Parties between Review Conferences— the accountability principle—must be instituted.
(13) Enhanced verification measures must be agreed and implemented.
Box text quoted from [6].
This May, after a month of arguments, the seventh NPT review ended in failure. According to Gunnar Westberg and John Loretz, Co-President and Program Director, respectively, of International Physicians for the Prevention of Nuclear War (IPPNW): “At the end of the day, the review collapsed over one issue: the refusal of the United States to build on the foundations for disarmament that were laid in 1995 and 2000, or even to acknowledge that those foundations exist” [4].
They continue: “The Bush administration may attempt to spin the meaning of the failed NPT review to suit its distaste for multilateral negotiations and for the UN as an institution. This would be akin to a teenager breaking the lawnmower and then telling his parents that he can't cut the lawn because the lawnmower doesn't work. One cannot deliberately break a consensus-based decision making process and then claim that multilateralism does not work” [4].
Michael Christ, Executive Director of IPPNW, told PLoS Medicine: “The failure of the 2005 NPT review exposed the underlying realities that stand as obstacles to achieving a world free of nuclear weapons: firstly, a stubborn refusal by the nuclear weapon states, particularly the US, to comply with their disarmament commitments and, conversely, an insistence that nuclear weapons are indispensable to their security and to the pursuit of their global interests; secondly, increasing levels of frustration and impatience among non-nuclear-weapons states, the overwhelming majority of which want nuclear disarmament; and thirdly, the increasingly unavoidable and dangerous contradiction between guaranteeing access to ‘peaceful’ uses of nuclear energy while at the same time ensuring that such uses do not become a platform for weapons development.”
Where Does That Leave the World in 2005?
In the current climate of increased global terrorism, the aftermath of the war in Iraq, and the uncertain situation in Iran and North Korea regarding nuclear weapons, where does the failure of the 2005 NPT review leave us?
Douglas Holdstock from the campaigning organisation Medact, said: “North Korea may have five to ten usable weapons. It is very unlikely to use them until directly attacked by the US. Iran will not have usable nuclear weapons for about five years.”
“The greatest risks of use,” he said, “are probably (1) India and Pakistan over Kashmir; (2) Israel against any nearby Islamic state, particularly if attacked by chemical or biological weapons; (3) India and China over border disputes; and (4) China and the US, for example, over Taiwan. Any of these could kill millions and cause widespread fallout, but probably not ‘nuclear winter’. This would only follow a US–Russia exchange, which is a low risk at present—but will remain as long as nuclear weapons exist.”
Michael Christ explained that about 5,000 US and Russian nuclear weapons are still on 24-hour hair trigger alert, ready to be launched at a moment's notice: “These are fallible machines being operated by fallible human beings—and we have had a number of frighteningly close calls with accidental nuclear war. Our [IPPNW's] calculations indicate that nearly 7 million Americans would die immediately from an accidental launch of weapons from a single Russian submarine. Furthermore, some of the nuclear powers, led by the US, are planning for a new generation of ‘useable’ battlefield nuclear weapons—‘bunker busters’ and ‘mini-nukes.’”
Gunnar Westberg thinks that having nuclear weapons is contagious: “If Russia and the USA say they need nuclear weapons for their security, of course smaller countries will feel the same, with stronger reasons. If the nuclear-weapons states do not abolish their arsenals, proliferation to many more countries cannot be stopped.”
The Campaign for Nuclear Disarmament argues that counter-proliferation methods are replacing the concept of non-proliferation. Ruth Tanner, Press Officer for the Campaign for Nuclear Disarmament UK, said: “The concept of non-proliferation, as enshrined in the NPT, is under threat from the drive by the US and UK towards a policy of counter-proliferation, rather than non-proliferation. Counter-proliferation policies also further undermine the multilateral non-proliferation regime through its possible substitution—as in the case of Iraq—by pre-emptive disarmament wars, carried out by a tiny minority of the international community.”
“Missile defence,” she said, “is clearly part of the counter-proliferation approach, for it enables first strike without fear of retaliation.”
Impacts on Health
“There are the enormous impacts on health and environment, documented in numerous studies, resulting from the development, manufacture, testing, stockpiling, maintenance, transport, dismantling, storage, and disposal of nuclear weapons,” said Michael Christ. “Every one of these steps poses direct risks to the health of the personnel involved and the general population. We [IPPNW] estimated 430,000 deaths worldwide from fatal cancer as a consequence of US atmospheric nuclear testing, from 1945 to 1963. Nuclear programs worldwide have left behind a toxic legacy that will affect human health and the environment for thousands of years. In the US alone, this folly cost taxpayers $5.5 trillion between 1940 and 1996. And spending is on the rise.”
He explained what IPPNW is doing to publicise the threat of nuclear weapons: “We are emphasizing the medical and moral imperative of nuclear disarmament. We must stigmatize nuclear weapons not on the basis of who owns them but for what they are and what they can do. These are not weapons at all—they are instruments of indiscriminate mass murder. They are Nazi crematoria mounted on missiles.”
The world's major health problems are all related, and are ultimately affected by how much money is spent on weapons, according to Douglas Holdstock: “Poverty, under-development, disease, [and] war, which [are] fuelled by the arms trade, climate change, and other environmental threats, such as over-population, are all inter-linked.” And reducing nuclear and other arms spending will free resources for better causes, he said.
What Can International Health-Care Workers Do?
Michael Christ reminds us what is at stake: “We are moving inexorably towards a major nuclear disaster of some form, and the medical dangers are just too profound to ignore for those concerned about and responsible for public health.”
He continued: “The heart of the problem is a lack of political will to rid the world of the only weapons that could extinguish most life on earth in a matter of hours. Creating that political will is our focus for the future.”
Douglas Holdstock said, “[Nobel Peace Laureate] Sir Joseph Rotblat says that to prevent nuclear war we must prevent all war, as the knowledge of how to make nuclear weapons will indeed always be with us. Rotblat ended his Nobel Prize acceptance speech by saying, ‘remember your humanity.’”
“At this difficult and dangerous time it is vital that we work for peace in the world,” said Ruth Tanner. “Nuclear weapons are a threat to the planet and its people and the rogue states that insist on maintaining their destructive arsenal are a minority in a world that wants to be free of nuclear weapons.”
“The NPT is still valid,” said Gunnar Westberg. “A strong international movement for a nuclear weapons convention, prohibiting nuclear weapons, is needed, and may be developing just now. It may work along the pattern of the Campaign to Ban Landmines.”
He continued: “We [physicians] are used to talking to people about questions of life and death. So we must tell the general public that nuclear weapons are the greatest threat to the survival of mankind, and the only intervention that will work is the complete abolition of all nuclear weapons. Now is the time to do this, in this period of low tension between the big powers, and before nuclear weapons proliferate to many more countries.”
“Nuclear weapons and mankind,” he added, “can in the long run not coexist. Either will be abolished. We have a choice.”
Conclusion
The world is in turmoil: terrorism, or at least the fear of terrorism, seems to have a stranglehold; world governments and the United Nations have an arbitrary way of dealing with “rogue states” (notice, for example, the differences in their treatment of Iraq, the Democratic Republic of Congo, Zimbabwe, and Myanmar); and international treaties can be broken on a whim, such as the recent deal in which the US agreed to share its civilian nuclear knowledge with India. Now is not the time to be gambling with the world's future and that of the human race by holding on to weapons that could destroy the planet thousands of times over. The countries that continue to have such weapons are potential destroyers, not the guardians of democracy, or the defenders of peace, or whatever they choose to call themselves. Democracy should be better than this.
Useful Links
International Physicians for the Prevention of Nuclear War:
www.ippnw.org
Medact:
www.medact.org
Campaign for Nuclear Disarmament:
www.cnduk.org
WMD Awareness Programme:
www.comeclean.org.uk
British American Security Information Council:
www.basicint.org
Citation: MacDonald R (2005) Nuclear weapons 60 years on: Still a global public health threat. PLoS Med 2(11): e301.
Abbreviations
IPPNWInternational Physicians for the Prevention of Nuclear War
NPTNon-Proliferation Treaty
NWSnuclear weapons states
==== Refs
References
[Anonymous] India and America: Now we are six The Economist 2005 July 23 13
WMD Awareness Programme Nuclear weapons: Treaties 2005 London WMD Awareness Programme Available: http://www.comeclean.org.uk/articles.php?articleID=21 . Accessed 1 August 2005
Campaign for Nuclear Disarmament Rejecting the logic of ‘counter-proliferation’: Disarmament is the key to global peace and security 2005 July London Campaign for Nuclear Disarmament Available: http://www.cnduk.org/pages/campaign/NPT05.pdf . Accessed 27 July 2005
Westburg G Loretz J A case of global medical malpractice 2005 June 21 Cambridge (Massachusetts) International Physicians for the Prevention of Nuclear War Available: http://www.ippnw.org/NPT2005WestbergLoretz.html . Accessed 27 July 2005
WMD Awareness Programme Nuclear weapons: Nuclear policy 2005 London WMD Awareness Programme Available: http://www.comeclean.org.uk/articles.php?articleID=25 . Accessed 1 August 2005
British American Security Information Council Oxford Research Group The Non-Proliferation Treaty Review Conference: Breakthrough or bust in '05? 2005 January London British American Security Information Council Available: http://www.basicint.org/nuclear/NPT/2005rc/nptoverview.htm . Accessed 27 July 2005
WMD Awareness Programme Nuclear weapons: Who has nuclear weapons? 2005 London WMD Awareness Programme Available: http://www.comeclean.org.uk/articles.php?articleID=22 . Accessed 1 August 2005
|
16104834
|
PMC1188256
|
CC BY
|
2021-01-05 10:39:21
|
no
|
PLoS Med. 2005 Nov 23; 2(11):e301
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020301
|
oa_comm
|
==== Front
PLoS MedPLoS MedpmedplosmedPLoS Medicine1549-12771549-1676Public Library of Science San Francisco, USA 10.1371/journal.pmed.0020315SynopsisInfectious DiseasesInfectious DiseasesMalariaTargeting of Endothelial Activation in Cerebral Malaria Synopsis9 2005 23 8 2005 2 9 e315Copyright: © 2005 Public Library of Science.2005This is an open-access article distributed under the terms of the Creative Commons Public Domain declaration, which stipulates that, once placed in the public domain, this work may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose.
Inhibition of Endothelial Activation: A New Way to Treat Cerebral Malaria?
==== Body
Malaria is one of the most serious of all tropical parasitic diseases: a severe and life-threatening form of which in humans is cerebral malaria, a complication that can occur in malaria caused by Plasmodium falciparum. This grave complication involves malarial infection of the red blood cells that accumulate within the very small capillaries that flow through the tissues of the brain. Even when treated, cerebral malaria has a fatality rate of 15% or more.
Numerous studies have pointed to a key role of tumor necrosis factor (TNF) and related proteins in the pathogenesis of cerebral malaria, and a clear relationship has been established between plasma concentrations of TNF and cerebral pathology. In experimental cerebral malaria, TNF-beta, now called lymphotoxin α (LT), has been shown to be a principal mediator of pathogenesis. LT and TNF belong to the same family, interact with a common receptor, and could act together during disease progression to effect the pathogenesis, according to recent evidence.
Now Georges Grau and colleagues describe the anti-inflammatory activity of a transcriptional inhibitor of TNF, called LMP-420, which might offer a new way for treating cerebral malaria. The aim of their study was to assess the ability of LMP-420 to inhibit the in vitro TNF and/or LT effects on brain endothelium, with particular attention to endothelial cell activation, adhesiveness for malarial parasites, and vesiculation.
Using an in vitro model of cerebral malaria based on human, brain-derived endothelial cells (HBEC-5i), they found that LMP-420 potently reduced endothelial activation, endothelial adhesiveness for P. falciparum–parasitized red blood cells, and endothelial MP release, three major features of cerebral malaria.
The results provide evidence for a dual inhibitory effect of LMP-420 on both TNF and LT in an in vitro model of cerebral malaria pathogenesis, when added either before or simultaneously with both cytokines, they said. LMP-420 also abolished the cytoadherence of ICAM-1-specific P. falciparum–parasitized red blood cells on these endothelial cells. Identical but weaker effects were observed when LMP-420 was added with LT. LMP-420 also caused a dramatic reduction of HBEC-5i vesiculation induced by TNF or LT stimulation.
Several molecules inhibiting TNF, such as monoclonal antibody to TNF or pentoxyfylline, have been tested in clinical trials of cerebral malaria but failed to improve disease outcome. These failures could be explained by the fact that LT was recently demonstrated to also have a crucial role in the pathogenesis of this cerebral syndrome, said the authors.
They conclude that the anti-inflammatory activity of LMP-420 might be useful in targeting the wide variety of diseases in which TNF and its related family members play a role, and could represent a novel, stable, and efficient therapeutic way to improve the outcome of patients with cerebral malaria.
Plasmodium-infected red blood cells bind to brain endothelial cells
However, the authors caution that the experimental in vitro results do not necessarily predict potential efficacy in either animal models or humans, especially since in their model the LMP-420 was administered before the disease process was established. Nevertheless, this avenue is a promising one to explore further.
|
0
|
PMC1188257
|
CC0
|
2021-01-05 10:40:32
|
no
|
PLoS Med. 2005 Sep 23; 2(9):e315
|
utf-8
|
PLoS Med
| 2,005 |
10.1371/journal.pmed.0020315
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 1611866510.1371/journal.pcbi.001002705-PLCB-RA-0061R2plcb-01-03-03Research ArticleBioinformatics - Computational BiologyEvolutionMicrobiologyEubacteriaArchaeaEukaryotesComparative Analyses of Fundamental Differences in Membrane Transport Capabilities in Prokaryotes and Eukaryotes Membrane Transport Systems in 141 OrganismsRen Qinghu Paulsen Ian T *The Institute for Genomic Research, Rockville, Maryland, United States of AmericaBork Peer EditorEMBL Heidelberg, Germany* To whom correspondence should be addressed. E-mail: [email protected] 2005 19 8 2005 1 3 e2724 3 2005 8 7 2005 Copyright: © 2005 Ren and Paulsen.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Whole-genome transporter analyses have been conducted on 141 organisms whose complete genome sequences are available. For each organism, the complete set of membrane transport systems was identified with predicted functions, and classified into protein families based on the transporter classification system. Organisms with larger genome sizes generally possessed a relatively greater number of transport systems. In prokaryotes and unicellular eukaryotes, the significant factor in the increase in transporter content with genome size was a greater diversity of transporter types. In contrast, in multicellular eukaryotes, greater number of paralogs in specific transporter families was the more important factor in the increase in transporter content with genome size. Both eukaryotic and prokaryotic intracellular pathogens and endosymbionts exhibited markedly limited transport capabilities. Hierarchical clustering of phylogenetic profiles of transporter families, derived from the presence or absence of a certain transporter family, showed that clustering patterns of organisms were correlated to both their evolutionary history and their overall physiology and lifestyles.
Synopsis
Membrane transporters are the cell's equivalent of delivery vehicles, garbage disposals, and communication systems—proteins that negotiate through cell membranes to deliver essential nutrients, eject waste products, and help the cell sense environmental conditions around it. Membrane transport systems play crucial roles in fundamental cellular processes of all organisms. The suite of transporters in any one organism also sheds light on its lifestyle and physiology. Up to now, analysis of membrane transporters has been limited mainly to the examination of transporter genes of individual organisms. But advances in genome sequencing have now made it possible for scientists to compare transport and other essential cellular processes across a range of organisms in all three domains of life.
Ren and Paulsen present the first comprehensive bioinformatic analysis of the predicted membrane transporter content of 141 different prokaryotic and eukaryotic organisms. The scientists developed a new computational application of the phylogenetic profiling approach to cluster together organisms that appear to have similar suites of transporters. For example, a group of obligate intracellular pathogens and endosymbionts possess only limited transporter systems in spite of the massive metabolite fluxes one would expect between the symbionts and their host. This is likely due to the relatively static nature of their intracellular environment. In contrast, a cluster of plant/soil-associated microbes encode a robust array of transporters, reflecting the organisms' versatility as well as their exposure to a wide range of different substrates in their natural environment.
Citation:Ren Q, Paulsen IT (2005) Comparative analyses of fundamental differences in membrane transport capabilities in prokaryotes and eukaryotes. PLoS Comp Biol 1(3): e27.
==== Body
Introduction
Membrane transport systems play essential roles in cellular metabolism and activities. Transporters function in the acquisition of organic nutrients, maintenance of ion homeostasis, extrusion of toxic and waste compounds, environmental sensing and cell communication, and other important cellular functions [1]. Various transport systems differ in their putative membrane topology, energy coupling mechanisms, and substrate specificities [2]. Among the prevailing energy sources are adenosine triphosphate (ATP), phosphoenolpyruvate, and chemiosmotic energy in the form of sodium ion or proton electrochemical gradients.
The transporter classification system (http://www.tcdb.org/) represents a systematic approach to classify transport systems according to their mode of transport, energy coupling mechanism, molecular phylogeny, and substrate specificity [2–5]. Transport mode and energy coupling mechanism serve as the primary basis for classification because of their relatively stable characteristics. There are four major classes of solute transporters in the transporter classification system: channels, primary (active) transporters, secondary transporters, and group translocators. Transporters of unknown mechanism or function are included as a distinct class. Channels are energy-independent transporters that transport water, specific types of ions, or hydrophilic small molecules down a concentration or electrical gradient; they have higher rates of transport and lower stereospecificity than the other transporter classes (e.g., Escherichia coli GlpF glycerol channel [6]). Primary active transporters (e.g., Lactococcus lactis LmrP multidrug efflux pump [7]) couple the transport process to a primary source of energy (ATP hydrolysis). Secondary transporters utilize an ion or solute electrochemical gradient, e.g., proton/sodium motive force, to drive the transport process. E. coli LacY lactose permease [8,9] is probably one of the best characterized secondary transporters [10]. Group translocators modify their substrates during the transport process. For example, E. coli MtlA mannitol PTS transporter phosphorylates exogenous mannitol using phosphoenolpyruvate as the phosphoryl donor and energy source and releases the phosphate ester, mannitol-1-P, into the cell cytoplasm [11,12]. Each transporter class is further classified into individual families and subfamilies according to their function, phylogeny, and/or substrate specificity [3].
Since the advent of genomic sequencing technologies, the complete sequences of over 200 prokaryotic and eukaryotic genomes have been published to date, representing a wide range of species from archaea to human. There are also more than 1,100 additional genome sequencing projects currently underway around the world (Gold Genomes Online Database, http://www.genomesonline.org/) [13,14]. Convenient and effective computational methods are required to handle and analyze the immense amount of data generated by the whole-genome sequencing projects. An in-depth look at transport proteins is vital to the understanding of the metabolic capability of sequenced organisms. However, it is often problematic to annotate these transport proteins by current primary annotation methods because of the occurrence of large and complex transporter gene families, such as the ATP-binding cassette (ABC) superfamily [15,16] and the major facilitator superfamily (MFS) [17,18], and the presence of multiple transporter gene paralogs in many organisms. We have been working on a systematic genome-wide analysis of cellular membrane transport systems. Previously, we reported a comprehensive analysis of the transport systems in 18 prokaryotic organisms [19,20] and in yeast [21]. Here we expand our analyses to 141 species and compare the fundamental differences in membrane transport systems in prokaryotes and eukaryotes. Phylogenetic profiling of transporter families and predicted substrates was utilized to investigate the relevance of transport capabilities to the overall physiology of prokaryotes and eukaryotes.
Results/Discussion
Numbers of Recognized Transporter Families and Proteins
A total of 40,678 transport proteins from 141 species (Table S1), including 115 Eubacteria, 17 Archaea, and 9 Eukaryota, were predicted by our analysis pipeline. They were classified into 134 families, including 7 families of primary transporters, 80 families of secondary transporters, 32 channel protein families, 2 phosphotransferase systems (PTSs), and 13 unclassified families. Some of these families are very large superfamilies with numerous members, such as the ABC superfamily and MFS, both of which are widely distributed in Eubacteria, Archaea, and Eukaryota. Some are small families with only a single or a few members. The distribution of transporter families varies significantly across the three domains of life (Figure 1). There are 42 eukaryotic-specific families, mostly ion channel families that exist exclusively in multicellular eukaryotic organisms like Drosophila melanogaster,
Arabidopsis thaliana, and humans. These channels are involved in processes like cell communication, signal transduction, and maintenance of internal homeostasis in a multicellular environment. Most of these families are restricted to a single organismal type. Many of them may have arisen later during evolution, after the separation of the three domains. Alternatively, some families may have diverged too extensively from their prokaryotic counterparts to be recognized as homologs. Interestingly, a bacterial homolog to the previously described “eukaryotic-specific” glutamate-gated ion channel (GIC) family of neurotransmitter receptors has now been characterized in Synechocystis [22,23], and its orthologs have been identified in other sequenced Cyanobacteria. The Synechocystis transporter binds glutamate and forms a K+-selective ion channel. These observations suggest that eukaryotic GIC family transporters arose from a primordial prokaryotic counterpart.
Figure 1 Venn Diagram Showing the Distribution of Transporter Families across the Three Domains of Life
There are 38 prokaryotic-specific transporter families, of which 22 families exist exclusively in Eubacteria, such as the bacterial sugar PTS systems (see below), and 16 are shared by Eubacteria and Archaea. In contrast to eukaryotic-specific families, which are usually limited to single species, the majority of prokaryotic-specific ones are broadly distributed among prokaryotes. There are no Archaea-specific transporter families currently known. Due to the very limited experimental characterization of Archaea species relative to Eubacteria and Eukaryota, many aspects of the physiology and biochemistry of Archaea are poorly understood [24]. We compared the annotation of membrane proteins in selected species of Archaea and Eubacteria in The Institute for Genomic Research's Comprehensive Microbial Resource database [25]. The percentage of the membrane proteins assigned to the role category of “hypothetical proteins” is significantly greater in Archaea than in Eubacteria (Figure S1). These observations suggest that the sparse functional characterization could be the primary reason for the lack of any known Archaea-specific transporter families.
There are 41 transporter families represented in all three domains of life, highlighting the fundamental importance of these families. These are presumably very ancient families shared by the last common ancestor of Archaea, Eukaryota, and Eubacteria. Most of them were found within the secondary transporter class. These ubiquitous transporter families function in the transport of a diverse spectrum of substrates, including sugars, amino acids, carboxylates, nucleosides, and various cations and anions. There are 14 families shared by Eubacteria and Eukaryota and 16 shared by Eubacteria and Archaea. Some of these families shared only in two domains may ultimately be discovered in all three domains once a greater diversity of organisms is sequenced.
The overall quantity of recognized transport proteins (Figure 2A) and the percentage relative to the total number of open reading frames (ORFs) (Figure 2B) were compared for the organisms analyzed. Between 2% and 16% of ORFs in prokaryotic and eukaryotic genomes were predicted to encode membrane transport proteins, emphasizing the importance of transporters in the lifestyles of all species. In general, eukaryotic species, especially multicellular eukaryotic organisms, exhibit the largest total number of transport proteins, e.g., Drosophila (682 transport proteins, 3.7% of ORFs), Arabidopsis (882, 3.5%), Caenorhabditis elegans (669, 4.1%), and humans (841, 3.0%). However, the transport proteins of eukaryotic species account for a relatively smaller percentage of total ORFs than in Eubacteria (average 9.3% ± 2.9%) and Archaea (average 6.7% ± 2.3%) species. Considerable variations in the quantity of transport proteins have been observed among species belonging to the same phylogenetic group. For example, α-Proteobacteria species exhibit a wide variety of lifestyles and corresponding differences in transporter content; they range from rhizosphere-dwelling organisms such as Mesorhizobium loti and Sinorhizobium meliloti [26] with 883 (12.1%) and 826 (13.3%) transport proteins each, to obligate intracellular pathogens or symbionts such as Rickettsia prowazekii and Wolbachia sp. with 57 (6.8%) and 65 (5.4%) transport proteins, respectively. Overall, prokaryotic obligate endosymbionts and intracellular pathogens, as well as the eukaryotic intracellular parasites (Plasmodium falciparum [27] and Encephalitozoon cuniculi [28]), possess the most limited repertoire of membrane transporters.
Figure 2 Numbers of Recognized Transport Proteins and Percentage of Total ORFs
The overall numbers of recognized transport proteins (A) and percentage of total ORFs encoding transport proteins (B) were compared for the 141 organisms analyzed. Species from distinct phylogenetic groups are labeled with different colors. The prokaryotic and eukaryotic obligate intracellular parasites/pathogens are marked with red stars.
Genome Size versus Diversity of Transporter Families and Numbers of Paralogs
Organisms with a larger genome size and therefore more ORFs generally encode a greater number of transporters [19,29]. In addition to transporters, regulatory genes, secondary metabolism genes, and transcription factors, also appear to increase with genome size [29–31]. Two major factors could contribute to the expansion of transporters in organisms with large genome sizes: (1) an increased number of distinct transporter families, and (2) a higher degree of gene duplication or expansion, leading to a greater number of paralogs in certain transporter families. To investigate the relationship between genome size and these two factors, we plotted the total number of ORFs from 141 organisms as a function of either the number of distinct transporter families (Figure 3A), or the average number of paralogs per family (Figure 3B). Prokaryotes and eukaryotes exhibit distinct differences. For prokaryotic species, there is a relatively linear relationship between the genome size and the number of transporter families (R
2 = 0.54) or average number of paralogs (R
2 = 0.65). As genome size increases, the rate of increase in the number of families per organism is approximately eight times greater than that of the average number of paralogs per family. The increase in genome size can only partially explain the expansion of transporter families and paralogs (as indicated by the correlation R
2 value). The strain-specific properties and lifestyles could also have an impact. For example, a group of α-Proteobacteria exhibit the most paralogs per family but have relatively lower diversity of transporter families. These organisms include rhizobial microsymbionts M. loti,
S. meliloti, and Bradyrhizobium japonicum [26], and a closely related plant pathogen, Agrobacterium tumefaciens (enclosed by a circle on Figure 3). All of these organisms have more ABC transporters than any other sequenced organisms [29]. ABC family transporters mediate the uptake of a variety of nutrients and the extrusion of drugs and metabolite wastes. Having a large complement of high-affinity ABC uptake systems may be an advantage for organisms in the competition among microbes for nutrients. Two Streptomyces species, St. avermitilis and St. coelicolor, also exhibit a similar trend, with a significant expansion of the ABC and MFS family transporters.
Figure 3 Number of Total ORFs versus Number of Distinct Transporter Families or Average Number of Paralogs per Family
The number of total ORFs in the genome for each of the 141 sequenced prokaryotic and eukaryotic organisms (x-axis) was plotted as a function of either the number of distinct transporter families (A) or the average number of paralogs per family (B) (y-axis). Blue diamonds represent prokaryotic organisms and red squares represent eukaryotic organisms. Trend line and power correlation R
2 value are shown for prokaryotes and eukaryotes, respectively. A group of α-Proteobacteria are enclosed by a circle (see text for discussion).
The number of eukaryotic species analyzed is smaller, so it is more difficult to draw robust conclusions. The single-celled eukaryotes such as the yeasts appear to display characteristics similar to those of the prokaryotes, showing expansions in both transporter families and paralogs as genome size increases, with the former being a more important factor. However, in multicellular eukaryotic organisms such as animals and plants, the tremendous number of paralogs in certain transporter families accounts for a significant portion of the increase of transporters. Although multicellular eukaryotes exhibit fewer transporter families than some of the prokaryotic species, they have generated an extraordinary number of paralogs by gene duplication or expansion within certain families, like the ABC superfamily, MFS, and the voltage-gated ion channel superfamily. For example, the Arabidopsis genome encodes 110 paralogs of the ABC superfamily [32,33] and 92 paralogs of the MFS.
These differences in the relative abundances of transporter paralogs and distinct transporter families probably represent fundamental differences in transporter needs or priorities of these organisms. Multicellular organisms with many apparently redundant transporter paralogs appear to be utilizing a strategy of specialization. Many of their closely related paralogous transporters are presumably expressed only in specific tissues or subcellular localizations, or at specific developmental time points. Many appear to be involved in cell–cell communication and signal transduction processes, emphasizing the importance of intercellular communication in complex multicellular organisms. In contrast, the single-celled prokaryotes and eukaryotes, with relatively fewer paralogs but a greater emphasis on numbers of different families of transporters, appear to be utilizing a strategy of diversification. This probably reflects that one of the primary roles of membrane transport systems in these organisms is nutrient acquisition. A greater diversity of transporter types presumably allows for a broader range of substrate utilization.
Distribution of Transporter Types According to Energy Coupling Mechanism
A wide range of variations were observed in the relative usage of energy coupling mechanisms to drive transport processes among the prokaryotes and eukaryotes analyzed. Table 1 shows the relative percentage of each transporter type in organisms from major phylogenetic groups. Transporters were categorized into five major types according to transport mode and energy coupling mechanism: primary transporters, secondary transporters, ion channels, group translocators, and unclassified. Primary and secondary carriers are ubiquitous, being present in all organisms analyzed. However, their percentage among the total transporters varies greatly (12%–78% for primary carriers and 17%–80% for secondary carriers). In prokaryotic and unicellular eukaryotic systems, primary and secondary carriers are the predominant types of transporters, together contributing more than 90% of the total transporters. Channel proteins make up a greater percentage (12%–43%) in higher eukaryotic organisms.
Table 1 The Relative Percentage of Each Transporter Type within Major Phylogenetic Groups
aNumber of organisms analyzed is indicated in the parenthesis.
NF, not found.
Compared to eukaryotes, prokaryotic organisms rely heavily on primary active transporters, largely because of the usage of ABC uptake systems that are absent in eukaryotes [34]. Organisms with the highest percentage of primary transporters generally belong to one of the three groups. (1) The first group includes organisms that lack a citrate cycle and an electron transfer chain, and therefore can only generate a proton motive force by indirect methods such as substrate-level phosphorylation followed by ATP hydrolysis. These organisms include Mycoplasma spp., spirochetes, Streptococcus spp., Tropheryma whipplei,
Mycobacterium leprae, Thermoanaerobacter tengcongensis, and Thermotoga maritime. ATP is their primary source of energy, and therefore is most frequently used to drive nutrient uptake and maintain ion homeostasis. (2) The second group includes photosynthetic organisms with the ability to synthesize an ATP pool via photosynthesis, including Synechocystis sp., Nostoc sp., and Thermosynechococcus elongates. (3) The third group is a group of α-Proteobacteria that possess a significant expansion of the ABC superfamily [29], including soil/plant-associated bacteria, such as M. loti [26], S. meliloti [26], A. tumefaciens, and related human/animal pathogens such as Brucella suis. Unlike the first two groups, in which the usage of primary transporters seems to be predicated on bioenergetic constraints, the expansion of the ABC transporter family in these α-Proteobacteria does not have any obvious energetic explanations. Instead, it may reflect an organismal requirement for high-affinity transport since ABC transporters typically show higher substrate affinities than most secondary transporters.
The PTS is only present in a subset of Eubacteria, while completely lacking in Archaea and Eukaryota. Gram-negative enteric bacteria, such as E. coli,
Shigella flexneri, and Salmonella typhimurium, as well as Gram-positive species associated with the human gastrointestinal tract, like Listeria monocytogenes and Lactobacillus plantarum, encode the most abundant PTS systems. Owing to the absorption capacity and efficiency of the intestine, these species have to compete with hundreds of other types of bacteria in an environment containing only small amounts of free carbohydrates or other easily absorbable forms of nutrients. The enrichment of sugar PTS systems in these species could be an advantage to thrive in their ecological niches.
Channel proteins contribute a relatively smaller percentage of transporters in the prokaryotic species we analyzed, and their functions in vivo are largely unknown. Nine organisms lack recognizable channels, including Chlamydia spp., T. whipplei,
Treponema pallidum,
Wolbachia sp., and R.
prowazekii, all of which are obligate intracellular pathogens/symboints. All other prokaryotic species, including all extremophiles sequenced to date, encode channel proteins, suggesting these channels could function in responding promptly to osmotic and other environmental stresses [35]. Intracellular pathogens and endosymbionts may not need water or ion channels because of their relatively static intracellular environment and may largely depend on their host organisms for maintenance of ion homeostasis.
The percentage of channel proteins increases significantly in multicellular eukaryotes. In animals, these consist largely of ion channels with communication roles, such as in signal transduction, or roles as sensors for external stimuli. For example, members in the ligand-gated ion channel family [36] and the GIC family [37] are activated by major excitatory (glutamate) and inhibitory neurotransmitters (GABA) and participate in neuronal communication in the brain [38]. Recent studies show that some subunits of ligand-gated ion channels and GIC-type channels are expressed prominently during embryonic and postnatal brain development, while others are expressed mainly in the adult brain, suggesting that a switch in subunit composition may be required for normal brain development [38]. In plants, approximately one-third of the channel proteins are aquaporins (water channels) [39], many of which show a cell-specific expression pattern in the root, emphasizing the importance of regulating and maintaining turgor pressure through the plant [40].
Three fungal species, Saccharomyces cerevisiae, Schizosaccharomyces pombe, and Neurospora crassa, possess the largest portion of secondary transporters (76%–80%), mainly because of the prominent gene expansion of two types of functionally diverse MFS family transporters: (1) drug efflux pumps, which could play roles in the secretion of secondary metabolites, toxic compounds, and signaling molecules, and (2) sugar symporters, which could allow a broader range of sugar utilization [41,42].
Phylogenetic Profiling of Transporter Family and Substrate Shows Strong Correlations to Organisms' Overall Physiology
The phylogenetic profile of a given protein is a string that encodes the presence or absence of that protein in every fully sequenced genome. Proteins that function together in a pathway or a common structural complex are likely to evolve in a correlated fashion, and therefore tend to be either preserved or eliminated together in a new species during evolution [43,44]. Phylogenetic profiling has been an effective way to detect conserved core genes, species-specific gene families, lineage-specific gene family expansions [45], and subcellular localization of proteins [46]. It can also facilitate the prediction of physical and functional interactions and assist in the deduction of the functions of genes that have no well-characterized homologs [47,48].
We have undertaken a novel application of phylogenetic profiling to investigate the presence or absence of transporter protein families across sequenced genomes. To our knowledge this represents the first application of a phylogenetic profiling approach using protein families rather than individual proteins as the unit of comparison. With the data on membrane transport systems from 141 fully sequenced organisms, we were able to construct the phylogenetic profiles for each transporter family (Figures 4 and S2). Hierarchical clustering of phylogenetic profiles showed a strong correlation between the observed clustering pattern and phylogeny, with Eubacteria, Archaea, and Eukaryota clearly separated into different clusters. Inside the bacterial cluster, Gram-positive bacteria, Proteobacteria, Chlamydia, and Cyanobacteria are also clearly defined into different groups. Given that the profiling approach solely utilizes presence or absence of a transporter family and does not use sequence similarity directly, this indicates that the types of transporters utilized by organisms are related to their evolutionary history. Additionally, the clustering appears to be influenced by habitat or lifestyle of organisms. For example, the obligate intracellular pathogens/symbionts and a collection of soil/plant-associated microbes are separated into two distinct superclusters (Figure 5).
Figure 4 Phylogenetic Profiling of Transporter Families
Phylogenetic profiles were created for each transporter family. Each profile is a string with 141 entries (number of organisms analyzed). If a given family is present in an organism, the value one is assigned at this position (red). If not, zero is assigned (black). Organisms and transporter families were clustered according to the similarity of their phylogenetic profiles.
Figure 5 Hierarchial Clustering of Phylogenetic Profiles of Obligate Intracellular Pathogens/Symbionts versus Soil/Plant-Associated Microbes
Detailed view of two clusters of organisms generated by hierarchical clustering of their phylogenetic profiles of transporter families: obligate intracellular pathogens/symbionts and soil/plant-associated microbes.
The obligate intracellular pathogens/symbionts cluster includes a group of phylogenetically diverse organisms, including Chlamydia spp. (pathogens); γ-Proteobacteria such as Buchnera spp., Wigglesworthia glossinidia brevipalpis, and Candidatus Blochmannia floridanus (endosymbionts); α-Proteobacteria such as Wolbachia sp. (endosymbiont) and R.
prowazekii (pathogen); Gram-positive-like organisms Mycoplasma spp. and T. whipplei (pathogens); Spirochetes such as Tr. pallidum and Borrelia burgdorferi (pathogens); and an archaeal symbiont, Nanoarchaeum equitans. Organisms in this cluster share an obligate intracellular lifestyle as well as reduced genome size. The clustering does not appear to be due to genome size alone as nonobligate intracellular organisms with small genome sizes do not fall into this cluster. One possibility is that the transport needs of these obligate intracellular organisms are more specialized than those of environmental organisms because of the much more static nature of their intracellular environments. This may have allowed them to shed, for example, transporters for alternative nitrogen/carbon sources, osmoregulatory functions, and ion homeostasis. Similar to their prokaryotic counterparts, two eukaryotic intracellular parasites, P. falciparum and En. cuniculi, form a distinct cluster separate from the other eukaryotes.
The soil/plant-associated microbe cluster also contains species from various phylogenetic groups, such as Actinobacteria (Corynebacterium and Streptomyces), Firmicutes (Bacillus and Oceanobacillus), α-Proteobacteria (Brucella, Agrobacterium, Mesorhizobium, Sinorhizobium, and Bradyrhizobium), β-Proteobacteria (Bordetella and Ralstonia), γ-Proteobacteria (Pseudomonas and Rhodopseudomonas), δ-Proteobacteria (Geobacter), Deinococcus (Deinococcus radiodurans), Planctomycetes (Pirellula sp.), and Bacteroidetes (Bacteroides thetaiotaomicron). All of these organisms possess a robust collection of transporter systems. It is unlikely that these species are merely clustered by their genome sizes because some species in this cluster have relatively smaller genome sizes, like Corynebacterium efficiens (3.0 Mb), D. radiodurans (3.2 Mb), and Brucella melitensis (3.2 Mb). In addition, hierarchical clustering of organisms exclusively by genome size generates clusters with no apparent phylogenetic relationship (data not shown). The similarity of phylogenetic profiles of organisms in this cluster probably reflects the versatility of these organisms and their exposure to a wide range of different substrates in their natural environment. The majority of species in this cluster can be free-living in the soil, and some are capable of living in a diverse range of environments. They generally share a broad range of transport capabilities for plant-derived compounds specifically and for organic nutrients in general. Interestingly, some of the human pathogens, e.g., Bordetella, Brucella, Bacillus anthracis [26], and Bacteroides thetaiotaomicron, are also grouped in this cluster. All of these pathogens have close relatives that are soil- or plant-associated environmental organisms [49–52], so their transport capabilities probably reflect a combination of their evolutionary heritage, original environmental niche, and current transport needs.
To compare the transport capabilities of organisms in the intracellular pathogen/symbiont cluster and the soil/plant-associated microbe cluster, we carried out statistical analysis on their number of transporters, percentage of ORFs encoding transport proteins, and compositions in each transporter type (data not shown). Organisms in the soil/plant-associated microbe cluster on average have about eight times as many transporters as those in the intracellular organism cluster (p < 0.0001; p-value denotes the confidence level that the correlation observed is significantly different from the null hypothesis). The difference in the relative percentage of ORFs that are transporters is smaller but still significant (1.5-fold increase, p < 0.0001), suggesting that systematic gene loss and genome compaction is one of the important factors in reducing the number of transport proteins in intracellular organisms. The residual transport systems conserved in these obligate intracellular organisms probably belong to the core essential genes required for the acquisition of key nutrients and metabolic intermediates. For example, a glutamate transporter is encoded in two obligate endosymboints: the GltP glutamate:proton symporter (DAACS family) [53] in Candidatus Blochmannia floridanus, and GltJKL ABC transporter [54] in Wigglesworthia glossinidia brevipalpis. These organisms have a truncated citrate cycle that begins with α-ketoglutarate and ends with oxaloactetate [55]. Their citrate cycle could be closed by the transamination of the imported glutamate to aspartate, catalyzed by an aspartate aminotransferase (AspC) that uses oxaloactetate as a cosubstrate and produces α-ketoglutarate. As to the distribution of transporter types, there is no significant difference between these two clusters although intracellular organisms show a higher degree of variation in each transporter type than the plant/soil-associated microbes. These variations may reflect the unique internal environment inside the host cells. All these observations illustrate how adaptation of an organism to certain living conditions leads to changes in its transporter repertoire and at the same time determines the set of transporters that the organism cannot afford to lose.
In addition to investigating the relationship between organisms based on their transporter profiles, we also examined the clustering of transporter families. The essentially ubiquitous families, like ABC, MFS, P(F)-type ATPase, that are present in virtually every organism we analyzed, are clustered together. Eukaryotic-specific families, most of which are single-organism-specific ion channels, are grouped together. Interestingly, the sodium-ion-dependent families, like neurotransmitter:sodium symporter, alanine/glycine:cation symporter, solute:sodium symporter, and divalent anion:sodium symporter [56–58], are clustered together. Transporters in these families are all symporters that utilize the sodium ion gradient to transport amino acid, solute, and/or divalent ions into cytoplasm. This clustering may suggest that these families co-occur in a specific set of organisms, presumably those reliant on sodium-ion-driven transport.
Previous studies have shown that transporters with similar functions characteristically cluster together in phylogenetic analyses; hence, substrate specificity appears to be a conserved evolutionary trait in transporters [19,20,59,60]. The phylogenetic profiles of predicted substrates for all 141 organisms were generated and clustered by MeV (see Figure S3). Overall, similar patterns were observed as with the clustering by families. Organisms were grouped together either by their phylogenetic history or by their physiology or living habits. Ubiquitous substrates (e.g., cation, amino acid, sugar, and phosphate) and eukaryotic-specific substrates (e.g., cholesterol, UDP-sugars, and phosphoenolpyruvate) each form distinct clusters.
Distribution of Transporter Families among Species in the Same Genus
With the transporter data from a great diversity of sequenced organisms, we were able to compare the distribution of transporter families in closely related species (i.e., from the same genus) (Figures 6 and S4). In most of the cases we studied, species from the same genus share highly parallel distributions of transporter families. For example, three Pseudomonas species, Ps. aeruginosa [61], Ps. putida [62] and Ps. syringae [63], all of which are metabolically versatile soil/plant-associated bacteria, show highly similar patterns of transporter family distribution. Among the 66 transporter families present in this genus, 47 are shared by all three species and 14 are shared by two species (Figure 6A). All three species encode transporters for a diverse spectrum of substrates, including sugars, amino acids, peptides, carboxylates, and various cations and anions.
Figure 6 Venn Diagrams Showing the Distribution of Transporter Families among Species Belonging to the Same Genus
(A) Transporter family distribution among three Pseudomonas species.
(B) Transporter family distribution among three Corynebacterium species.
The distribution of transporter families in three Corynebacterium species represents an exception. Co. glutamicum [64] and Co. efficiens [65] are widely used in the industrial production of amino acids like glutamic acid and lysine by fermentation. The closely related Co. diphtheriae [66], however, is a human pathogen causing the respiratory illness diphtheria and lacks amino acid productivity. Compared to the other two species, Co. diphtheriae shows a dramatically different transporter family profile (Figure 6B). There are eight families specific to Co. diphtheriae, while only one for Co. glutamicum and three for Co. efficiens. More importantly, Co. diphtheriae uses totally different mechanisms to transport potassium ion and C4-dicarboxylates than the other two species. In Co. diphtheriae, potassium ions are transported into cytoplasm via a Trk family K+:H+ symporter [67], while both Co. glutamicum and Co. efficiens encode a KUP family potassium ion uptake permease [68]. Co. diphtheriae utilizes the DcuABC antiporter system [69] for the uptake of C4-dicarboxylate, while the other species use the ATP-independent tripartite periplasmic symporter systems (TRAP-T family) [70]. The common orthologs of transporters in families specific to one or two Corynebacterium species were identified in sequenced high-GC Gram-positive bacteria, and the phylogenetic trees were constructed by the neighbor-joining method (data not shown). For those families with orthologs in Co. glutamicum and Co. efficiens but not in Co. diphtheriae, orthologs were also identified in the majority of high-GC Gram-positive species. The trees of transport protein are similar to the 16S rRNA tree, suggesting certain transporter families in Co. efficiens are missing because of specific gene losses. By contrast, Co. diphtheriae–specific transporter families, like Dcu, DcuC, and Trk families, tend to have either no apparent orthologs or only distantly related homologs in other sequenced high-GC Gram-positive species, suggesting possible evolutionary gene acquisition events in Co. diphtheriae. The recent finding that both gene loss and horizontal gene transfer are responsible for the functional differentiation in amino acid biosynthesis of the three Corynebacterium species [71] further supports this conclusion.
All three Corynebacterium species share 41 transporter families. Interestingly, although Co. diphtheriae shows no amino acid productivity and has a reduced genome size [71], all the major types of amino acid exporters in Co. glutamicum [72] are conserved in Co. diphtheriae, e.g., the LysE family transporter for the export of basic amino acids, the RhtB family transporter for threonine efflux, the ThrE family transporter for threonine and serine export, and the LIV-E family transporter (BrnFE in Co. glutamicum), which is a two-component efflux pump exporting branched-chain amino acids [73]. The only difference observed among these organisms is the number of paralogs in the RhtB family: three in Co. glutamicum, two in Co. efficiens, and only one in Co. diphtheriae. The phylogenetic tree of the RhtB family suggests that gene duplication took place in the common ancestor of Corynebacterium, and that specific gene loss was responsible for the single RhtB transporter in Co. diphtheriae.
Conclusion
The rapid expansion of complete genome sequencing enabled us to conduct analyses of transporter capabilities on the whole-genome level. By comparing the membrane transport systems in Eubacteria, Archaea, and Eukaryota, we could draw conclusions as follows. (1) Eukaryotic species generally encode a larger number of transporters, but transporters account for a smaller percentage of total ORFs in eukaryotic than in prokaryotic species. Prokaryotic obligate intracellular pathogens and endosymbionts, as well as the eukaryotic parasites, possess the most limited repertoire of membrane transporters. (2) Organisms with a larger genome size tend to have a higher number of transporters. In prokaryotes and unicellular eukaryotes, this increase is primarily due to increased diversity of types of transporter. In multicellular eukaryotes, this increase is largely due to the greater number of paralogs by gene duplication or expansion in certain transporter families. (3) The distribution of different transporter types according to transport mode and energy coupling mechanism generally correlates with organisms' primary mechanism of energy generation. Compared to eukaryotes, prokaryotic species rely heavily on primary (active) transporters. Primary type transporters in Eubacteria and Archaea account for a much larger percentage of total transporters than any other transporter type. This phenomenon may be related to the absence of ABC-type uptake permeases in eukaryotes and, in some cases, the bioenergetic requirements and environmental constraints of prokaryotic organisms. (4) Energy-independent channel proteins are far more numerous in multicellular organisms and are often involved in cell–cell communication and signal transduction processes. Many channels are restricted to a single organismal type. The expression of different subunits of a channel in a timely fashion may be an essential step during embryonic development in mammals. (5) The PTS is only present in a subset of Eubacteria, and is completely absent in Archaea and Eukaryota. The expansion of sugar PTS systems in species dwelling in the gastrointestinal tract could provide the advantage to thrive in their ecological niches. (6) Hierarchical clustering of the phylogenetic profiles of transporter families showed that the distribution of transporter families appears to reflect a combination of evolutionary history and environment and lifestyle factors. (7) The distribution pattern of transporter families in species belonging to the same genus is usually parallel, with some notable exceptions that may reflect specific environmental differences.
Materials and Methods
We developed a semi-automated pipeline to annotate transport systems genome-wide, input the data into TransportDB database, and visualize the result through a Web interface [74]. The complete protein sequences from specific organisms were first searched against our curated database of transport proteins for similarity to known or putative transport proteins using BLAST [75,76]. All of the query proteins with significant hits (E-value < 0.001) were collected and searched against the NCBI nonredundant protein database and Pfam database [77]. Transmembrane protein topology was predicted by TMHMM [78]. A Web-based interface was created to facilitate the annotation processes, which incorporates number of hits to the transporter database, BLAST and HMM search E-value and score, number of predicted transmembrane segments, and the description of top hits to the nonredundant protein database. We also set up direct links between transporter classification family and COG classification [79] so that COG-based searches can inform the transporter annotation. The results can be viewed at the TransportDB Web site (http://www.membranetransport.org/).
To analyze the phylogenetic profiles of transporter families and predicted substrates, we assigned a profile to each transporter family or substrate. Each profile is a string with 141 entries (number of species analyzed). If a given family is present or a given substrate is transported in certain species, the value one was assigned at these positions (red for transporter families/purple for predicted substrates). If not, zero was assigned (black). Transporter families or substrates were clustered according to the similarity of their phylogenetic profiles using The Institute for Genomic Research's microarray multi-experiment viewer (MeV) [80] with two-dimensional hierarchical clustering as described by Eisen et al. [81].
Supporting Information
Figure S1 Comparison of the Percentage of Membrane Proteins with Six or More Transmembrane Segments That Were Annotated as “Hypothetical Protein” in Selected Archaea and Eubacteria
(37 KB PDF)
Click here for additional data file.
Figure S2 Detailed View of the Hierarchical Clustering of Phylogenetic Profiles of Transporter Families
(A) Clustering of species.
(B) Clustering of transporter families.
(722 KB PPT)
Click here for additional data file.
Figure S3 Phylogenetic Profiling of Predicted Transporter Substrates
Phylogenetic profiles were created for each predicted substrate. Each profile is a string with 141 entries (number of organisms analyzed). If a specific substrate is transported in a given organism, the value one is assigned at this position (purple). If not, zero is assigned (black). Organisms and substrates were clustered according to the similarity of their phylogenetic profiles.
(671 KB PDF)
Click here for additional data file.
Figure S4 Venn Diagrams Showing the Distribution of Transporter Families among Species Belonging to the Same Genus
(A) Transporter family distribution among three Bordetella species.
(B) Transporter family distribution among three Chlamydia species.
(C) Transporter family distribution among three Mycobacterium species.
(D) Transporter family distribution among three Pyrococcus species.
(E) Transporter family distribution among three Streptococcus species.
(F) Transporter family distribution among three Vibrio species.
(947 KB PDF)
Click here for additional data file.
Table S1 List of 141 Organisms Analyzed in This Study
(194 KB DOC)
Click here for additional data file.
We would like to thank Dr. Jonathan A. Eisen for comments and suggestions relating to the phylogenetic profiling analyses and Dr. Julie C. Dunning Hotopp for assistance with the MeV software and for critical reading of the manuscript. We also thank Robert L. Koenig for his contribution to the synopsis.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. QR and ITP conceived and designed the experiments and wrote the paper. QR analyzed the data and contributed materials/analysis tools.
Abbreviations
ABCadenosine triphosphate–binding cassette; ATP, adenosine triphosphate
GICglutamate-gated ion channel
MFSmajor facilitator superfamily
ORFopen reading frame
PTSphosphotransferase system
==== Refs
References
Saier MH Jr 1999 Classification of transmembrane transport systems in living organisms VanWinkle L Biomembrane transport San Diego Academic Press 265 276
Saier MH Jr 2000 A functional-phylogenetic classification system for transmembrane solute transporters Microbiol Mol Biol Rev 64 354 411 10839820
Saier MH Jr 1999 A functional-phylogenetic system for the classification of transport proteins J Cell Biochem 75 Suppl 32 84 94
Busch W Saier MH Jr 2002 The transporter classification (TC) system, 2002 Crit Rev Biochem Mol Biol 37 287 337 12449427
Busch W Saier MH 2004 The IUBMB-endorsed transporter classification system Mol Biotechnol 27 253 262 15247498
Sweet G Gandor C Voegele R Wittekindt N Beuerle J 1990 Glycerol facilitator of Escherichia coli: Cloning of glpF and identification of the glpF product J Bacteriol 172 424 430 2152911
Bolhuis H Poelarends G van Veen HW Poolman B Driessen AJ 1995 The lactococcal lmrP gene encodes a proton motive force-dependent drug transporter J Biol Chem 270 26092 26098 7592810
Newman MJ Foster DL Wilson TH Kaback HR 1981 Purification and reconstitution of functional lactose carrier from Escherichia coli
J Biol Chem 256 11804 11808 7028742
Viitanen P Newman MJ Foster DL Wilson TH Kaback HR 1986 Purification, reconstitution, and characterization of the lac permease of Escherichia coli
Methods Enzymol 125 429 452 3520229
Abramson J Smirnova I Kasho V Verner G Iwata S 2003 The lactose permease of Escherichia coli: Overall structure, the sugar-binding site and the alternating access model for transport FEBS Lett 555 96 101 14630326
Elferink MG Driessen AJ Robillard GT 1990 Functional reconstitution of the purified phosphoenolpyruvate-dependent mannitol-specific transport system of Escherichia coli in phospholipid vesicles: Coupling between transport and phosphorylation J Bacteriol 172 7119 7125 2123863
Postma PW Lengeler JW Jacobson GR 1993 Phosphoenolpyruvate:carbohydrate phosphotransferase systems of bacteria Microbiol Rev 57 543 594 8246840
Bernal A Ear U Kyrpides N 2001 Genomes OnLine Database (GOLD): A monitor of genome projects world-wide Nucleic Acids Res 29 126 127 11125068
Janssen P Goldovsky L Kunin V Darzentas N Ouzounis CA 2005 Genome coverage, literally speaking: The challenge of annotating 200 genomes with 4 million publications EMBO Rep 6 397 399 15864286
Davidson AL Chen J 2004 ATP-binding cassette transporters in bacteria Annu Rev Biochem 73 241 268 15189142
Schneider E Hunke S 1998 ATP-binding cassette (ABC) transport systems: Functional and structural aspects of the ATP-hydrolyzing subunits/domains FEMS Microbiol Lett 22 1 20
Marger MD Saier MH Jr 1993 A major superfamily of transmembrane facilitators that catalyse uniport, symport and antiport Trends Biochem Sci 18 13 20 8438231
Saier MH Jr Beatty JT Goffeau A Harley KT Heijne WH 1999 The major facilitator superfamily J Mol Microbiol Biotechnol 1 257 279 10943556
Paulsen IT Nguyen L Sliwinski MK Rabus R Saier MH Jr 2000 Microbial genome analyses: Comparative transport capabilities in eighteen prokaryotes J Mol Biol 301 75 100 10926494
Paulsen IT Sliwinski MK Saier MH Jr 1998 Microbial genome analyses: Global comparisons of transport capabilities based on phylogenies, bioenergetics and substrate specificities J Mol Biol 277 573 592 9533881
Paulsen IT Sliwinski MK Nelissen B Goffeau A Saier MH Jr 1998 Unified inventory of established and putative transporters encoded within the complete genome of Saccharomyces cerevisiae
FEBS Lett 430 116 125 9678606
Chen GQ Cui C Mayer ML Gouaux E 1999 Functional characterization of a potassium-selective prokaryotic glutamate receptor Nature 402 817 821 10617203
Mayer ML Olson R Gouaux E 2001 Mechanisms for ligand binding to GluR0 ion channels: Crystal structures of the glutamate and serine complexes and a closed apo state J Mol Biol 311 815 836 11518533
Bult CJ White O Olsen GJ Zhou L Fleischmann RD 1996 Complete genome sequence of the methanogenic archaeon, Methanococcus jannaschii
Science 273 1058 1073 8688087
Peterson JD Umayam LA Dickinson T Hickey EK White O 2001 The comprehensive microbial resource Nucleic Acids Res 29 123 125 11125067
Ruder K Winstead ER Gibbs MS 2004 A quick guide to sequenced genomes Rockville (Maryland) Genome News Network Available: http://www.genomenewsnetwork.org/resources/sequenced_genomes/genome_ guide_p1.shtml . Accessed 13 July 2005.
Gardner MJ Hall N Fung E White O Berriman M 2002 Genome sequence of the human malaria parasite Plasmodium falciparum
Nature 419 498 511 12368864
Katinka MD Duprat S Cornillot E Metenier G Thomarat F 2001 Genome sequence and gene compaction of the eukaryote parasite Encephalitozoon cuniculi
Nature 414 450453
Konstantinidis KT Tiedje JM 2004 Trends between gene content and genome size in prokaryotic species with larger genomes Proc Natl Acad Sci U S A 101 3160 3165 14973198
Cases I de Lorenzo V Ouzounis CA 2003 Transcription regulation and environmental adaptation in bacteria Trends Microbiol 11 248 12823939
Jordan IK Makarova KS Spouge JL Wolf YI Koonin EV 2001 Lineage-specific gene expansions in bacterial and archaeal genomes Genome Res 11 555 565 11282971
Sanchez-Fernandez R Davies TG Coleman JO Rea PA 2001 The Arabidopsis thaliana ABC protein superfamily, a complete inventory J Biol Chem 276 30231 30244 11346655
The Arabidopsis Genome Initiative 2000 Analysis of the genome sequence of the flowering plant Arabidopsis thaliana
Nature 408 796 815 11130711
Dean M Allikmets R 2001 Complete characterization of the human ABC gene family J Bioenerg Biomembr 33 475 479 11804189
Kung C Blount P 2004 Channels in microbes: So many holes to fill Mol Microbiol 53 373 380 15228520
Hong X 1998 Identification of major phylogenetic branches of inhibitory ligand-gated channel receptors J Mol Evol 47 323 9732459
Nakanishi S Masu M 1994 Molecular diversity and functions of glutamate receptors Annu Rev Biophys Biomol Struct 23 319 348 7919785
Lujan R Shigemoto R Lopez-Bendito G 2005 Glutamate and GABA receptor signalling in the developing brain Neuroscience 130 567 580 15590141
Johanson U Karlsson M Johansson I Gustavsson S Sjovall S 2001 The complete set of genes encoding major intrinsic proteins in Arabidopsis provides a framework for a new nomenclature for major intrinsic proteins in plants Plant Physiol 126 1358 1369 11500536
Javot H Maurel C 2002 The role of aquaporins in root water uptake Ann Bot 90 301 313 12234142
Galagan JE Calvo SE Borkovich KA Selker EU Read ND 2003 The genome sequence of the filamentous fungus Neurospora crassa
Nature 422 859 868 12712197
Borkovich KA Alex LA Yarden O Freitag M Turner GE 2004 Lessons from the genome sequence of Neurospora crassa : Tracing the path from genomic blueprint to multicellular organism Microbiol Mol Biol Rev 68 1 108 15007097
Pellegrini M Marcotte EM Thompson MJ Eisenberg D Yeates TO 1999 Assigning protein functions by comparative genome analysis: Protein phylogenetic profiles Proc Natl Acad Sci U S A 96 4285 4288 10200254
Pellegrini M 2001 Computational methods for protein function analysis Curr Opin Chem Biol 5 46 50 11166647
Vandepoele K Van de Peer Y 2005 Exploring the plant transcriptome through phylogenetic profiling Plant Physiol 137 31 42 15644465
Marcotte EM Xenarios I van der Bliek AM Eisenberg D 2000 Localizing proteins in the cell from their phylogenetic profiles Proc Natl Acad Sci U S A 97 12115 12120 11035803
Levesque M Shasha D Kim W Surette MG Benfey PN 2003 Trait-to-gene: A computational method for predicting the function of uncharacterized genes Curr Biol 13 129 133 12546786
Kriventseva EV Biswas M Apweiler R 2001 Clustering and analysis of protein families Curr Opin Struct Biol 11 334 339 11406384
Paulsen IT Seshadri R Nelson KE Eisen JA Heidelberg JF 2002 The Brucella suis genome reveals fundamental similarities between animal and plant pathogens and symbionts Proc Natl Acad Sci U S A 99 13148 13153 12271122
Xu J Bjursell MK Himrod J Deng S Carmichael LK 2003 A genomic view of the human–Bacteroides thetaiotaomicron symbiosis Science 299 2074 2076 12663928
Parkhill J Sebaihia M Preston A Murphy LD Thomson N 2003 Comparative analysis of the genome sequences of Bordetella pertussis,
Bordetella parapertussis and Bordetella bronchiseptica
Nat Genet 35 32 40 12910271
Gerlach G von Wintzingerode F Middendorf B Gross R 2001 Evolutionary trends in the genus Bordetella
Microbes Infect 3 61 72 11226855
Tolner B Ubbink-Kok T Poolman B Konings WN 1995 Cation-selectivity of the L-glutamate transporters of Escherichia coli,
Bacillus stearothermophilus and Bacillus caldotenax: Dependence on the environment in which the proteins are expressed Mol Microbiol 18 123 133 8596452
Linton KJ Higgins CF 1998 The Escherichia coli ATP-binding cassette (ABC) proteins Mol Microbiol 28 5 13 9593292
Zientz E Dandekar T Gross R 2004 Metabolic interdependence of obligate intracellular bacteria and their insect hosts Microbiol Mol Biol Rev 68 745 770 15590782
Reizer J Reizer A Saier MH Jr 1994 A functional superfamily of sodium/solute symporters Biochim Biophys Acta 1197 133 166 8031825
Saier MH Jr Eng BH Fard S Garg J Haggerty DA 1999 Phylogenetic characterization of novel transport protein families revealed by genome analyses Biochim Biophys Acta 1422 1 56 10082980
Pajor AM 2000 Molecular properties of sodium/dicarboxylate cotransporters J Membr Biol 175 1 8 10811962
Pao SS Paulsen IT Saier MH Jr 1998 Major facilitator superfamily Microbiol Mol Biol Rev 62 1 34 9529885
Jack DL Paulsen IT Saier MH 2000 The amino acid/polyamine/organocation (APC) superfamily of transporters specific for amino acids, polyamines and organocations Microbiology 146 1797 1814 10931886
Stover CK Pham XQ Erwin AL Mizoguchi SD Warrener P 2000 Complete genome sequence of Pseudomonas aeruginosa PA01, an opportunistic pathogen Nature 406 959 964 10984043
Nelson KE Weinel C Paulsen IT Dodson RJ Hilbert H 2002 Complete genome sequence and comparative analysis of the metabolically versatile Pseudomonas putida KT2440 Environ Microbiol 4 799 808 12534463
Buell CR Joardar V Lindeberg M Selengut J Paulsen IT 2003 The complete genome sequence of the Arabidopsis and tomato pathogen Pseudomonas syringae pv. tomato DC3000 Proc Natl Acad Sci U S A 100 10181 10186 12928499
Kalinowski J Bathe B Bartels D Bischoff N Bott M 2003 The complete Corynebacterium glutamicum ATCC 13032 genome sequence and its impact on the production of aspartate-derived amino acids and vitamins J Biotechnol 104 5 25 12948626
Nishio Y Nakamura Y Kawarabayasi Y Usuda Y Kimura E 2003 Comparative complete genome sequence analysis of the amino acid replacements responsible for the thermostability of Corynebacterium efficiens
Genome Res 13 1572 1579 12840036
Cerdeno-Tarraga AM Efstratiou A Dover LG Holden MTG Pallen M 2003 The complete genome sequence and analysis of Corynebacterium diphtheriae NCTC13129 Nucleic Acids Res 31 6516 6523 14602910
Schlosser A Meldorf M Stumpe S Bakker EP Epstein W 1995 TrkH and its homolog, TrkG, determine the specificity and kinetics of cation transport by the Trk system of Escherichia coli
J Bacteriol 177 1908 1910 7896723
Trchounian A Kobayashi H 1999 Kup is the major K+ uptake system in Escherichia coli upon hyper-osmotic stress at a low pH FEBS Lett 447 144 148 10214935
Engel P Kramer R Unden G 1994 Transport of C4-dicarboxylates by anaerobically grown Escherichia coli . Energetics and mechanism of exchange, uptake and efflux Eur J Biochem 222 605 614 8020497
Kelly DJ Thomas GH 2001 The tripartite ATP-independent periplasmic (TRAP) transporters of bacteria and archaea FEMS Microbiol Rev 25 405 424 11524131
Nishio Y Nakamura Y Usuda Y Sugimoto S Matsui K 2004 Evolutionary process of amino acid biosynthesis in Corynebacterium at the whole genome level Mol Biol Evol 21 1683 1691 15163767
Eggeling L Sahm H 2003 New ubiquitous translocators: Amino acid export by Corynebacterium glutamicum and Escherichia coli
Arch Microbiol 180 155 160 12879215
Kennerknecht N Sahm H Yen MR Patek M Saier MH Jr 2002 Export of L-isoleucine from Corynebacterium glutamicum: A two-gene-encoded member of a new translocator family J Bacteriol 184 3947 3956 12081967
Ren Q Kang KH Paulsen IT 2004 TransportDB: A relational database of cellular membrane transport systems Nucleic Acids Res 32 D284 D288 14681414
Altschul SF Gish W Miller W Myers EW Lipman DJ 1990 Basic local alignment search tool J Mol Biol 215 403 410 2231712
Altschul SF Gish W 1996 Local alignment statistics Methods Enzymol 266 460 480 8743700
Sonnhammer EL Eddy SR Birney E Bateman A Durbin R 1998 Pfam: Multiple sequence alignments and HMM-profiles of protein domains Nucleic Acids Res 26 320 322 9399864
Krogh A Larsson B von Heijne G Sonnhammer EL 2001 Predicting transmembrane protein topology with a hidden Markov model: Application to complete genomes J Mol Biol 305 567 580 11152613
Tatusov RL Natale DA Garkavtsev IV Tatusova TA Shankavaram UT 2001 The COG database: New developments in phylogenetic classification of proteins from complete genomes Nucleic Acids Res 29 22 28 11125040
Saeed AI Sharov V White J Li J Liang W 2003 TM4: A free, open-source system for microarray data management and analysis Biotechniques 34 374 378 12613259
Eisen MB Spellman PT Brown PO Botstein D 1998 Cluster analysis and display of genome-wide expression patterns Proc Natl Acad Sci U S A 95 14863 14868 9843981
|
16118665
|
PMC1188273
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Aug 19; 1(3):e27
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010027
|
oa_comm
|
==== Front
PLoS Comput BiolPLoS Comput. BiolpcbiplcbploscompPLoS Computational Biology1553-734X1553-7358Public Library of Science San Francisco, USA 1611866610.1371/journal.pcbi.001003105-PLCB-RA-0106R1plcb-01-03-05Research ArticleBioinformatics - Computational BiologyMolecular Biology - Structural BiologyHomo SapienFunctional Coverage of the Human Genome by Existing Structures, Structural Genomics Targets, and Homology Models Functional and Structural SpaceXie Lei Bourne Philip E *San Diego Supercomputer Center and Department of Pharmacology, University of California, San Diego, California, United States of AmericaThornton Janet EditorEuropean Bioinformatics Institute, United Kingdom*To whom correspondence should be addressed. E-mail: [email protected] 2005 19 8 2005 24 7 2005 1 3 e3116 5 2005 18 7 2005 Copyright: © 2005 Xie and Bourne.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.The bias in protein structure and function space resulting from experimental limitations and targeting of particular functional classes of proteins by structural biologists has long been recognized, but never continuously quantified. Using the Enzyme Commission and the Gene Ontology classifications as a reference frame, and integrating structure data from the Protein Data Bank (PDB), target sequences from the structural genomics projects, structure homology derived from the SUPERFAMILY database, and genome annotations from Ensembl and NCBI, we provide a quantified view, both at the domain and whole-protein levels, of the current and projected coverage of protein structure and function space relative to the human genome. Protein structures currently provide at least one domain that covers 37% of the functional classes identified in the genome; whole structure coverage exists for 25% of the genome. If all the structural genomics targets were solved (twice the current number of structures in the PDB), it is estimated that structures of one domain would cover 69% of the functional classes identified and complete structure coverage would be 44%. Homology models from existing experimental structures extend the 37% coverage to 56% of the genome as single domains and 25% to 31% for complete structures. Coverage from homology models is not evenly distributed by protein family, reflecting differing degrees of sequence and structure divergence within families. While these data provide coverage, conversely, they also systematically highlight functional classes of proteins for which structures should be determined. Current key functional families without structure representation are highlighted here; updated information on the “most wanted list” that should be solved is available on a weekly basis from http://function.rcsb.org:8080/pdb/function_distribution/index.html.
Synopsis
The sequencing of the human genome provides biologists with new opportunities to understand the molecular basis of physiological processes and disease states. To take full advantage of these opportunities, the three-dimensional structures of the gene products are needed to provide the appropriate level of detail. Since protein structure determination lags behind protein sequence determination, an important and ongoing question becomes: what degree of coverage of the human proteome do we have from experimental structures, and what can we infer by modeling? Or, turning the question around: what structures do we need to determine (the “most wanted list”) to further our understanding of the human condition? This paper addresses these questions through integration of existing data resources correlated using comparative functional features, namely the Gene Ontology, which describes biochemical process, molecular function, and cellular location for all types of proteins, and the Enzyme Commission classification for enzymes. Genetic disease states are linked through the Online Mendelian Inheritance in Man resource. Readers can ask their own questions of the resource at http://function.rcsb.org:8080/pdb/function_distribution/index.html. The resource should prove particularly useful to the structural genomics community as it strives to undertake large-scale structure determination with a goal of improving the understanding of protein functional space.
Citation:Xie L, Bourne PE (2005) Functional coverage of the human genome by existing structures, structural genomics targets, and homology models. PLoS Comp Biol 1(3): e31.
==== Body
Introduction
The three-dimensional structure of a protein is an essential component in elucidating the biological function(s) at the molecular level and in understanding the details of molecular recognition. Traditional structural biology supports a paradigm in which biochemical evidence of function is confirmed and further understood through the study of structure [1]. Structural genomics [2] has changed this paradigm, being motivated by a variety of criteria, including a desire to increase the coverage of known fold space [3]. Concomitantly, complete genome sequences are becoming available at an increasing rate and both putative functions and structures defined for coding regions. Even given the limitations of these assignments, it is an appropriate time to assess the current coverage of protein structure space from a functional perspective relative to the perceived functional coverage of complete genomes, notably human. Further, the registering of structural genomics targets (sequences subject to structure determination) by most projects worldwide [4] provides an excellent opportunity to assess what the perceived coverage of functional space by structure will likely be going forward. This paper makes this assessment, discusses where a change of strategy in selecting targets may be appropriate, and reports on functional classes that are well represented in the human genome but without the existence of structures—the so-called “most wanted list.”
Many authors have noted the structural and functional bias in the Protein Data Bank (PDB), but few have attempted to quantify it [5–8]. Rather, general statements are made that refer to the limitations associated with structure determination methods, such as the propensity for small, globular, soluble proteins solved by X-ray crystallography and nuclear magnetic resonance. Beyond physical limitations, there is a bias toward proteins identified as potential drug targets and a historical bias toward structures that, without the benefit of modern techniques, were, from the point of view of protein isolation and structure determination, the most tractable. Where does that put us today, and how can we estimate this bias? A problem that has thwarted such studies is the lack of a common reference frame. This problem has been partially addressed by systems of consistent nomenclature; notwithstanding, depth of coverage is neither complete nor consistent across protein families. Recently, the Enzyme Commission (EC) classification has been used to study the relationships among sequence, structure, and functions [9–15]. Similarly, the Gene Ontology (GO) [16], while still evolving, provides a consistent view of molecular function, biological process, and cell component beyond enzymes. Further, with consistent sequence annotation as a common feature between resources describing structure and human genetic disease, structure–disease relationships can be inferred. Inference requires that care must be taken to assess the statistical significance of the outcome. Using EC and GO to define a common functional framework and highly significant sequence relationships to infer relationships between structure (either solved or under study) and disease, we can measure the biased nature of the PDB and the structures under consideration by structural genomics and suggest protein structures that should be determined to further our understanding of structure and function space.
Results/Discussion
The relationship between protein structures and their function(s) is complex. A single structure superfamily often displays variations in function. Conversely, the same function can be achieved by proteins with different structures [10,11]. Domain recombination and shuffling leads to further functional diversity [17–19]; hence, this study is undertaken at the level of both single domain and whole protein to provide an in-depth view of the functional distribution of protein structures. Single-domain coverage is defined such that at least one domain in the protein has structural information available from the PDB, structural genomics, or homology models. Whole-protein coverage means that structure information for all domains, including their organization, can be directly or indirectly inferred. Similarities between functional distributions from the human genome and from experimental structures or theoretical models are measured with Kendall's tau correlation, which ranges from −1.0 to 1.0. A large positive value indicates that two measurements have similar ranks. The structure–function relationship analysis is based on non-redundant sequence clusters with less than 40% sequence identity and 90% overlap, since functional similarity usually breaks down below these thresholds [10,15].
The Functional Bias of PDB Structures
As stated, several studies have noted structural and functional bias in the PDB [5–8]. In general, protein domains such as transmembrane domains, low complexity regions, and disordered regions, which are not suited to current structure determination methods by X-ray crystallography and nuclear magnetic resonance, are highly underrepresented in the PDB. The columns labeled “PDB/Genome” in Tables 1–4 quantify this bias relative to the known functional classification within the human genome using EC (Table 1) and GO (Tables 2–4) classifications. This bias is examined from the perspective of both a single domain and the whole structure, since many proteins have intracellular and extracellular domains that have been solved without their domain spanning regions. For example, proteins associated with transporter activity (Table 2) have the lowest coverage at the domain level (21.0%), but are further underrepresented at the structure level (12.1%) because of the presence of transmembrane domains. Proteins with two or more contiguous domains, where each of the domains has structure information available, may result in different structures when those domains are swapped. This impacts the observed relationship between values of coverage and correlation computed with Kendall's tau (see Materials and Methods) for single domains versus whole structures, as will be described subsequently.
Table 1 Coverage/Kendall's Tau Correlations for Major Categories of Enzyme for Both Single Domains and Whole Proteins
Current values for nodes of these major branches can be determined from http://function.rcsb.org:8080/pdb/function_distribution/index.html.
Table 2 Coverage/Kendall's Tau Correlations for Major Categories of GO Molecular Function for Both Single Domains and Whole Proteins
Current values for nodes of these seven major branches and other eight minor categories can be determined from http://function.rcsb.org:8080/pdb/function_distribution/index.html.
a Eight minor categories are not listed in the table, and can be browsed from the Web site.
Beyond obvious experimental limitations, skewed functional distributions of PDB structures are observed for almost all types and levels of EC and GO classifications (Tables 1–4). For example, consider classification by EC number at all levels (only the top-level EC classification is given in Table 1, but current values for all levels of the EC hierarchy are available from the Web site). The correlation coefficients by Kendall's tau between the genome sequences and PDB structures with single-domain coverage for EC: *.*.*.* (all), 2.*.*.* (transferases), EC 2.7.*.* (transferring phosphorous-containing groups), and EC 2.7.1.* (phosphotransferase with an alcohol group as acceptor) are 0.867, 0.889, 0.806, and 0.383, with coverage of 29.9%, 25.4%, 28.7%, and 32.1%, respectively. Thus, even for one of the most structurally studied superfamilies, the protein kinase-like superfamily (all belong to EC 2.7.1), the structures of the majority of atypical kinases (proteins that phosphorylate a variety of substrates) have not been determined, and the protein kinase family itself is slightly underrepresented. The Kendall's tau correlation coefficient is only 0.383 and 0.192 for single-domain and whole-protein coverage, respectively, for the protein kinase-like superfamily.
The functional bias of PDB structures is also notable when using GO molecular function annotations that extend beyond enzyme activity. A total of 16,211 proteins within the human genome can be annotated at this time. As shown in Table 2, according to their Kendall's tau at the whole structure level, several subcategories are underrepresented, notably transporter activity (already noted) and structural molecule activity. Looking deeper (refer to http://function.rcsb.org:8080/pdb/function_distribution/index.html), there are 16 GO subcategories of molecular function associated with structural molecule activity. Twelve of them have been mapped to the human genome. The most structurally underrepresented proteins include the structural constituent of ribosome (two structures but 180 annotations), of myelin sheath (zero structures and two annotations), of epidermis (zero structures and six annotations), of tooth enamel (zero structures and five annotations), of bone (zero structures and three annotations), of chorion (zero structures and one annotation), and of cell wall (zero structures and one annotation).
Table 3 provides the distribution of protein domains and whole structures according to the subcategories of GO biological process. A total of 14,876 proteins within the human genome can be annotated at this time. Thus, biological process is less well characterized than molecular function, presumably since molecular function cannot necessarily be related to a role in a complex biological process. Notwithstanding, both single-domain and whole-protein structures with an identified role in cellular process are underrepresented.
Table 3 Coverage/Kendall's Tau Correlations for Major Categories of GO Biological Process for Both Single Domains and Whole Proteins
Current values for nodes of these five major branches and other two minor categories can be determined from http://function.rcsb.org:8080/pdb/function_distribution/index.html.
a Two minor categories—viral life of cycle and biological process unknown—are not listed in the table, and can be browsed from the Web site.
Table 4 Coverage/Kendall's Tau Correlations for Major Categories of GO Cell Component for Both Single Domains and Whole Proteins
Current values for nodes of these five major branches and other two minor categories can be determined from http://function.rcsb.org:8080/pdb/function_distribution/index.html.
a Two minor categories—virion and cell component unknown—are not listed in the table, and can be browsed from the Web site.
Table 4 provides the distribution of protein domains according to the subcategories of GO cell component. Overall, the distribution between PDB structures and the human genome is comparable, with a Kendall's tau of 0.714; however, proteins identified within the cell (GO 5623) are underrepresented at both the structure and domain levels. Coverage is not as favorable as distribution: only 38.3% of the subcategories of cell location have at least one structure domain representative. Of those, the vast majority are annotated as cell (4,599 out of 8,936 gene clusters). Under cell (see http://function.rcsb.org:8080/pdb/function_distribution/index.html), there are 12 subcategories that have been assigned to the human genome, five have no structure representation and seven have at least one structure domain representative (membrane [3,354 gene and 236 structure clusters], intracellular [1,237 gene and 169 structure clusters], cell surface [23 gene and four structure clusters], cell projection [26 gene and one structure cluster], cell fraction [621 gene and 75 structure clusters], apical part of cell [six gene and one structure cluster], and basal part of cell (two gene and one structure cluster]). As expected, membrane is structurally underrepresented, and intracellular and cell fraction is structurally overrepresented. There is no structural information available for five small subcategories of cell: site of polarized growth (three gene clusters), periplasmic space (three gene clusters), midbody (one gene cluster), external encapsulation (two gene clusters), and cell soma (one gene cluster).
Structural Coverage of Human Genetic Diseases
Three-dimensional protein structures are important in understanding the mechanisms of human genetic diseases [20], predicting the effect of non-synonymous single nucleotide polymorphisms [20,21], and developing new personalized medicines [22]. For example, a recent study highlights the application of PDB structure and homology models in understanding a predisposition to breast and ovarian cancer [23]. However, the current identification and coverage of human genetic disease space, as identified by the Online Mendelian Inheritance in Man (OMIM), is limited: 218 non-redundant human genome sequence clusters, 46 structure clusters, and 34 structural genomics target clusters. The PDB currently covers 69.9% of the disease categories described by OMIM, but the distribution based on class of disease is uneven. For example, diseases of the central nervous system have the largest single representation (20 structure clusters) with a disproportionately large number of structural genomics targets (23 target clusters). Blood and lymph based diseases have a disproportionate number of ten solved structures, but an underrepresentation of targets (three clusters). Conversely, diseases of the ear, nose, and throat are underrepresented (six structure and seven target clusters). Overall, cancers have an appropriate level of structures and targets; however, digging deeper reveals a different situation. For example, there are no structures available for the five human proteins that have been associated with prostate cancer, although homology models can be inferred for domains such as prostate specific kallikrein of serine proteases [24]. Data showing measurable differences in protein and gene regulatory networks between the early- and the late-stage prostate cancer [25] only highlight the need to further understand the structural basis of this disease. Human genetic disease distributions, while limited, are undoubtedly influenced by historical precedent, preventative and treatable conditions, and social and, hence, funding pressures.
The Contributions of Homology Modeling
While the number of three-dimensional structures of proteins has increased close to the near-exponential rate predicted by Dickerson in 1978 as number of structures = exp(0.19 × year) [26], there are still a vast number of protein sequences without structure information available. Homology modeling can potentially provide putative structure information for these sequences to facilitate our understanding of their function and evolution [27,28]. Reliable homology modeling usually requires that the query sequence share at least 30% sequence identity with the template structure for each domain [27]. Domain rearrangements and lack of domain structures reduce the effectiveness of homology modeling for whole structures, as shown in the columns labeled “Model/Genome” in Tables 1–4. In almost all EC and GO classifications, coverage and distribution falls for whole structures versus domains. From a biological perspective, modeling of only a subset of domains within a structure limits the value of modeling.
As expected, the distribution of homology models is highly correlated with the availability of PDB structures. Single-domain coverage across the whole human genome indicates that our ability to provide homology models for domains in the different GO molecular function categories varies from 32% to 75%. For the modeling of whole proteins, coverage drops, varying from 16% to 41%. Transporter activity and signal transducer activity is the most difficult to model at the whole-protein level. GO functional coverage for signal transducer activity drops from 0.541 to 0.155. Thus, while catalytic domains involved in signal transduction are well represented and can be modeled in 54% of cases, these data quantitatively show that the associated non-transmembrane domains of the whole protein are significantly underrepresented, thereby limiting our ability to model whole proteins in 38% of cases.
Considering enzymes alone, our ability to homology model single domains is fairly evenly distributed across all major classes (Table 1). At the whole-protein level, this picture changes. Retaining a high Kendall's tau even as coverage drops significantly could imply that functional diversity comes primarily from domain recombination rather than from new domains that cannot be modeled. Indeed, it has recently been reported that contemporary ligases evolved by domain fusions [29], a fact supported by a relatively small drop in Kendall's tau from 1.000 to 0.745 for single domains versus the whole protein.
Low correlation within a functional class implies that homology models can be inferred from structures in different functional subclasses and other species. For example, in the oxidoreductases (EC 1.x.x.x ), five classifications (1.7, acting on other nitrogenous compounds as donors; 1.9, acting on a heme group of donors; 1.10, acting on diphenols and related substances as donors; 1.17, acting on –CH2 groups; and 1.97, other oxidoreductase) are not structurally covered at all. However, with the exception of 1.97, other oxidoreductase, the proteins in the four remaining subclasses can be modeled from structural templates present in the other functional subcategories, implying a close evolutionary relationship within this functional class.
Conversely, experimental structural coverage is more critical for functional classes with more distinct evolutional origins, such as the protein kinase-like superfamily, which is in the transferase category. It has been suggested [30] that atypical kinases diverged early in evolution from protein kinases; therefore, homology models of atypical kinases derived from protein kinases are likely insufficient to infer their functional and evolutional roles. In 13% of cases, while the homology model is identified to belong to the protein kinase-like superfamily, the specific family cannot be determined.
The Contribution of Structural Genomics
One approach to the selection of structural genomics targets has been to focus on increasing the coverage of fold space [31–33]. A recent review suggests that the first phase of structural genomics has been successful in this regard [34]. It is anticipated that functional roles will be given greater precedence in future phases of the project [35–37]. If so, a question to address is: does the current complement of structural genomics targets and the structures solved by these projects reduce the functional bias present in the PDB? The short answer is yes, but only significantly in some functional categories (Tables 1–4, columns labeled “SG/Genome”) and assuming two to three times the number of structures than we have now (based on the relative numbers of clusters between structure genomics targets and PDB structures, given a 40% sequence identity cutoff).
Within the enzymes (Table 1), ligases will benefit the most and lyases the least. Based on GO molecular function (Table 2), structural molecule activity and transcription regulator activity (single domain) will be impacted the most; binding, catalytic activity, signal transducer activity, and transporter activity the least. In terms of GO biological processes (Table 3), structural genomics will contribute almost nothing to our understanding of behavior and about equally to cellular processes, development, physiological processes, and regulation of biological processes. Finally, current structural genomics targets will not contribute to our understanding of extracellular region of cell component (Table 4). The most notable impact of structural genomics overall is in our potential understanding of transcription regulator activity, which shows an improvement in coverage from 0.542 to 0.750 and an improved Kendall's tau of 0.636 to 0.925 for a single domain.
Drilling down into one of these categories, the previously described structurally underrepresented GO class for molecular function—namely, structural molecule—becomes better populated such that targets will increase the coverage of the structural constituent of tooth enamel (one structure but five annotations), of myelin sheath (one structure and two annotations), and of ribosome (48 structures and 180 annotations). There remains no anticipated experimental structure information for the structural constituent of epidermis, bone, chorion, and cell wall (total 11 annotations).
Given these findings, it is timely to consider the choice of structural genomics targets. It has been suggested that solving the structures of proteins from the 5,000 Pfam families will cover more of fold space than focusing on a single genome [38]. Here, we look at target selection from a functional perspective and provide a tool for comparing the functional coverage by the existing PDB and what the existing complement of structural genomics targets do to that functional coverage. The remainder of the paper considers one application of the tool in providing a strategy for selecting structures that could be used to maximize our understanding of structure–function relationships with respect to the human genome.
Defining Structures That Should Be Determined
To date, approximately 50% of human genes (16,211 terms for GO molecular function, 14,876 terms for GO biological process, and 13,322 terms for GO cell component) have at least one GO annotation. However, approximately 70% of these GO molecular function categories are yet to be covered by experimental structures with even one identifiable domain. The structural coverage of the human genome is even lower with respect to sequence space: approximately 10% coverage by structure at 40% sequence identity. Stated another way, 5% of the human genome, which covers 30% of functional space, has structure representation for at least one domain in a protein. If all current structural targets were determined, it is estimated that coverage of the human genome and functional space would increase to 20% and 50%, respectively. Homology modeling would increase genome and functional coverage to 40% and 60%, but what these putative high-throughput models add to our understanding of molecular function remains questionable. When taking domain recombination into account, the functional coverage of the human genome by existing experimental structures and anticipated structures being determined by structure genomics decreases to approximately 25% of the functional space.
This lack of coverage perhaps calls for a new strategy to select targets for structure determination. Here, one such strategy is outlined for choosing targets to increase the coverage of functional space. It is based on the following criteria: (1) functional categories are preferred where proteins with experimental or theoretical three-dimensional models are underrepresented; (2) from (1), proteins without SCOP superfamily assignments are preferred; (3) if the protein is identified as being associated with a disease or is identified in multiple functional categories, it has a higher priority; and (4) less experimentally tractable proteins—for example, those with transmembrane segments—can be filtered out.
From our initial analysis, approximately 2,000 non-redundant human genes with GO annotation have no experimental structure in the PDB, nor are they identified structural genomics targets or amenable to homology modeling. Of this 2,000, approximately 50% include transmembrane domains. After removing transmembrane and low-complex regions, about 1,800 include at least one domain that is potentially solvable. The most understudied proteins of this 1,800 are various types of transporters and receptors. It should be noted that it requires fewer targets to cover this functional space than the equivalent sequence space.
Ranked by the size of the cluster of proteins, examples of the most pressing biological molecule functions for which structural representation is needed and soluble structure domains are probably present are listed here (Tables 5 and 6) and at http://function.rcsb.org:8080/pdb/function_distribution/index.html, which is updated regularly. For catalytic activity, most of them are involved in protein syntheses and gene regulation. For binding, most of them are involved in signal transduction and have additional benefit as potential drug targets.
Table 5 Most Wanted Structures According to EC Numbersa
The proteins are clustered with 40% sequence identity.
a Data for clusters of fewer than three can be obtained from http://function.rcsb.org:8080/pdb/function_distribution/index.html.
Table 6 Most Wanted Structures According to GO Classificationa
The proteins are clustered with 40% sequence identity.
a Data for clusters of fewer than five can be obtained from http://function.rcsb.org:8080/pdb/function_distribution/index.html.
Several genes without experimental structures and not found in the structural genomics target list are annotated by both GO and disease terms (see http://function.rcsb.org:8080/pdb/function_distribution/index.html). For example, congenital adrenal hyperplasia is associated with three gene clusters. Two of them are annotated with oxygen binding (GO ID: 19825), and one with steroid 11-beta-monooxygenase activity (GO ID: 4507).
In summary, by using common annotation as found in the GO and the EC classification scheme, we have been able to correlate the biological functions of proteins and their constituent domains for both experimentally derived structures and those under determination by structural genomics projects worldwide. Further, by using empirical sequence limitations known from homology modeling experiments and by clustering human genome sequences according to sequence identity, we can estimate the impact that current structure determination strategies will have on our understanding of structure–function relationships from homology modeling. Finally, by introducing relationships between gene products and known disease states, we have provided pointers for choosing structures to be determined to have the maximum impact on our understanding of human genetic disease. To facilitate these choices, a Web resource has been established at http://function.rcsb.org:8080/pdb/function_distribution/index.html to allow readers to make their own assessments of the progress of structural biology. The resource will be updated on a weekly basis to provide a current view. The resource itself will be the subject of a separate publication.
Materials and Methods
Date sources and annotation mapping.
The human genome sequences (version 26.35.1) were downloaded from Ensembl database [39]. Wild-type sequences associated with PDB structures were generated by associating the structural sequence with that from UniProt [40] using database cross references records. Subsequently, all wild-type PDB sequences of the human proteins were mapped to the genes in the human genome through sequence alignment using Blast [41]. A gene was considered to have a structure representation if it had 100% sequence identity with the wild-type sequence of the PDB structure. Structural genomics targets were taken from targetdb [42], the worldwide repository of all sequences representing structures being attempted. Among more than 5,000 registered human target sequences, there were 3,141 and 4,784 targets mapped to the 3,200 and 4,218 Ensembl human genes with sequence identity 100% and greater than 90%, respectively. The 4,784 targets with sequence identity above 90% were used in our analysis, with 2,180 of them having GO or EC terms assigned.
Sequences were assigned GO terms from the EBI GOA resource (http://www.ebi.ac.uk/GOA). The query sequence was aligned with the UniProt GO annotated sequence with Blast [41]. If the Blast sequence identity was above 40%, and the overlap was above 90%, the annotated GO terms were mapped to the query gene (16,211 for GO molecular function, 14,876 for GO biological process, and 13,322 for GO cell component). The threshold is based on the observation that below 40% sequence identity with global alignment, the functional similarity relationship breaks down [10,15]. Sequences were also mapped to enzyme classification numbers with the annotations and sequences in the UniProt database as the reference. The 40% sequence identity and 90% overlap threshold was also applied to EC mapping.
Genome sequences were masked for low-complexity regions, coiled-coils, and transmembrane helical domains, using SEG [43], Coils [44], and TMHMM [45], respectively. SCOP superfamily domains [46] of unmasked regions of human genome sequences were assigned with HMMER [47]. A set of hidden Markov Models of SCOP domains was taken from SUPERFAMILY 1.65 [48]. Given the current stage of homology modeling, the model was usually reliable when the sequence identity was above 30% between the query sequence and the template structure [25]. Thus, only those assigned domains with sequence identity above 30% in the alignment were considered as homology models. The sequence regions that were not assigned by SCOP domains were further parsed with Pfam 16.0 [49]. The remaining unmasked sequence segments that were not annotated by either SCOP or Pfam but longer than 30 residues were considered as novel domains. Moreover, for contiguous domains, their orders were recorded in the database. The two domains were considered as contiguous with each other if they were not separated by the filtered sequence segments.
All genome sequences were clustered with 40% sequence identity and 90% overlap using CD-HIT [50].
For PDB structures and structural genomics targets, SCOP domains and their arrangements were computed with the same procedure as for genome sequences.
The original mapping of structures to OMIM numbers was taken from SWISS-PROT [51]. The mapping of genome sequences to OMIM numbers was from NCBI [52]. These mappings were recorded and used from the PDB beta site [26].
Data analysis.
For each functional or structural category, the number of sequence or structure clusters in the subcategory was normalized with that of sequence clusters from the genome. The overall similarity between two distributions—for example, the PDB structure and the human genome—was measured with Kendall's tau correlation coefficient τ [53]. For N pairs of measurements (xi, yi), each of them has N(N−1)/2 pairs of data points. τ is computed as:
con is defined as the number of pairs where (xi, xj) ranks the same as (yi, yj). dis is the number of pairs where (xi, xj) ranks the opposite to (yi, yj). ey is the number of pairs where yi = yj, and ex is the number of pairs where xi = xj.
Kendall's tau correlation coefficient ranges from −1.0 to 1.0. If two measurements have the similar ordering, it will be close to 1.0. The opposite ordering will give values close to −1.0. The coverage was also computed and defined as the ratio between the number of functional categories that have at least one structure representative and all functional categories.
Data access.
Data were warehoused in a single relational database where relations represent the mappings between the individual data sources. From a user's perspective, data appear in a multi-dimensional space. Each of the functional or structural categories is considered one dimension in the multi-dimensional space. A PDB structure or a genome sequence occupies a cube in this space. Any combination of two dimensions can be selected, and the distributions corresponding to the selected dimensions are calculated and displayed. The dimensions are organized in a hierarchal fashion according to their functional or structural taxonomies. Thus, data mining tasks such as drill-down or roll-up are supported. The database is accessible from http://function.rcsb.org:8080/pdb/function_distribution/index.html.
Supporting Information
Accession Numbers
The UniProt (http://www.pir.uniprot.org/) accession number for prostate specific kallikrein of serine proteases is P07288.
This work is supported by the Protein Data Bank through a multi-agency grant (NSF DBI 9814284), and PEB is supported in part by a National Institutes of Health grant (GM63208).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. LX and PEB conceived and designed the experiments, analyzed the data, and wrote the paper. LX performed the experiments and contributed reagents/materials/analysis tools.
A previous version of this article appeared as an Early Online Release on July 24, 2005 (DOI: 10.1371/journal.pcbi.0010031.eor).
Abbreviations
ECEnzyme Commission
GOGene Ontology
OMIMOnline Mendelian Inheritance in Man
PDBProtein Data Bank
==== Refs
References
Thornton JM Todd AE Milburn D Borkakoti N Orengo CA 2000 From structure to function: Approaches and limitations Nat Struct Biol 7 991 994 11104008
Brenner SE Levitt M 2000 Expectations from structural genomics Protein Sci 9 197 10739263
Portugaly E Linial M 2000 Estimating the probability for a protein to have a new fold: A statistical computational model Proc Natl Acad Sci U S A 97 5161 10792051
Westbrook J Feng Z Chen L Yang H Berman HM 2003 The Protein Data Bank and structural genomics Nucleic Acids Res 31 489 491 12520059
Peng K Obradovic Z Vucetic S 2004 Exploring bias in the Protein Data Bank using contrast classifiers Pac Symp Biocomput 2004 435 446
Brenner SE Chothia C Hubbard TJ 1997 Population statistics of protein structures: Lessons from structural classifications Curr Opin Struct Biol 7 369 9204279
Gerstein M 1998 How representative are the known structures of the proteins in a complete genome? A comprehensive structural census Fold Des 3 497 9889159
Liu J Rost B 2002 Target space for structural genomics revisited Bioinformatics 18 922 12117789
Devos D Valencia A 2000 Practical limits of function prediction Proteins: Struct Funct Genet 41 98 107 10944397
Hegyi H Gerstein M 1999 The relationship between protein structure and function: A comprehensive survey with application to the yeast genome J Mol Biol 288 147 164 10329133
Todd AE Orengo CA Thornton JM 2001 Evolution of function in protein superfamilies, from a structural perspective J Mol Biol 307 1113 1143 11286560
Wilson CA Kreychman J Gerstein M 2000 Assessing annotation transfer for genomics: Quantifying the relations between protein sequence, structure and function through traditional and probabilistic scores J Mol Biol 297 233 249 10704319
Martin ACR Orengo CA Hutchinson EG Jones S Karmirantzou M 1998 Protein folds and functions Structure 6 875 884 9687369
Rost B 2002 Enzyme function less conserved than anticipated J Mol Biol 318 595 608 12051862
Tian WD Skolnick J 2003 How well is enzyme function conserved as a function of pairwise sequence identity? J Mol Biol 333 863 882 14568541
Consortium TGO 2000 Gene Ontology: Tool for the unification of biology Nat Genet 25 25 29 10802651
Hegyi H Gerstein M 2001 Annotation transfer for genomics: Measuring functional divergence in multi-domain proteins Genome Res 11 1632 1640 11591640
Apic G Huber W Teichmann SA 2003 Multi-domain protein families and domain pairs: Comparison with known structures and a random model of domain recombination J Struct Funct Genomics 4 67 78 14649290
Littler SJ Hubbard SJ 2005 Conservation of orientation and sequence in protein domain-domain interactions J Mol Biol 345 1265 1279 15644220
Wang Z Moult J 2001 SNPs, protein structure, and disease Hum Mutat 17 263 270 11295823
Sunyaev S Ramensky V Bork P 2000 Towards a structural basis of human non-synonymous single nucleotide polymorphisms Trends Genet 16 198 200 10782110
Sander C 2000 Genomic medicine and the future of health care Science 287 1977 1978 10755952
Mirkovic N Marti-Renom MA Weber BL Sali A Monteiro AN 2004 Structure-based assessment of missense mutations in human BRCA1: Implications for breast and ovarian cancer predisposition Cancer Res 64 3790 3797 15172985
Carvalho AL Sanz L Barettino D Romero A Calvete JJ 2002 Crystal structure of a prostate kallikrein isolated from stallion seminal plasma: A homologue of human PSA J Mol Biol 322 325 337 12217694
Lin B White JT Lu W Xie T Utleg AG 2005 Evidence for the presence of disease-perturbed networks in prostate cancer cells by genomics and proteomic analyses: a systems approach to disease Cancer Res 65 3081 3091 15833837
Deshpande N Addess KJ Bluhm WF Merino-Ott JC Townsend-Merino W 2005 The RCSB Protein Data Bank: A redesigned query system and relational database based on the mmCIF schema Nucleic Acids Res 33 D233 D237 15608185
Marti-Renom MA Stuart AC Fiser A Sanchez R Melo F 2000 Comparative protein structure modeling of genes and genomes Annu Rev Biophys Biomol Struct 29 291 325 10940251
Baker D Sali A 2001 Protein structure prediction and structural genomics Science 294 93 96 11588250
Shuman S Lima CD 2004 The polynucleotide ligase and RNA capping enzyme superfamily of covalent nucleotidyltransferases Curr Opin Struct Biol 14 757 764 15582400
Scheeff ED Bourne PE 2005 Structural evolution of the protein kinase-like superfamily Submitted.
Sanchez R Pieper U Melo F Eswar N Marti-Renom MA 2000 Protein structure modeling for structural genomics Nat Struct Biol 7 986 990 11104007
Brenner SE 2000 Target selection for structural genomics Nat Struct Biol 7 967 969 11104002
Vitkup D Melamud E Moult J Sander C 2001 Completeness in structural genomics Nat Struct Biol 8 559 566 11373627
Todd AE Marsden RL Thornton JM Orengo CA 2005 Progress of structural genomics initiatives: An analysis of solved target structures J Mol Biol In press.
Yee A Pardee K Christendat D Savchenko A Edwards AM 2003 Structural proteomics: Toward high-throughput structural biology as a tool in functional genomics Acc Chem Res 36 183 189 12641475
Kyogoku Y Fujiyoshi Y Shimada I Nakamura H Tsukihara T 2003 Structural genomics of membrane proteins Acc Chem Res 36 199 206 12641477
Lundstrom K 2004 Structural genomics on membrane proteins: The MePNet approach Curr Opin Drug Discov Devel 7 342 346
Chandonia JM Brenner SE 2005 Implications of structural genomics target selection strategies: Pfam5000, whole genome, and random approaches Proteins 58 166 179 15521074
Hubbard T Barker D Birney E Cameron G Chen Y 2002 The Ensembl genome database project Nucleic Acids Res 30 38 41 11752248
Bairoch A Apweiler R Wu CH Barker WC Boeckmann B 2005 The Universal Protein Resource (UniProt) Nucleic Acids Res 33 D154 D159 15608167
F AS Madden TL Schaffer AA Zhang J Zhang Z 1997 Gapped BLAST and PSI-BLAST: A new generation of protein database search programs Nucleic Acids Res 25 3389 3402 9254694
Chen L Oughtred R Berman HM Westbrook J 2004 A target registration database for structural genomics projects Bioinformatics 20 2860 2862 15130928
Wootton JC 1994 Non-globular domains in protein sequences: Automated segmentation using complexity measures Comput Chem 18 269 285 7952898
Lupas AN Van Dyck M Stock J 1991 Predicting coiled coils from protein sequences Science 252 1162 1164 2031185
Krogh A Larsson B von Heijne G Sonnhammer ELL 2001 Predicting transmembrane protein topology with a hidden Markov model: Application to complete genomes J Mol Biol 305 567 580 11152613
Murzin AG Brenner SE Hubbard T Chothia C 1995 SCOP: A structural classification of proteins database for the investigation of sequences and structures J Mol Biol 247 536 540 7723011
Eddy SR 1998 Profile hidden Markov models Bioinformatics 14 755 763 9918945
Gough J Karplus K Hughey R Chothia C 2001 Assignment of homology to genome sequences using a library of hidden Markov models that represent all proteins of known structure J Mol Biol 313 903 919 11697912
Bateman A Coin L Durbin R Finn RD Hollich V 2004 The Pfam Protein Families Database Nucleic Acids Res 32 D138 D141 14681378
Li W Jaroszewski L Godzik A 2001 Clustering of highly homologous sequences to reduce the size of large protein database Bioinformatics 17 282 283 11294794
Bairoch A Apweiler R 2000 The SWISS-PROT protein sequence database and its supplement TrEMBL in 2000 Nucleic Acids Res 28 45 48 10592178
Wheeler DL Barret T Benson DA Bryant SH Canese K 2005 Database resources of the National Center for Biotechnology Information Nucleic Acids Res 33 D39 D45 15608222
Press WH Flannery BP Teukolsky SA Vetterling WT 1992 Numerical recipes in C: The art of scientific computing Cambridge Cambridge University Press 994 p.
|
16118666
|
PMC1188274
|
CC BY
|
2021-01-05 09:18:22
|
no
|
PLoS Comput Biol. 2005 Aug 19; 1(3):e31
|
utf-8
|
PLoS Comput Biol
| 2,005 |
10.1371/journal.pcbi.0010031
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1612125410.1371/journal.pgen.001001905-PLGE-RA-0033R2plge-01-02-08Research ArticleCancer BiologyDevelopmentGenetics/Gene FunctionGenetics/Disease ModelsDanio (zebrafish)Teleost FishesVertebratesAnimalsEukaryotesThe Zebrafish Mutants dre, uki, and lep Encode Negative Regulators of the Hedgehog Signaling Pathway Hedgehog Repressors in ZebrafishKoudijs Marco J 1den Broeder Marjo J. 1Keijser Astrid 1Wienholds Erno 1Houwing Saskia 1van Rooijen Ellen M. H. C. 1Geisler Robert 2van Eeden Fredericus J. M. 1*1 Hubrecht Laboratory, The Netherlands Institute for Developmental Biology, Utrecht, The Netherlands
2 Max-Planck-Institut für Entwicklungsbiologie, Tübingen, Germany
Mullins Mary EditorUniversity of Pennsylvania School of Medicine, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 19 8 2005 1 2 e1924 2 2005 23 6 2005 Copyright: © 2005 Koudijs et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Proliferation is one of the basic processes that control embryogenesis. To identify factors involved in the regulation of proliferation, we performed a zebrafish genetic screen in which we used proliferating cell nuclear antigen (PCNA) expression as a readout. Two mutants, hu418B and hu540A, show increased PCNA expression. Morphologically both mutants resembled the dre (dreumes), uki (ukkie), and lep (leprechaun) mutant class and both are shown to be additional uki alleles. Surprisingly, although an increased size is detected of multiple structures in these mutant embryos, adults become dwarfs. We show that these mutations disrupt repressors of the Hedgehog (Hh) signaling pathway. The dre, uki, and lep loci encode Su(fu) (suppressor of fused), Hip (Hedgehog interacting protein), and Ptc2 (Patched2) proteins, respectively. This class of mutants is therefore unique compared to previously described Hh mutants from zebrafish genetic screens, which mainly show loss of Hh signaling. Furthermore, su(fu) and ptc2 mutants have not been described in vertebrate model systems before. Inhibiting Hh activity by cyclopamine rescues uki and lep mutants and confirms the overactivation of the Hh signaling pathway in these mutants. Triple uki/dre/lep mutants show neither an additive increase in PCNA expression nor enhanced embryonic phenotypes, suggesting that other negative regulators, possibly Ptc1, prevent further activation of the Hh signaling pathway. The effects of increased Hh signaling resulting from the genetic alterations in the uki, dre, and lep mutants differ from phenotypes described as a result of Hh overexpression and therefore provide additional insight into the role of Hh signaling during vertebrate development.
Synopsis
In a screen aimed at finding genes that control proliferation in the zebrafish embryo, three mutants were identified. Mutants showed an increase in size of several structures including the brain, the retina, and the fins. Surprisingly, although size was increased in the embryos, adults remained small. Cloning of these genes revealed that increased Hedgehog signaling was at the basis of the phenotype, because all three genes encoded known repressors of the Hedgehog signaling pathway: Ptc2, Su(Fu), and Hip.
Hedgehog is known to play a role in proliferation. For instance, ectopic Hedgehog signaling can lead to several tumors including basal cell carcinoma and medulloblastoma. However, the phenotypes were still a surprise, because earlier experiments in zebrafish embryos suggested that activation should lead to patterning rather than proliferation defects. Current models of the pathway predict that these genes act independently to inhibit the signal but curiously, redundancy amongst these genes was not found, because triple mutants looked like the single mutants.
The conclusion is that weak activation of Hedgehog signaling can already lead to stimulation of growth in the absence of patterning defects, and that the Hedgehog signal is probably kept in check by the last inhibitor: Ptc1. A mutant for the ptc1 gene has recently been created and will put the model to the test.
Citation:Koudijs MJ, den Broeder MJ, Keijser A, Wienholds E, Houwing S, et al. (2005) The zebrafish mutants dre, uki, and lep encode negative regulators of the hedgehog signaling pathway. PLoS Genet 1(2): e19.
==== Body
Introduction
During development, proliferation is one of the key processes in the formation of an embryo, but how it is controlled spatiotemporally is still poorly understood. A tight regulation of proliferation is necessary during development and the remaining lifespan of an organism, as disrupted regulation might result in tumorigenesis. Several essential developmental signaling pathways are reported to control embryogenesis and many of these are involved in regulating proliferation in vertebrates and invertebrates. These basic developmental pathways all involve receptor ligation of highly conserved sets of secreted peptides like the TGF-β superfamily [1], FGF [2], Wnt [3], and Hedgehog (Hh) [4]. The Hh signaling pathway is highly conserved throughout evolution and has been documented to control proliferation [5]. In our current understanding, Hh proteins are expressed in a signaling cell, secreted and bound to the 12-transmembrane receptor Ptc (Patched) on a neighboring cell. Upon this binding, Ptc is thought to be internalized into endosomes where it is unable to repress the activity of Smo (smoothened) [6,7]. The signal is transmitted to the downstream proteins Cos2 (Costal2), Fused, Su(fu) (suppressor of fused), and one of the at least three members of the Gli family of zinc finger transcription factors [4]. In the presence of Hh, the Gli protein can be activated and transported to the nucleus where it activates genes mainly involved in patterning, proliferation, and cell structure [8]. Multiple genes are described to limit the activity of Hh signaling. Besides its own receptors Ptc1 and Ptc2, Hip (Hedgehog interacting protein) [9] is also expressed at the membrane in response to Hh activity. All three are involved in sequestering Hh to limit its effective range [10]. Further down the pathway, casein kinase I (CKI), glycogen synthase 3β (GSK3β), and protein kinase A (PKA) are involved in the processing or degradation of the Gli transcription factor [4]. The nuclear activity of Gli proteins is inhibited by Cos2 (Costal2) [11–13] and Su(fu) [14–18], which are both reported to be involved in tethering Gli in the cytoplasm, preventing overactivation of the pathway.
Hh signaling regulates multiple developmental processes in specific tissues in vertebrates and invertebrates [4]. In addition to the role of Hh during development, it is necessary to tightly regulate its activity during adulthood, where its aberrant activation is reported to predispose to malignant types of tumors in bone [19], pancreas [20], gut [21], skin, and brain [22,23]. Mutations in the negative regulator Su(fu) have been reported to predispose to medulloblastomas [24]. The formation of medulloblastomas has also been observed in patients suffering from Nevoid basal-cell carcinoma syndrome (NBCCS), where mutations in the Ptc protein have been identified [25].
Here we report about a forward genetic screen, performed to identify proliferation mutants. In this screen we used an in situ hybridization (ISH) approach in which we used PCNA expression levels as a specific marker for proliferation. Two zebrafish mutants, called hu418B and hu540A, were identified showing increased levels of proliferation at 40 h post fertilization (hpf). After 4 d, a combination of phenotypes was observed, similar to a known class of mutants from the Tübingen large-scale zebrafish screen [26]. These mutants, known as dre (dreumes), uki (ukkie), and lep (leprechaun), were identified based on their eye [27], ear [28], and pectoral fin [29] phenotypes. Both proliferation mutants from our screen are shown to be additional alleles of the uki mutation. In addition to the previously described phenotypes, adult dre mutants specifically show a disturbed regulation of chondrocyte differentiation in the branchial arches.
Positional cloning of this class of mutants identified mutations in negative regulators of Hh signaling. The dre, uki, and lep mutants encode the zebrafish homologs of the negative regulators Su(fu) [30], Hip [9], and Ptc2 [31]. As a result, the Hh signaling pathway is aberrantly activated. Treating mutant embryos with cyclopamine, a specific inhibitor of Hh signaling, rescues the phenotypes of uki and lep mutants. In an attempt to enhance the level of proliferation, double and triple mutants were generated showing equal levels of proliferation, compared to ukihu418B and ukihu540A, and no additive effect on the embryonic phenotypes. This suggests that additional regulators are still capable of inhibiting the Hh pathway, preventing further activation. In this report we describe the identification of the first vertebrate su(fu) and ptc2 mutants, and three nonsense mutations in Hip, all showing similar phenotypes. All mutants demonstrate the effects of aberrant activation of Hh signaling, which differs from all previously described Hh mutants in the zebrafish, which mainly show inhibition of Hh activity. This class of mutants will therefore contribute to the understanding of the role of Hh signaling during vertebrate development.
Results
A Genetic Screen for Proliferation Mutants
To identify mutants showing altered levels of embryonic proliferation, we used proliferating cell nuclear antigen (PCNA) as a readout, a commonly used marker in proliferation studies. PCNA is a protein that cooperates with DNA polymerase δ during DNA replication and repair [32]. We found that it was difficult to use the standard antibody (PC10, Novocastra Laboratories Ltd, Newcastle upon Tyne, United Kingdom) in a whole-mount procedure, but PCNA ISH gave robust results. Early in development, all cells are PCNA positive and expression gradually diminishes as patterning and differentiation proceed. At 40 hpf, a number of tissues are still positive, and these correspond to ones known to develop late, e.g., the pectoral fins, the gut, and the branchial arches. Furthermore, the cells that line the lumen of the neural tube, or cells in several brain folds that are derived from that region, are PCNA positive (Figure 1A). Finally, a ring of cells around the lens, called the ciliary marginal zone (CMZ), is PCNA positive (Figure 1B). The CMZ is known to contain stem cells that continue to generate retinal cells throughout life in lower vertebrates [33]. BrdU (bromo-2-deoxy-uridine) labeling experiments have shown that these regions indeed contain actively dividing cells (Figure 1C) [34]. Later in development, expression further decreases and at 120 hpf, PCNA expression becomes almost undetectable.
Figure 1 PCNA and Proliferation Patterns
(A) PCNA pattern as scored during the screen. In a dorsal view at 40 hpf, PCNA staining is observed in the medial and posterior part of the tectum, and in the cerebellum, the neural crest (arrowhead), and the pectoral fin (arrow).
(B) In a sideview (42 hpf), a ring of positive cells is visible around the lens.
(C) In a whole-mount BrdU labeling from day 3.5 to 4.5, similar regions are labeled indicating that PCNA RNA expression prefigures where BrdU will be incorporated.
(D and E) Sibling and hu418B mutants, respectively, showing increased PCNA labeling in the CMZ, but most prominently in the tectum.
We screened 100 mutagenized genomes for mutations that affect the level of PCNA expression. Several mutants with an increased expression were noted, but the majority showed typical degeneration/apoptosis phenotypes. Two mutants, however, showed an increase in PCNA expression, most prominently visible in the tectum, and did not show degeneration. These mutants, named hu418B and hu540A, show an increased expression of PCNA in the peripheral retina and the tectum (Figure 1D and 1E). These regions are known to be highly proliferative and are thought to contain stem cells. The increase can be observed at 36 hpf, but not at 24 hpf when a larger proportion of the embryo is PCNA positive. Furthermore, expression still undergoes a general reduction as the embryo ages. At 96 hpf, PCNA is difficult to detect in both mutant and wild-type embryos using an ISH approach. To exclude the possibility that a low level of apoptotic cell death is responsible for the increase of PCNA expression, we performed a whole-mount terminal transferase dUTP nick-end labeling (TUNEL) assay on hu418B mutants and siblings at 40 hpf. No difference in the level of apoptosis could be observed (data not shown), indicating that these mutants purely display increased levels of PCNA, independent of apoptosis.
Altered Level of Proliferation Affects Several Structures of the Developing Embryo
Morphological analysis of hu418B and hu540A mutants shows abnormalities that correspond with an increased level of cell proliferation. At 72 hpf, the volume of the head is increased (Figure 2A and 2B), mainly in the region of the tectum, where an increased level of proliferation is observed. Additionally, hu418B and hu540A mutant embryos show a reduced size of the pupil, whereas the overall size of the eye is unaffected (Figure 2C and 2D). Measurements revealed that both length and width of the pupil are significantly decreased (Figure 2E). However, the formation of the lens is normal (data not shown). This eye phenotype might result from the increased rate of proliferation in the CMZ. The retina is reported to grow as a result of several division steps of retinal stem cells in the CMZ [35]. An increased level of proliferation of these cells might cause the retina to overgrow the lens, reducing the size of the pupil. Additionally, the pectoral fins of the hu418B and hu540A mutants were enlarged (Figure 2F and 2G). Dorsoventrally the fins have approximately increased in size by 50% (0.01 < p < 0.02, n = 6), the area has increased by 65% (p < 0.001, n = 3) (Figure 2H and 2I).
Figure 2 Phenotypes of ukihu418B Mutant Embryos
(A and B) Lateral view of a wild-type (wt) and ukihu418B mutant.
(B) Showing an increased volume of the head. The size of the pupil is reduced in the ukihu18B mutant without affecting the size of the eye.
(C–E) The size of the pupil is reduced in the uki
hu18B mutant compared to wild-type, without affecting the size of the eye. Measurements revealed that the length and width of the pupil is significantly reduced in the uki418B mutant (n = 5, *p < 0,001).
(F–H) Pectoral fins showing the increased size of an uki418B mutant in which the dorsal/ventral (D/V) size of the pectoral fin is increased by 50%, (n = 3, *0.01 < p < 0.02).
(I) The anterior/posterior size is not significantly affected, but the fin area has increased by 65% (n = 3, p < 0.001).
(J and K) ukihu418B mutants lack the dorsolateral septum in the ear (arrowhead). Scale bar is 100 μm.
An additional phenotype is observed in the otic vesicle in the hu418B and hu540A mutants, in which the dorsolateral septum is missing (Figure 2J and 2K). However, the otoliths are correctly positioned, indicating that the anlage of the ear is correct.
This phenotypic combination was already observed in a class of mutants identified in a large-scale screen [26], covering the dre, uki, and lep mutants [27–29]. Complementation analysis revealed that hu418B and hu540A are additional alleles of uki (now referred to as ukihu418B and ukihu540A). Of this complementation group, the ukihu418B mutant shows the most consistent and strongest phenotype and is therefore used for further experiments. The morphological phenotype of dre and lep is slightly weaker and no clear difference in PCNA expression can be detected in the dre and lep mutants using an ISH approach.
Adult Mutants Show Additional Phenotypes
Raising homozygous uki, dre, and lep mutants demonstrates that approximately 10% of the uki and lep mutants reach 2 mo of age, and all die before the third month. Only dre mutants can be raised in significant numbers (50%) for 3 mo, with a maximum lifespan of 9 mo (5%). All mutants stay infertile and show a dwarfism phenotype [29]. This could be a result of the absence of growth hormone, which is secreted by the pituitary gland. However, sectioning an adult dre mutant revealed the presence of a pituitary gland (data not shown). The adenohypophysis secretes multiple hormones that are reported to play a role in the development of a dwarfism phenotype [36]. However, ISH experiments show that expression levels of growth hormone (GH), proopiomelanocortin (POMC), prolactin (PRL) and thyroid stimulating hormone (TSH) are not altered in dre mutants (data not shown).
Histological analysis of a 5-mo-old dre mutant (n = 4) and sibling (n = 3) reveals an additional phenotype concerning the gills of the adult fish. The gills contain branchial arches that are composed of multiple primary lamellae, formed by a stack of single chondrocytes (Figure 3A and 3B). To increase the area for sufficient oxygen uptake, these primary lamellae are branched into a large number of secondary lamellae (Figure 3A). However, in the dre mutant, the degree of branching to form secondary lamellae is severely diminished. The primary lamellae contain large clusters of cells, which morphologically resemble chondrocytes (Figure 3C). The strictly organized stacks of single chondrocytes are mainly absent. Occasionally, lines of chondrocytes appear to branch instead of the epithelium (Figure 3D). To investigate whether these clusters are indeed composed of chondrocytes we performed an Alcian Blue staining (Figure 3E and 3F) staining differentiated chondrocytes. In a wild-type fish the presence of differentiating chondrocytes could be confirmed in these single cell stacks. However, in the mutant, the clusters of cells were shown to be Alcian Blue negative (Figure 3F). On the other hand, the chondrocytes in the region where the primary lamellae are attached to the skeleton are Alcian Blue positive. This suggests that the chondrocytes in the gills are specifically affected in the dre mutant. One of the possibilities is that these Alcian Blue–negative cells are not able to properly differentiate during the development of the branchial arches. To examine whether these cells are in an earlier stage of cartilage formation, a Periodic Acid Schiff (PAS) staining was performed to detect ovotransferrin, a glycoprotein transiently expressed by differentiating hypertrophic chondrocytes, before they become Alcian Blue positive [37]. Both the mutant and the wild-type sections are negative for the PAS staining (data not shown). We therefore suggest that the chondrocyte-like cells in the mutant are prehypertrophic chondrocytes, which are arrested in their proliferative stage, which are therefore unable to finally differentiate into mature chondrocytes. This branchial arch phenotype appeared to be dre specific because primary lamellae in uki and lep mutants are able to branch (Figure 3G and 3H) and do not contain these foci of chondrocytes.
Figure 3 Patterning of the Branchial Arches in a 5-Mo-Old dre Mutant
(A and B) The strict organization of the brachial arch into primary (p) and secondary (s) lamellae in a wild-type situation (100× magnification). Higher magnification shows stacks of single chondrocytes in the primary lamellae.
(C and D) Sectioning of a dre mutant shows disturbed patterning, resulting in the absence of secondary lamellae and the presence of foci of chondrocyte-like cells in the primary lamellae (arrowsHE, hematoxylin and eosin stain; wt, wild-type.
(E and F) Alcian Blue staining reveals the presence of differentiated chondrocytes in the wild-type (wt) primary lamellae, but not in the dre mutant, indicating that the differentiation of these chondrocytes is affected (200×).
(G and H) Branchial arches of uki and lep mutants appear to be wild-type (wt).
The dre Locus is Encoding the Suppressor of Fused Protein
To identify the genes responsible for the observed phenotypes, we intended to positionally clone this class of mutants. The dre mutation was roughly mapped to linkage group (LG) 13 near z5395. Linkage analysis on 765 mutants reveals that the mutation was positioned close to z5395, leaving nine recombinants (0.6 cM) and z25745, leaving six recombinants (0.4 cM) (Figure 4A). Both markers mapped north of the mutation (referring to the MGH mapping panel at the Zebrafish Information Network at http://zfin.org). We identified an assembled contig of the Zv2 zebrafish genome assembly containing these markers, called ctg11890 (http://www.ensembl.org). We assumed that, based on the physical distance between these two markers, the mutation could be located on this contig. Several simple sequence length polymorphisms (SSLPs) mapping to this contig were tested for linkage, leaving zero recombinants with 11890.2A (Figure 4A). However, no marker was identified on this contig that would enclose the mutation on the south side. The closest marker was positioned in a region containing four predicted genes (Figure 4A). One of those, the β-mannosidase precursor gene, manba, was not considered to be a likely candidate. The other three candidates were screened for mutations by direct sequence analysis of all 28 predicted exons. This analysis revealed several silent mutations and one missense mutation in the third exon of Su(fu) [30], changing a threonine (ACG) to a lysine (AAG) at amino acid position 111 (Figure 4B), indicated as T111K. This residue is highly conserved in a stretch of eight amino acids: GFELTFRL, from bacteria (Bacillus circulans) to human (Figure 4C). No additional mutations affecting protein sequence could be identified in the other predicted coding regions, so we expected this mutation to be responsible for the dre phenotype. To test the hypothesis whether a loss of function of Su(fu) was able to phenocopy the mutants, we injected morpholino antisense oligonucleotides (MO) targeted against the predicted initiation codon of Su(fu) (Figure 5). The characteristic eye (Figure 5A and 5B) and ear (Figure 5E and 5F) phenotype of the dre mutant could be phenocopied effectively (75% phenocopy in two different strains, n = 58), in contrast to a control MO. Besides the eye and ear phenotype, the MO also induced a somite phenotype (Figure 5I and 5J). The normally chevron-shaped somites become partially flat, an effect previously described [30]. This could be due to a maternal component, which can be inhibited by the MO, enhancing the phenotype. Alternatively, the dretm146d allele may be a partial loss of function allele. To distinguish between these possibilities, we injected up to 25 ng of MO against a splice site, thereby affecting only the zygotic component of Su(fu). This results in a clear phenocopy of the dre mutant without a somite defect (>90% phenocopy in two different strains, n = 66). Additionally, injecting the same amount of splice MO into dre mutants, does not enhance the phenotype (>95% phenocopy, n = 71), suggesting this allele of Su(fu) to be a severe loss of function.
Figure 4 Positional Cloning of the dre Mutant
(A) Schematic representation of assembled contig 11890 of the Zv2 genome assembly. SSLP markers z5395 and z25745 and newly identified SSLPs 11890.2A and 11890.2 were closely linked with the dre locus. Remaining recombinants of a complete panel of 765 mutant embryos are indicated. Four genes were predicted in the region of marker 11890.2A that encode Su(fu), TRC8, ubiquitin conjugating enzyme E2, and β-mannosidase precursor protein.
(B) The dre mutation is a C to an A substitution, changing a threonine to a lysine.
(C) Multiple alignment of Su(fu) homologs revealed that the induced mutation changes an amino acid in a highly conserved region of Su(fu).
Figure 5 MO Injection Experiments against Su(fu), Hip, and Ptc2
(A) Dorsal view of the eye showing the lens in the eye chamber.
(B–D) Dorsal view of embryos injected with the indicated MOs, resulting in a phenocopy of dre, uki, and lep mutants.
(E) A wild-type ear showing the presence of the dorsolateral septum (arrow), which is not present after injections with the indicated MOs (F–H, arrow).
(I) Injections with control MOs against the initiation codon of Su(fu) results in chevron-shaped somites with an angle of 97°.
(J) Injection of MOs against Su(fu) results in a more obtuse angle of the somite (126°).
The uki and lep Loci Encode Negative Regulators of the Hh Signaling Pathway
The similarity of phenotypes between dre, uki, and lep mutants suggested that all the mutants encode negative regulators of Hh signaling. Positional cloning of the uki and lep mutations was therefore initiated by linkage analysis of SSLPs neighboring 14 candidate genes, all members of the Hh signaling pathway. The ukihu418B mutation is tightly linked with marker z13452 and z27361 on LG 1, enclosing Hip. Sequence analysis of all predicted coding sequence of Hip revealed a nonsense mutation in exon 5, changing a tyrosine to a stop codon at position 295 (Y295STOP) of the transcript encoding 694 amino acids (Figure 6A). Sequence analysis of the ukihu540A mutant identified a stop codon in exon 5 at position 285 (Y285STOP) of the Hip protein. The ukitc256d allele contained a premature stop codon at position 406 (Y406STOP) in exon 7 (Figure 6A). Amino acids 285, 295, and 406 of the zebrafish homolog of Hip corresponds to amino acids 293, 303, and 414 in human Hip. The identified mutation in the ukihu418B mutant should lead to a truncated protein without a membrane anchor, possibly resulting in a malfunctioning protein.
Figure 6 Premature Stop Codons Were Identified in the uki and lep Mutant in Hip and Ptc2, Respectively
(A) Schematic representation of the genomic organization of the Hip gene. All three alleles of the uki mutation contain premature stop codons positioned in exon 5 and exon 7.
(B) Representation of the protein structure of Ptc2 shows that the identified nonsense mutation is positioned after the sixth transmembrane domain, probably resulting in a malfunctioning protein.
(C) ISH experiments show that Ptc1 expression is increased in uki and lep (arrow), confirming the aberrant activation of the Hh signaling pathway.
Positional cloning of the leptj222 mutants was performed in a similar way. No recombination events were detected with marker z11948 and four newly identified SSLP markers on contig 10160 in genome assembly Zv2, containing Ptc2 (http://www.ensembl.org). The zebrafish Ptc2 protein has a transcript of 3,732 base pairs, encoding for 1,244 amino acids [31]. Exon sequencing identified a T to A substitution changing a tyrosine to a premature stop at position 590 (Figure 6B) directly after the sixth transmembrane domain. When this transcript is expressed, Ptc2 misses the second large extracellular domain known to be necessary for Hh binding and probably the inhibitory capability on Smo. We therefore expect this truncated protein to be a functional null.
To confirm that the identified mutations in these genes are responsible for the phenotypes, we injected wild-type embryos with an MO against the initiation codon of Hip and a splice MO for Ptc2. For the uki mutant, a clear phenocopy could be observed after 4 d, affecting the head, eyes, and ears (see Figure 5C and 5G) (60% phenocopy in two different strains, n = 64). Injection of wild-type embryos with 20 ng of Ptc2 splice MO resulted in a phenocopy of the ear and eye phenotype, however with a lower success rate (12/44; 28.5%) (see Figure 5D and 5H). This might be an effect of the positive feedback loop on Ptc2 when the Hh signaling pathway is activated, counteracting the efficiency of the MO. These experiments show that the uki and lep mutant phenotypes are indeed caused by Hip and Ptc2.
Loss of a negative regulator of Hh signaling should increase Hh activity, for which Ptc1 expression is generally used as a readout. An ISH was performed on dre, uki, and lep mutants, resulting in a slight increase in Ptc1 expression only for uki and lep mutants (Figure 6C). dre mutants do not show a significant increase in Ptc1 expression (data not shown).
Taken together, we conclude that aberrant activation of the Hh signaling pathway is responsible for the uki, dre, and lep mutant phenotypes.
Double and Triple Mutants Do Not Enhance the Phenotypes
Because the increase in proliferation can only significantly be detected in uki mutants, we initiated the generation of double and triple mutants in an attempt to enhance the level of proliferation. Current models suggest that the three genes should independently inhibit Hh signaling, therefore a higher level of proliferation could be expected. Analyzing PCNA expression in the progeny of two uki+/−/dre+/−/lep+/− carriers shows increased PCNA expression for a small subset, which upon genotyping turned out to be mainly uki mutants. Additionally, after sorting 96 genotyped embryos into all the possible genotypic combinations, it turned out that double and triple mutants do not show an obvious increase of the strength of the morphological phenotypes (Figure 7A–7E). The only morphological difference in double or triple mutants, compared to single mutants, comprises the ear (Figure 7F–7I). In a wild-type situation, the semicircular canal is formed after the ingrowth of the epithelial projections from the outline of the otic vesicle, which fuses in the center of the ear to form the ear lumen (Figure 7F). In the uki/lep double and uki/dre/lep triple mutants, all epithelial projections fail to grow toward the lumen of the ear (Figure 7G and 7I). These findings suggest that other negative regulators might still be present to prevent further activation of the Hh signaling pathway.
Figure 7 Phenotypic Analysis of dre/uki/lep Triple Mutants
(A) Wild-type (wt) embryo at 96 hpf.
(B–E) The indicated double and triple mutants do not show severe enhancement of the phenotype.
(F–I) dre/lep double mutants have an ear phenotype comparable with a single mutant. In the uki/lep and dre/uki/lep triple mutants, the epithelial projections (arrows) fail to grow out to fuse in the middle of the ear to form the ear lumen.
The uki and lep Mutants Can Be Rescued by Cyclopamine Treatment
To further prove that the described mutants are a result of increased Hh signaling, we attempted to inhibit Hh signaling, and thereby rescue the mutant phenotypes, by cyclopamine treatment. Cyclopamine is an inhibitor of Hh signaling acting on the level of Smo [38] at the initial stage of the signal transduction pathway. Treating lep mutants with limited amounts of cyclopamine (3 μM) clearly rescued the eye and ear phenotype. Genotyping of all apparent wild-type embryos identified the presence of lep mutants (Table 1). Wild-type embryos were unaffected using this concentration. However, a treatment using 25 μM of cyclopamine clearly mimicked the syu phenotype [39], showing the functionality of the cyclopamine.
Table 1 Rescuing Experiment Using Cyclopamine on uki, dre, and lep Mutants
Data are the numbers of embryos. Rescue experiment for uki, dre, and lep by inhibiting Hh activity using cyclopamine. lep mutants can be fully rescued using 3 μM cyclopamine, a concentration not affecting siblings. uki mutants can be rescued using 25 μM cyclopamine. However, the cyclopamine affects the development of the siblings, shown by the curly tail phenotype. Cyclopamine treatment did not rescue dre mutants.
aWeak mutant and curly tail.
The eye phenotype of uki mutant can be partially rescued using 10 or 15 μM cyclopamine, and fully rescued using 25 μM (Table 1). However, using 25 μM of cyclopamine, a subset of both siblings and mutants in the same clutch are affected, shown by a curly tail. Apparently, uki mutants are not protected against the effects of cyclopamine. This shows a limitation of using cyclopamine for rescuing the uki mutant, which is not the case for rescuing lep mutants, in which a much lower concentration is able to fully rescue without any side effects.
When dre mutants are treated with 25 μM cyclopamine, the mutant eye phenotype can be observed in combination with a curly tail, which is a result of the cyclopamine (data not shown). Increasing the cyclopamine concentration to 50 or 75 μM severely affected the development of all embryos, and therefore the eyes and ears could not be analyzed. Thus dre mutants are not protected against the effects of cyclopamine, but they also cannot be rescued by a cyclopamine treatment. The latter is expected because dre/Su(fu) acts downstream of Smo. These results emphasize that the phenotypes in this class of mutants are a result of an increased level of Hh activity.
Discussion
Proliferation Is Increased in the hu418B and hu540A Mutants
In our forward genetic screen, we were able to identify two proliferation mutants based on altered expression levels of PCNA and detected by an ISH approach. The identified mutants show an increase in the level of PCNA expression after 40 hpf, which is ectopically expressed in the developing tectum and in the CMZ of the eye. Unfortunately, no mutants were identified with a decreased proliferation rate. This could be explained by the fact that this screen covered only 1%–10% (100 genomes) of the zebrafish genome, leaving several additional genes to be identified in a larger screen. The limited amount of identified mutants suggests that there might be a high level of redundancy in controlling proliferation. Mutants showing an altered level of proliferation associated with increased apoptosis were excluded due to their frequent occurrence. We speculate that an increase in the amount of apoptosis results in an increase in proliferation/PCNA expression as part of a wounding/repair response [40], on which altered levels of proliferation are a secondary effect. Possibly, this secondary defect has obscured some interesting early defects.
TUNEL experiments have shown that the increase in proliferation in the hu418B and hu540A mutants is not associated with increased apoptosis. The increased PCNA expression could therefore be a result of impaired regulation of proliferation. The phenotypes observed after 4 d are similar to a previously described class of mutants identified in the large-scale Tübingen zebrafish screen [26], containing dre, uki, and lep. Complementation analysis revealed that both hu418B and hu540A are additional alleles of the uki mutation. dre and lep mutants are weaker than uki as judged from morphology and do not show a comparable increase of PCNA expression. Possibly, the increase in proliferation in the uki mutant might reflect a specific function for Hip, but considering the overall morphological similarity of the mutants, it is more likely that the difference is due to other factors (see below).
Aberrant Hh Activation Is Responsible for the dre, uki, and lep Mutants
We show that dre, uki, and lep encode components of the Hh signaling pathway. The lesion in the dre mutant is a missense mutation in the su(fu) gene. The mutation, changing a threonine to a lysine, has been proposed as a potential PKC target site [41]. It is positioned in a highly conserved N-terminal region shown to be involved in binding the Gli protein [42]. Crystal structure analysis revealed that this threonine is buried and therefore suggested to be unimportant for the activity of Su(fu) [42]. However, our data indicate that this residue is crucial for the proper functioning of Su(fu), therefore it might become accessible for certain kinases due to conformational changes.
By injecting MOs against Su(fu), the mutant phenotype of dre could be copied, confirming that the dre locus encodes Su(fu). However, Su(fu) MOs against the initiation codon of Su(fu) induce a somite phenotype [30], which could be explained in a situation in which the Su(fu) MO also affects a maternal contribution. This is confirmed by the finding that an MO against a splice site does not result in a somite phenotype. Additionally, the phenotypes of dre can not be enhanced by the splice MO, suggesting this allele of Su(fu) to be a strong loss of function or a null. The similarity of the phenotypes within this mutant class suggested that the uki and lep mutants are also a result of aberrant activation of Hh. Linkage analysis of markers near multiple candidate genes confirmed this. Premature stop codons in Hip and Ptc2 were identified to be responsible for the uki and lep mutants, respectively, which was confirmed by the MO-induced phenocopies. In the leptj222 mutant, the identified premature stop is positioned at amino acid 590 directly after the sixth putative transmembrane domain [31], only producing the N-terminal half of the protein. In Drosophila, multiple alleles of Ptc have been analyzed, showing that expression of either the N- or C-terminal half alone will abolish its function [43]. We therefore expect that this severe truncation will abolish Ptc2 protein function in the mutant. The three nonsense mutations in the Hip protein result in all cases in a comparable phenotype, suggesting these alleles to be nulls. As a result, the Hh signaling pathway will be aberrantly activated, confirmed by the increase in Ptc1 expression in uki and lep mutants. However, the effect of the overactivation of the Hh pathway is subtle compared to Hh overexpression data. This might be a result of a restricted expression pattern of these negative regulators. ISH experiments showed that Su(fu) is ubiquitously expressed until 24 hpf [30] and becomes more anteriorly restricted at 42 hpf (data not shown). The expression pattern of Ptc2 is generally overlapping Ptc1 expression, with some minor differences [31]. Hip transcripts can be detected in the adaxial cells in the developing trunk and in the head, generally resembling the expression pattern of Ptc1 (Figure S1A). Within the developing trunk, two rows of adaxial cells are shown to be Hip positive at 24 hpf (Figure S1A and S1B). Hip expression is reduced in uki mutants at 24 hpf, which is probably due to nonsense-mediated decay (data not shown). At 40 hpf the pectoral fins, some branchial arches, and the tectum opticum are Hip positive, linking the expression pattern with the observed embryonic phenotypes (Figure S1C and S1D). Combining these expression patterns suggests that the subtle phenotypes observed in these mutants are not a result of a restricted pattern of one of these negative regulators, but are probably due to other negative regulators, most likely Ptc1, preventing further activation of the pathway.
Phenotypic Consequences of Aberrant Hh Activation
The ukihu418B and ukihu540A mutants were picked up showing an increased level of proliferation in the developing brain. Hh activity is reported to be involved in the proliferation of cells in the central nervous system [44–49]. The increase in the volume of the head of the ukihu418B mutant is therefore in line with previous studies in which the growth of the brain is shown to be regulated by the activation of Shh-Gli1 signaling [50,51].
All the affected structures and tissues in the described mutants are known to be under control of Hh signaling during development. Hh signaling is one of the key regulators in the development of the eye, in which the formation of the retina is driven by a wave of Hh signal, secreted by the cells of the ganglion cell layer [52]. As a result, Hh controls proliferation of multiple cell types of the eye like photoreceptors and glia [53]. All described mutants show a decreased size of the pupil, which might be due to an overgrowing activity of the cells of the retina. Therefore the lens is not visible from a dorsal view, but no defects are observed in the lens itself.
The increased fin size in the ukihu418B mutant embryos could be linked to impaired Hh signaling. It is the opposite of the phenotype of syu mutants, in which finbuds are established, but fail to grow out, due to the absence of Shh signaling [39]. The observed phenotypes in the dre, uki, and lep mutants can therefore be linked to aberrant activation of the Hh signaling pathway.
Surprisingly, dre, uki, and lep mutants can be grown for several months, but remain small and are infertile. One explanation for the dwarfism phenotype involves the absence of growth hormone secreted by the pituitary gland. The formation of the pituitary gland is reported to be regulated by Hh signaling [54,55], indicating that its functioning might be hampered in these mutants and resulting in the observed small phenotype. However, sectioning of a dre mutant embryo revealed that the pituitary gland is morphologically present. The secreted hormones of the adenohypophysis are reported to be involved in the development of a dwarfism phenotype [36]. However, dre mutants do not show obvious altered levels of POMC, TSH, PRL, and GH expression, indicating that the formation of the adenohypophysis is not affected in this class of mutants. Currently, we are examining a potential role for the IGF signaling pathway in the development of the dwarfism phenotype.
The dre mutants show an abnormality in the development of the branchial arches. Normally, the primary lamellae are strictly patterned and intensely branched into secondary lamellae. Occasionally, the chondrocyte-like cells appear to form the secondary lamellae itself instead of the branching of epithelial cells. Branching of the mammalian lungs is reported to be regulated by Shh activity [56], in which increased Shh activity disrupts branching and increases the level of proliferation. The primary lamellae gain their rigidity by stacks of single chondrocytes. However, in the dre mutant the primary lamellae contain large clusters of prehypertrophic chondrocytes, which might be unable to start the differentiation process. The formation of cartilage is reported to be tightly regulated by the action of Indian hedgehog (Ihh) and parathyroid hormone-related protein (PTHrP) [57,58]. In this process the amount of Ihh acts like a sensor, thereby limiting the group of cells that are stimulated to enter the differentiation stage [57]. This might be deregulated in the dre mutant, in which an aberrant activation of Ihh signaling increases the amount of PTHrP, thereby preventing hypertrophic differentiation. The prehypertrophic chondrocytes therefore remain in their proliferative stage and might form the observed clusters in the primary lamellae. The branchial arch phenotype has not been observed in uki/Hip and lep/Ptc2 mutants, suggesting that this is a unique function for Su(fu). Therefore Su(Fu) could be modulating signals via Ptc1 as well. This is in agreement with results on Su(Fu) morphants that mimic Ptc1 morphants in their somite phenotype [30]. If the Su(fu) allele is a strong loss of function and affects signals via Ptc1 and Ptc2, why is the phenotype not any stonger? In addition to rescue by maternal protein, the role of Su(fu) in Hh inhibition may be accessory rather than absolutely central. This has been shown in Drosophila, in which complete inactivation of Su(Fu) does not lead to a full Hh overactivation phenotype [59].
Complex Regulation of Hh Activity
We have shown that aberrant activation of the Hh signaling pathway is responsible for the dre, uki, and lep mutants. Nevertheless, none of the mutants that was identified shows a typical phenotype described for aberrant Hh activation as was obtained by the overexpression of dnPKA or Shh [30,60]. Surprisingly, triple mutants of su(fu), Hip, and ptc2 still do not show a further increased PCNA expression or a strongly enhanced phenotype. This could be explained in a scenario in which a slight activation of the Hh pathway exceeds a certain threshold upon which Ptc1 will be expressed via an autoregulatory loop, preventing further activation of the pathway. Inhibiting Ptc1 functioning in these triple mutants could probably result in the expected Hh-overexpression phenotypes. Ptc1/Ptc2 double morphants were described, confirming this idea [30].
The ability of cyclopamine to rescue the uki and lep mutants is in line with expectations. Cyclopamine acts on the level of Smo and can revert the effect of upstream components such as Hip and Ptc2. The ability of cyclopamine to rescue the effects of mutations in Ptc is also documented for cell lines [38]. Because Su(fu) acts downstream of Smo (the point where cyclopamine acts), it is likely to be independent of the presence of an upstream signal. Similar results have been reported in a system with Gli2 overexpression, in which cyclopamine was unable to revert the effects [38]. Indeed, we find that dre mutants cannot be rescued by cyclopamine treatment.
Aberrant postnatal activation of the Hh signaling pathway is implicated in various types of neuronal and epithelial tumors. However, no obvious tumors have been observed in the mutants and heterozygotes. Future experiments using additional mutants, like a p53 knockout, might induce tumorigenesis in a background with aberrant Hh activity.
Materials and Methods
Strains and screening methods.
ukitc256d, leptj222, and dretm146d stocks were obtained from the Max Planck Institute for Developmental Biology stock center in Tübingen, Germany. ENU mutagenesis was performed on TL males as described [61]. F1 families were generated and interbred to obtain F2 families. Inbreeding generated F3 embryos for screening. Embryos were incubated in PTU and dechorionated by pronase treatment according to standard protocols (http://zfin.org). Staging of embryos was according to Kimmel et al. [62]. Embryos were fixed in 4% paraformaldehyde/PBS at 40 hpf. During screening, PCNA whole-mount ISH (WISH) was performed on an Abimed 96-wells ISH robot (Intavis Bioanalytical Instruments, Cologne, Germany; settings available on request). Mutants that were detected by morphological screening at 24 hpf were processed separately for WISH, along with two wild-types as controls. In addition, morphological screening was performed at 72 hpf on a duplicate clutch.
In situ hybridization.
Manual ISH was carried out as described [63]. An antisense probe for PCNA was generated by linearizing EST clone fc43g05 (MPMGp609L0932, RZPD Deutsches Ressourcenzentrum für Genomforschung, Berlin, Germany; http://www.rzpd.de), using SalI and transcription using SP6 polymerase. Ptc1 probe synthesis was performed according to Concordet et al. [64]. ISH for POMC, TSH, GH, and PRL was performed according to Herzog et al. [65]. A 2-kb fragment of Hip was amplified from cDNA, using primers 5′-AATTTGTGCTCTTGTTAGCC-3′ and 5′-AGTGAGGTCCAGCAGGTAAG-3′, cloned and subsequently transcribed.
TUNEL assay and BrdU labeling.
To determine the amount of apoptosis, a whole-mount TUNEL assay was performed on 40 hpf embryos as described [66]. The presence of mutants was confirmed by genotyping the analyzed embryos. BrdU labeling was performed according to a previous report [67].
Histology.
Adult fish were fixed in 4% paraformaldehyde at 4 °C for 4 d and subsequently decalcified in 0.25M EDTA (pH 8) for 2 d. Paraffin sections (6 μm thick) were stained with eosin, Alcian Blue, or PAS in combination with hematoxylin using standard protocols.
Measurements.
The size of the fins and pupils was determined on a Zeiss Axioplan microscope (Carl Zeiss, Jena, Germany) using a micrometer. The area of the pectoral fins was measured by determining the amount of pixels of a scanned photograph using Paint Shop Pro version 5 (Jasc Software, Corel Corporation, Ottawa, Ontario, Canada).
Genetic mapping and positional cloning of dre.
The rough genome mapping of dretm146d to LG 13 was performed by bulked segregation analysis of F2 embryos and genome scanning with SSLPs [68]. To fine map the mutation, a mapping strain was created by crossing a dre/+ male in a Tübingen background to a wild-type WIK female. F2 fish carrying the dre mutation were crossed; 765 mutant F3 embryos were collected and genomic DNA was isolated. In total, 35 SSLPs from the Massachusetts General Hospital [69] were used for positional cloning on LG 13 in the region around 40 cM. Marker names and primer sequences can be obtained on request. Additionally, 12 SSLPs were identified on assembled contig 11890 of version Zv2 of the zebrafish genome assembly (http://www.ensembl.org) that contained the two most closely linked markers (z5395 and z25743) from the MGH database. Linkage analysis of these newly identified markers was performed to enclose the mutation. All predicted exon sequences south (corresponding to the MGH map) of the most closely linked SSLP marker were amplified and sequenced using DyeNamic ET (Amersham Biosciences, Little Chalfont, United Kingdom) protocols. All used zebrafish genome sequence data were produced by the Sanger Institute (http://www.sanger.ac.uk).
Candidate gene approach to positional clone uki and lep.
Mapping crosses were generated crossing a female ukihu418B or leptj222 carrier in Tübingen background, with a WIK or TL male, respectively. Fourteen known regulators of the Hh signaling pathway were selected, and homologs were identified in the Zv2 zebrafish genome build at http://www.ensembl.org. A likely map position of the identified contigs was determined using the comparative map at http://zfin.org and http://www.sanger.ac.uk. Forty SSLPs surrounding these candidates were analyzed for linkage on 10 uki or lep mutants and two siblings. In the case of linkage to a candidate gene, all predicted exons were amplified and sequenced. Predicted Hip exons were obtained using EST clone fc52e12 and additional homology-based assembly of the transcript.
MO antisense knockdown.
MOs (Gene Tools, LLC, Philomath, Oregon, United States) were designed against the predicted initiation codon or splice donor site of Su(fu), Hip, and Ptc2, along with a five-mismatch MO as a control. Their sequences are as follows (mismatches in lower case): Su(fu) MO: 5′-GCTGCTAGGCCGCATCTCATCCATC-3′, Su(fu) mismatched control: 5′-GCTGgTAcGCCGaATCTgATaCATC-3′, Su(fu) splice MO: 5′-TGACATTCTTACTCGTGAACTCTGT-3′; Hip MO: 5′-AATGCTTCATTTTTGCAGGGATGA-3′, Hip mismatched control: 5′-AATGgTTgATTTTTcCAGcGATcA-3′, Ptc2 splice MO: 5′-CTAGCAAATAAGCCATACCTGTTGT-3′, Ptc2 control 5-mm–splice MO: 5′-CTAcCAAAaAAcCCATAaCTGaTGT-3′. MOs were diluted in water to a stock concentration of 50 ng/nl. Ranges from 0.33 to 25 ng of MO were injected into one- to four-cell stage TL or ABxTL embryos and screened for the expected phenotypes 4 d after fertilization.
Cyclopamine treatment.
Progeny from a cross of two uki, dre, or lep heterozygotes were grown in embryo medium in the presence of various concentrations of cyclopamine (from a 10 mM stock in 96% ethanol), ranging from 2 to 75 μM, administered at 5.5 hpf. Control embryos were treated with equal amounts of 96% ethanol. Genotyping was performed to identify mutants in the clutch of treated embryos.
Supporting Information
Figure S1 Expression Pattern of Hip in Wild-Type Embryos
(A and B) At 24 hpf, Hip is expressed in the brain and in two lines of adaxial cells in the developing tail.
(C and D) At 42 hpf, Hip expression is reduced in the somites. The mid-hindbrain boundary, some branchial arches, and the pectoral fins are Hip positive. Hip expression is also detected in the tectum (arrowhead).
(9.8 MB TIF)
Click here for additional data file.
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank) accession numbers for the described ESTs, genes, and proteins are Bacillus circulans, (CAD41946), zebrafish Hip EST clone fc52e12 (AI878265), PCNA EST clone fc43g05 (AI794381), su(fu) (NP_958466), and ptc2 (CAB39726), and human Hip (AAH25311).
We would like to thank Peter Lanser, Chris Jopling, Carina van Rooijen, Gerwen Lammers, and Marit Kosters for help during the screen. We thank Dr. Dana Jongejan-Zivkovic for advice and suggestions, and J. Korving, Harry Begthel, and Evelyn Groot for the sectioning experiments. Probes for the analysis on the adenohypophysis were kindly provided by M. Hammerschmidt. RG was supported by the German Human Genome Project (DHGP Grant 01 KW 9919), and MJK was supported by NWO genomics grant 050-10-024.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. MJK and FJMV conceived and designed the experiments. MJK, MJDB, AK, SH, EW, EMHCV, and FJMV performed the experiments. MJK, MJDB, SH, and FJMV analyzed the data. AK, SH, and RG contributed reagents/materials/analysis tools. MJK wrote the paper.
Abbreviations
CMZciliary marginal zone
ENUethyl-nitrosourea
GHgrowth hormone
HhHedgehog
hpfhours post fertilization
IhhIndian hedgehog
ISHin situ hybridization
LGlinkage group
MOmorpholino antisense oligonucleotide
PASPeriodic Acid Schiff
PCNAproliferating cell nuclear antigen
POMCproopiomelanocortin
PRLprolactin
PTHrPparathyroid hormone-related protein
PTUphenylthiourea
SSLPsimple sequence length polymorphism
TSHthyroid stimulating hormone
TUNELterminal transferase dUTP nick-end labeling
==== Refs
References
Massague J Chen YG 2000 Controlling TGF-beta signaling Genes Dev 14 627 644 10733523
Szebenyi G Fallon JF 1999 Fibroblast growth factors as multifunctional signaling factors Int Rev Cytol 185 45 106 9750265
Wodarz A Nusse R 1998 Mechanisms of Wnt signaling in development Annu Rev Cell Dev Biol 14 59 88 9891778
Ingham PW McMahon AP 2001 Hedgehog signaling in animal development: Paradigms and principles Genes Dev 15 3059 3087 11731473
Wetmore C 2003 Sonic hedgehog in normal and neoplastic proliferation: Insight gained from human tumors and animal models Curr Opin Genet Dev 13 34 42 12573433
Denef N Neubuser D Perez L Cohen SM 2000 Hedgehog induces opposite changes in turnover and subcellular localization of patched and smoothened Cell 102 521 531 10966113
Incardona JP Gruenberg J Roelink H 2002 Sonic hedgehog induces the segregation of patched and smoothened in endosomes Curr Biol 12 983 995 12123571
Yoon JW Kita Y Frank DJ Majewski RR Konicek BA 2002 Gene expression profiling leads to identification of GLI1-binding elements in target genes and a role for multiple downstream pathways in GLI1-induced cell transformation J Biol Chem 277 5548 5555 11719506
Chuang PT McMahon AP 1999 Vertebrate Hedgehog signalling modulated by induction of a Hedgehog-binding protein Nature 397 617 621 10050855
Nybakken K Perrimon N 2002 Hedgehog signal transduction: Recent findings Curr Opin Genet Dev 12 503 511 12200154
Robbins DJ Nybakken KE Kobayashi R Sisson JC Bishop JM 1997 Hedgehog elicits signal transduction by means of a large complex containing the kinesin-related protein costal2 Cell 90 225 234 9244297
Sisson JC Ho KS Suyama K Scott MP 1997 Costal2, a novel kinesin-related protein in the Hedgehog signaling pathway Cell 90 235 245 9244298
Wang G Amanai K Wang B Jiang J 2000 Interactions with Costal2 and suppressor of fused regulate nuclear translocation and activity of cubitus interruptus Genes Dev 14 2893 2905 11090136
Ding Q Fukami S Meng X Nishizaki Y Zhang X 1999 Mouse suppressor of fused is a negative regulator of sonic hedgehog signaling and alters the subcellular distribution of Gli1 Curr Biol 9 1119 1122 10531011
Kogerman P Grimm T Kogerman L Krause D Unden AB 1999 Mammalian suppressor-of-fused modulates nuclear-cytoplasmic shuttling of Gli-1 Nat Cell Biol 1 312 319 10559945
Methot N Basler K 2000 Suppressor of fused opposes hedgehog signal transduction by impeding nuclear accumulation of the activator form of Cubitus interruptus Development 127 4001 4010 10952898
Murone M Luoh SM Stone D Li W Gurney A 2000 Gli regulation by the opposing activities of fused and suppressor of fused Nat Cell Biol 2 310 312 10806483
Dunaeva M Michelson P Kogerman P Toftgard R 2003 Characterization of the physical interaction of Gli proteins with SUFU proteins J Biol Chem 278 5116 5122 12426310
Jones KB Morcuende JA 2003 Of hedgehogs and hereditary bone tumors: Re-examination of the pathogenesis of osteochondromas Iowa Orthop J 23 87 95 14575257
Thayer SP di Magliano MP Heiser PW Nielsen CM Roberts DJ 2003 Hedgehog is an early and late mediator of pancreatic cancer tumorigenesis Nature 425 851 856 14520413
Berman DM Karhadkar SS Maitra A Montes De Oca R Gerstenblith MR 2003 Widespread requirement for Hedgehog ligand stimulation in growth of digestive tract tumours Nature 425 846 851 14520411
Vorechovsky I Unden AB Sandstedt B Toftgard R Stahle-Backdahl M 1997 Trichoepitheliomas contain somatic mutations in the overexpressed PTCH gene: Support for a gatekeeper mechanism in skin tumorigenesis Cancer Res 57 4677 4681 9354420
Vorechovsky I Tingby O Hartman M Stromberg B Nister M 1997 Somatic mutations in the human homologue of Drosophila patched in primitive neuroectodermal tumours Oncogene 15 361 366 9233770
Taylor MD Liu L Raffel C Hui CC Mainprize TG 2002 Mutations in SUFU predispose to medulloblastoma Nat Genet 31 306 310 12068298
Fujii K Kohno Y Sugita K Nakamura M Moroi Y 2003 Mutations in the human homologue of Drosophila patched in Japanese nevoid basal cell carcinoma syndrome patients Hum Mutat 21 451 452
Haffter P Granato M Brand M Mullins MC Hammerschmidt M 1996 The identification of genes with unique and essential functions in the development of the zebrafish, Danio rerio
Development 123 1 36 9007226
Heisenberg CP Brand M Jiang YJ Warga RM Beuchle D 1996 Genes involved in forebrain development in the zebrafish, Danio rerio
Development 123 191 203 9007240
Whitfield TT Granato M van Eeden FJ Schach U Brand M 1996 Mutations affecting development of the zebrafish inner ear and lateral line Development 123 241 254 9007244
van Eeden FJ Granato M Schach U Brand M Furutani-Seiki M 1996 Genetic analysis of fin formation in the zebrafish, Danio rerio
Development 123 255 262 9007245
Wolff C Roy S Ingham PW 2003 Multiple muscle cell identities induced by distinct levels and timing of hedgehog activity in the zebrafish embryo Curr Biol 13 1169 1181 12867027
Lewis KE Concordet JP Ingham PW 1999 Characterisation of a second patched gene in the zebrafish Danio rerio and the differential response of patched genes to Hedgehog signalling Dev Biol 208 14 29 10075838
Lu X Tan CK Zhou JQ You M Carastro LM 2002 Direct interaction of proliferating cell nuclear antigen with the small subunit of DNA polymerase delta J Biol Chem 277 24340 24345 11986310
Perron M Harris WA 2000 Retinal stem cells in vertebrates Bioessays 22 685 688 10918298
Mueller T Wullimann MF 2002 BrdU-, neuroD (nrd)- and Hu-studies reveal unusual non-ventricular neurogenesis in the postembryonic zebrafish forebrain Mech Dev 117 123 135 12204253
Wetts R Serbedzija GN Fraser SE 1989 Cell lineage analysis reveals multipotent precursors in the ciliary margin of the frog retina Dev Biol 136 254 263 2478403
Nica G Herzog W Sonntag C Hammerschmidt M 2004 Zebrafish pit1 mutants lack three pituitary cell types and develop severe dwarfism Mol Endocrinol 18 1196 1209 14988429
Gentili C Bianco P Neri M Malpeli M Campanile G 1993 Cell proliferation, extracellular matrix mineralization, and ovotransferrin transient expression during in vitro differentiation of chick hypertrophic chondrocytes into osteoblast-like cells J Cell Biol 122 703 712 8393014
Taipale J Chen JK Cooper MK Wang B Mann RK 2000 Effects of oncogenic mutations in Smoothened and Patched can be reversed by cyclopamine Nature 406 1005 1009 10984056
Schauerte HE van Eeden FJ Fricke C Odenthal J Strahle U 1998 Sonic hedgehog is not required for the induction of medial floor plate cells in the zebrafish Development 125 2983 2993 9655820
Smith PG Liu M 2002 Impaired cutaneous wound healing after sensory denervation in developing rats: Effects on cell proliferation and apoptosis Cell Tissue Res 307 281 291 11904764
Stone DM Murone M Luoh S Ye W Armanini MP 1999 Characterization of the human suppressor of fused, a negative regulator of the zinc-finger transcription factor Gli J Cell Sci 112 ( Pt 23): 4437 4448 10564661
Merchant M Vajdos FF Ultsch M Maun HR Wendt U 2004 Suppressor of fused regulates Gli activity through a dual binding mechanism Mol Cell Biol 24 8627 8641 15367681
Johnson RL Milenkovic L Scott MP 2000 In vivo functions of the patched protein: Requirement of the C terminus for target gene inactivation but not Hedgehog sequestration Mol Cell 6 467 478 10983992
Goodrich LV Milenkovic L Higgins KM Scott MP 1997 Altered neural cell fates and medulloblastoma in mouse patched mutants Science 277 1109 1113 9262482
Wallace VA 1999 Purkinje-cell-derived Sonic hedgehog regulates granule neuron precursor cell proliferation in the developing mouse cerebellum Curr Biol 9 445 448 10226030
Wechsler-Reya RJ Scott MP 1999 Control of neuronal precursor proliferation in the cerebellum by Sonic Hedgehog Neuron 22 103 114 10027293
Dahmane N Ruiz-i-Altaba A 1999 Sonic hedgehog regulates the growth and patterning of the cerebellum Development 126 3089 3100 10375501
Rowitch DH S-Jacques B Lee SM Flax JD Snyder EY 1999 Sonic hedgehog regulates proliferation and inhibits differentiation of CNS precursor cells J Neurosci 19 8954 8965 10516314
Kenney AM Rowitch DH 2000 Sonic hedgehog promotes G(1) cyclin expression and sustained cell cycle progression in mammalian neuronal precursors Mol Cell Biol 20 9055 9067 11074003
Dahmane N Sanchez P Gitton Y Palma V Sun T 2001 The Sonic Hedgehog-Gli pathway regulates dorsal brain growth and tumorigenesis Development 128 5201 5212 11748155
Palma V Ruiz i Altaba A 2004 Hedgehog-GLI signaling regulates the behavior of cells with stem cell properties in the developing neocortex Development 131 337 345 14681189
Neumann CJ Nuesslein-Volhard C 2000 Patterning of the zebrafish retina by a wave of sonic hedgehog activity Science 289 2137 2139 11000118
Jensen AM Wallace VA 1997 Expression of Sonic hedgehog and its putative role as a precursor cell mitogen in the developing mouse retina Development 124 363 371 9053312
Sbrogna JL Barresi MJ Karlstrom RO 2003 Multiple roles for Hedgehog signaling in zebrafish pituitary development Dev Biol 254 19 35 12606279
Herzog W Zeng X Lele Z Sonntag C Ting JW 2003 Adenohypophysis formation in the zebrafish and its dependence on sonic hedgehog Dev Biol 254 36 49 12606280
Pepicelli CV Lewis PM McMahon AP 1998 Sonic hedgehog regulates branching morphogenesis in the mammalian lung Curr Biol 8 1083 1086 9768363
Vortkamp A Lee K Lanske B Segre GV Kronenberg HM 1996 Regulation of rate of cartilage differentiation by Indian hedgehog and PTH-related protein Science 273 613 622 8662546
Lanske B Karaplis AC Lee K Luz A Vortkamp A 1996 PTH/PTHrP receptor in early development and Indian hedgehog-regulated bone growth Science 273 663 666 8662561
Preat T 1992 Characterization of Suppressor of fused, a complete suppressor of the fused segment polarity gene of Drosophila melanogaster
Genetics 132 725 736 1468628
Hammerschmidt M Bitgood MJ McMahon AP 1996 Protein kinase A is a common negative regulator of Hedgehog signaling in the vertebrate embryo Genes Dev 10 647 658 8598293
van Eeden FJ Granato M Odenthal J Haffter P 1999 Developmental mutant screens in the zebrafish Methods Cell Biol 60 21 41 9891329
Kimmel CB Ballard WW Kimmel SR Ullmann B Schilling TF 1995 Stages of embryonic development of the zebrafish Dev Dyn 203 253 310 8589427
Thisse C Thisse B Schilling TF Postlethwait JH 1993 Structure of the zebrafish snail1 gene and its expression in wild-type, spadetail and no tail mutant embryos Development 119 1203 1215 8306883
Concordet JP Lewis KE Moore JW Goodrich LV Johnson RL 1996 Spatial regulation of a zebrafish patched homologue reflects the roles of sonic hedgehog and protein kinase A in neural tube and somite patterning Development 122 2835 2846 8787757
Herzog W Sonntag C Walderich B Odenthal J Maischein HM 2004 Genetic analysis of adenohypophysis formation in zebrafish Mol Endocrinol 18 1185 1195 14752054
Cole LK Ross LS 2001 Apoptosis in the developing zebrafish embryo Dev Biol 240 123 142 11784051
Kimmel CB Miller CT Kruze G Ullmann B BreMiller RA 1998 The shaping of pharyngeal cartilages during early development of the zebrafish Dev Biol 203 245 263 9808777
Geisler R 2002 Mapping and cloning Dahm CN-VaR Zebrafish: A practical approach Oxford (United Kingdom) Oxford University Press 175 212
Shimoda N Knapik EW Ziniti J Sim C Yamada E 1999 Zebrafish genetic map with 2000 microsatellite markers Genomics 58 219 232 10373319
|
16121254
|
PMC1189072
|
CC BY
|
2021-01-05 08:00:57
|
no
|
PLoS Genet. 2005 Aug 19; 1(2):e19
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010019
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1612125510.1371/journal.pgen.001002005-PLGE-RA-0085R3plge-01-02-05Research ArticleDiabetes - Endocrinology - MetabolismMolecular Biology - Structural BiologyNephrologyUrologyGenetics/Gene DiscoveryGenetics/Gene FunctionGenetics/Functional GenomicsGenetics/Disease ModelsMus (Mouse)Homo (Human)MammalsDiabetes Insipidus in Mice with a Mutation in Aquaporin-2 Aquaporin-2 Mutant MiceLloyd David J 1Hall Frank Wesley 2Tarantino Lisa M 1Gekakis Nicholas 1*1 Genomics Institute of the Novartis Research Foundation, La Jolla, California, United States of America
2 Department of Pathology, Scripps Clinic, La Jolla, California, United States of America
Beier David EditorHarvard Medical School, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 19 8 2005 1 2 e2018 4 2005 28 6 2005 Copyright: © 2005 Lloyd et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Congenital nephrogenic diabetes insipidus (NDI) is a disease characterized by failure of the kidney to concentrate urine in response to vasopressin. Human kindreds with nephrogenic diabetes insipidus have been found to harbor mutations in the vasopressin receptor 2 (Avpr2) gene or the vasopressin-sensitive water channel aquaporin-2 (Aqp2) gene. Development of a treatment is rendered difficult due to the lack of a viable animal model. Through forward genetic screening of ethylnitrosourea-mutagenized mice, we report the identification and characterization of a mouse model of NDI, with an F204V mutation in the Aqp2 gene. Unlike previously attempted murine models of NDI, our mice survive to adulthood and more exactly recapitulate the human disorder. Previous in vitro experiments using renal cell lines suggest recessive Aqp2 mutations result in improper trafficking of the mutant water pore. Using these animals, we have directly proven this hypothesis of improper AQP2 translocation as the molecular defect in nephrogenic diabetes insipidus in the intact organism. Additionally, using a renal cell line we show that the mutated protein, AQP2-F204V, is retained in the endoplasmic reticulum and that this abnormal localization can be rescued by wild-type protein. This novel mouse model allows for further mechanistic studies as well as testing of pharmacological and gene therapies for NDI.
Synopsis
Nephrogenic diabetes insipidus (NDI) is a disease marked by excessive urination and thirst. Normally, the hypothalamus senses situations where water is limited and signals to the kidney to increase water reabsorption from urine. The signaling molecule secreted by the hypothalamus is arginine vasopressin (AVP), which binds to a specific protein on the surface of kidney cells, AVP receptor (AVPR2). AVP binding to its receptor on kidney cells begins a series of biochemical events that ultimately results in the insertion of a protein, aquaporin 2 (AQP2), into the outer surface of the kidney cell. As its name suggests, AQP2 facilitates the reuptake of water from the urinary space into the cell, thus concentrating the urine and conserving water. Congenital NDI is caused by mutations in either the water channel, AQP2, or in the receptor, AVPR2. While these mutations have been studied extensively in the lab, work in live animals has been very limited. This report describes the first viable mouse model of NDI. Previous models have been attempted by targeted mutation, i.e., genes known to be involved in the disease have been altered in the mouse, a so-called reverse genetic approach. Reverse genetic approaches have so far failed to produce a viable mouse model of NDI. Here the authors take a forward genetic approach in which genes are mutated at random and animals are screened for disease-like properties. As well as proving hypotheses that come from lab studies, this model opens the door to the testing of gene therapy or other therapies for treatment of NDI.
Citation:Lloyd DJ, Hall FW, Tarantino LM, Gekakis N (2005) Diabetes insipidus in mice with a mutation in aquaporin-2. PLoS Genet 1(2): e20.
==== Body
Introduction
Nephrogenic diabetes insipidus (NDI) is a disease characterized by excessive urination and thirst, despite normal production of the antidiuretic hormone arginine vasopressin (AVP) [1]. The inherited forms are either X-linked as a consequence of mutation of the Avpr2 gene [2], or autosomal due to mutation of the Aqp2 gene [3]. Aquaporin-2 (AQP2) is a pore-forming protein belonging to a family of water channels [4], and it is expressed in collecting-duct principal cells in the kidney [5]. Generally these proteins permit the passage of water through the plasma membrane (PM) of cells, several of which carry out this role specifically in the process of water reabsorption from urine in the kidney. It has been established that aquaporins, although functional as a monomer, tetramerize before their insertion into the plasma membrane [4,6]. Furthermore these proteins can also be differentially targeted to distinct regions of the PM; for example, AQP2 is routed to the apical membrane of cells surrounding the collecting duct, whereas other aquaporins (AQP3 or 4) are inserted into the basolateral face. Unlike all other family members, AQP2 is not constitutively inserted into the plasma membrane. Under basal conditions, the protein resides in subapical intracellular vesicles; however, under conditions requiring water retention AQP2 translocates to the apical membrane, permitting water reabsorption [7,8]. For this process to occur, AVP binds its receptor, AVPR2, on the basolateral face of the collecting duct cells, leading to a rise in intracellular cAMP, ultimately resulting in phosphorylation of AQP2 at serine 256 by cAMP-dependent protein kinase [9] and its redistribution to the plasma membrane.
The importance of AQP2 redistribution has been highlighted by functional characterization of Aqp2 mutations resulting in severe NDI in humans [3,10]. Recessive Aqp2 mutations are generally thought to produce an abnormally localized and, in most instances, misfolded water pore that responds abnormally to an increase in cAMP [6,11]. Furthermore, dominant mutations have been described and found to misroute both the mutant and the wild-type protein to the basolateral membrane [6,12].
Several mouse models of diabetes insipidus have been generated [13–17]. In an attempt to recapitulate human NDI, mice have been generated with mutations in Aqp2 and Avpr2 [15,18]. Yang and colleagues created a mouse with a T126M knock-in mutation in the Aqp2 gene. Unexpectedly, homozygous mutant mice died within 6 d after birth. Interestingly, AVPR2-deficient male pups also die within the first week after birth. Together these models suggest that the mouse may be a highly sensitive organism with regard to water homeostasis, and is unable to survive with polyuria.
In a forward genetic screen, a mouse with an Aqp2 mutation was identified. The purpose of this study was to characterize this murine model of recessive nephrogenic DI. We now report a novel F204V mutation in the Aqp2 gene. This allele of Aqp2 was found to cause the first mouse model of NDI to survive past the first week of life. Molecular analyses concluded that mutant AQP2 adopts a different subcellular localization in renal collecting-duct cells, and was resistant to translocation induced by desmopressin, an agonist of AVP. In vitro studies using the Madin-Darby canine kidney (MDCK) cell line demonstrated an endoplasmic reticulum pattern for the mutant protein, and apparent resistance to translocation. These data conclusively prove that autosomal recessive NDI is a consequence of improper AQP2 routing in the intact mammal.
Results
In a forward genetic screen that used ethylnitrosourea (ENU) to induce mutations in a founder animal whose offspring were then screened for abnormal whole body metabolism [19,20], we found a family of mice that urinated and drank excessively. Serum and urine analysis showed that plasma glucose levels were normal and there was no glucose in the urine (unpublished data). Hence, this was an example of diabetes insipidus.
The disorder in these mice segregated in a monogenic, autosomal recessive manner, making Aqp2 a candidate gene. Sequencing of Aqp2 coding region of affected mice identified a thymine to guanine (T to G) transversion (Figure 1A), which is predicted to lead to a valine for phenylalanine substitution at amino acid 204 of the protein (F204V).
Figure 1 Analysis of Aqp2 Sequence and Phenotype in Mutant Mice
(A) Chromatographic traces of Aqp2 F204V mutation. The box shows the mutated codon, TTC (Phe) to GTC (Val) at position 204.
WT, wild type; Mut, mutant.
(B) Amino acid conservation of mouse AQP2 (residues 194–214). The boxed residue indicates phenylalanine at position 204.
hAQP2, human AQP2; mAQP1, mouse AQP1; mAQP2, mouse AQP2; rAQP2, rat AQP2; xAQP2, Xenopus AQP2.
(C) Urine production (ml) and water consumption (ml) of 58 F2 mice over a 24-h period (both sexes, aged 10–22 wk). Mutant mice (black squares) exhibit overt polyuria and polydipsia compared to littermate wild-type (white triangles) and heterozygous (grey circles) mice.
(D) Urine osmolality and concentrating ability in Aqp2 mutant and their littermates (10–22 wk, both sexes), before (white bars) and after (black bars) dDAVP treatment. Wild type (WT; n = 12); heterozygote (Het; n = 20); mutant (Mut; n = 9). Data represent averages ± standard error of the mean, **p < 0.01; ***p < 0.001.
AQP2 is a six-transmembrane water channel, and F204 lies near the extracellular face of the sixth membrane spanning domain, a region rich in hydrophobic amino acids. This and the other membrane-spanning domains are conserved among vertebrate species. The phenylalanine at position 204 is particularly well conserved (Figure 1B), not only among vertebrate AQP2 proteins, but also among others members of this family.
Aqp2F204V/F204V mice have dramatically increased urine production, in some cases producing an amount of urine in 24 h that exceeds their body weight, compared to their heterozygous or wild-type littermates. Such loss of water would rapidly lead to dehydration were it not compensated by increased water intake. Indeed, mutant mice also dramatically increase their water intake (Figure 1C) compared to their heterozygous or wild-type littermates. This phenotype—increased urinary output and water intake—showed complete concordance with homozygosity of the F204V mutation in the 58 animals tested.
Diabetes insipidus can be defined as an inability to concentrate urine where appropriate. Compared to wild-type or heterozygous littermates, Aqp2F204V/F204V mice produce very dilute urine (Figure 1D). Basal urine concentration in mutant mice is about 161 mOsm, compared to about 1,293 mOsm in wild-type mice (p < 0.001). Normally, urine concentration is under the control of the hypothalamus, which, in response to hypovolemia or hypernatremia [21], secretes AVP. The synthetic AVP analog, 1-deamino-8-D-arginine vasopressin (dDAVP; also called desmopressin), is a potent agonist of AVPR2. When administered to wild-type mice, dDAVP leads to a dramatic increase in urine concentration, from 1,293 to 5,885 mOsm (4.6-fold; Figure 1D). With similar treatment, mutant mice concentrate their urine to a lesser but still significant extent, from 161 to 470 mOsm (2.9-fold), indicating that these animals are not only unable to concentrate their urine properly but are also defective in their response to dDAVP. The smaller response to dDAVP indicates some residual activity of the mutant AQP2 channel, which must be sufficient to allow survival of the individual, in contrast to the T126M knock-in mouse [18].
Multiple heterozygous matings yielded 101 animals, which appeared at a ratio of 26:49:26, near the expected Mendelian wild type, heterozygote, and mutant frequencies, respectively, indicating that there is no reduced viability associated with this mutation. Other than the increased urine production and water intake, there was no overt phenotype in mutant mice, save distended kidneys, which appeared variably in adult animals (Figure 2A). Although not specifically measured, mutant mice seem to have a normal lifespan. The one animal that was followed lived to 18 mo, typical for animals in our colony.
Figure 2 Anatomy and Histology of Mouse Kidneys
(A) Gross anatomy of an affected mouse (8-mo-old male). This shows the enlargement and cystic dilatation of the renal pelvis. There is thinning of the overlying renal parenchyma imparting a translucent appearance to portions of the kidney and collecting system. The bladder is also dilated.
(B) Left kidney from mutant mouse (right) shown in (A) compared to a kidney from an age-sex matched unaffected littermate (left).
(C) Hematoxylin and eosin stained section of ureter from a mutant mouse, showing normal histology despite bloating of the kidney.
(D) Hematoxylin and eosin stained histologic section of a kidney from a 4-wk-old female mutant mouse. The mutant kidney shows marked dilatation of the renal pelvis with blunting of the papilla. There is preservation of the cortex and medulla.
Aqp2F204V/F204V mice suffer from severe hydronephrosis (Figure 2A and 2B), presumably as a consequence of an inability to cope with the extreme polyuria. We found distended kidneys in all Aqp2F204V/F204V mice; however, the degree of inflation was variable in affected mice and worsened with age. Severe hydronephrosis has previously been observed in double Aqp1/Aqp3 knock-out mice [17], and appears at 6 wk. Even at 4 wk, Aqp2F204V/F204V mice had hydronephrosis. Histologic sections from Aqp2F204V/F204V mice demonstrated marked dilatation of the renal pelvis yet normal morphology of the ureter (Figure 2C and 2D). In particular, the muscularis propria was neither hypertrophied nor thinned. There was the normal festooned appearance of the urothelium, and this transitional epithelium was of normal thickness. There was thinning of the kidney as measured from renal capsule to renal pelvis. However, the morphologic features of the glomeruli and proximal/distal tubules were unremarkable (Figure 2D).
As shown previously [18,22], immunoblotting revealed three different forms of AQP2, due to different degrees and forms of glycosylation (Figure 3A). Previous reports have demonstrated that nonglycosylated protein appears as a 29 kDa band, while complex glycosylated protein runs as a smear between 35 and 45 kDa. A short-lived intermediate form of 31 kDa representing core, high-mannose glycosylation of AQP2 is apparent from pulse-chase labeling experiments [22]. Compared to that from the kidneys of wild-type animals, AQP2 from mutant animals was reduced in both the high molecular weight, diffuse form and the lowest molecular weight form, but enriched in the intermediate molecular weight form (Figure 3A). Heterozygous animals showed intermediate amounts of all three forms. The nature of these glycosylated forms was revealed by digestion with endoglycosidase H, which specifically cleaves mannose-rich carbohydrate from the protein backbone. Treatment of endogenous AQP2 from kidneys of wild-type, heterozygous, and mutant animals specifically affected the intermediate molecular weight form (Figure 3B). The presence of some mature glycosylated proteins (35–45 kDa) in Aqp2F204V/F204V mice presumably permits their survival compared to Aqp2T126M/T126M mice, and is consistent with a diminished response to dDAVP.
Figure 3 Immunoblot Analyses of AQP2 from Mouse Kidneys
(A) Western blot analyses of total kidney membranes from littermate mice. An intermediate form of AQP2 at 31 kDa was identified in kidney membranes from a mutant mouse (Mut) and partially in a heterozygous mouse (Het).
(B) Total kidney membranes were subjected to endoglycosidase H treatment (Endo H) prior to Western blotting. High-mannose (h.m.) glycosylated proteins that have not exited the ER are sensitive to endoglycosidase H digestion.
In humans, recessive alleles of Aqp2 are postulated to cause NDI because they do not properly translocate to the apical cell surface in response to AVP. This postulate comes solely from in vitro studies in which mutant Aqp2 cDNAs corresponding to human disease mutations are transfected into kidney cell lines. In general, such recessive alleles, when visualized immunocytochemically, fail to localize to AVP-responsive vesicles. Rather, they get trapped in the endoplasmic reticulum (ER). Our mouse model of NDI affords the first opportunity to test this hypothesis in a mature animal. As shown in Figure 4A (top row of photomicrographs), AQP2 (stained red) normally localized to the subapical region of collecting duct cells in kidneys of wild-type mice. Upon stimulation with dDAVP, AQP2 translocated to or near the cell surface (Figure 4A, second row). In kidneys taken from mutant animals, however, AQP2 was distributed randomly throughout the cell in the basal state (Figure 4A, third row), while AQP3 (green) appropriately localized to the basolateral surface [23]. Furthermore, upon dDAVP stimulation, AQP2-F204V failed to translocate to the cell surface (Figure 4A, bottom row). To confirm these findings, the staining was repeated in kidneys taken from two further mice for each class, wild-type or mutant, with or without dDAVP treatment, with identical results.
Figure 4 AQP2 Subcellular Localization and Translocation in Mouse Kidney Collecting Ducts and MDCK Cell Lines
(A) Immunohistochemistry on collecting ducts in kidney sections from an AQP2-F204V mutant (Mut) mouse and an age-sex matched wild-type (WT) littermate. Mice were injected intraperitoneally with PBS (NT) or dDAVP before sacrificing and fixation of the kidneys. Kidneys sections were immunostained for AQP2 (red) and the basolateral marker AQP3 (green). The images were merged and an area of the cytoplasm was magnified (zoom). Note that mutant AQP2 is not properly localized to the subapical compartment, nor does it respond to dDAVP.
(B) MDCK cell lines, stably transfected with constructs encoding mouse WT or AQP2-F204V, were treated with and without 150 μM forskolin for 90 min, after which cells were fixed, permeabilized, and subjected to immunocytochemistry. AQP2 is shown in green, and the basolateral marker Na+/K+-ATPase is shown in red, alongside the nuclear stain DAPI. The z-profile images were reconstructed from multiple z-sections, along the dotted line. Mutant AQP2 fails to localize to the cell surface upon forskolin stimulation. Rather, the perinuclear staining is consistent with an ER localization of mutant AQP2.
(C) The MDCK cell line expressing AQP2-F204V was grown on fibronectin-coated coverslips until tight junctions formed, at which point the cells were treated with 150 μM forskolin for 90 min. Cells were fixed, permeabilized, and sequentially immunoblotted for AQP2 (green) and calnexin (red), an ER marker. The merged image shows that AQP2-F204V colocalizes with the endoplasmic reticulum marker. Scale bar refers to 10 μm.
To investigate the mechanism of defective translocation of AQP2-F204V, we turned to transfection of MDCK cells. Stable cell lines expressing mouse wild-type AQP2 and AQP2-F204V were established. Immunoblots of protein extracts from stable cell lines showed that MDCK cells recapitulate the glycosylation defect seen in mutant mice (unpublished data). The wild-type protein was again present in three different forms. Cells expressing AQP2-F204V lacked the 35–45 kDa form and were enriched in the core-glycosylated 31 kDa form.
In transfected, unstimulated MDCK cells, wild-type AQP2 (stained green) appeared in a punctate pattern distributed throughout the subapical region (Figure 4B, left column photomicrographs), consistent with vesicular compartmentalization. AQP2-F204V, on the other hand, appeared in a punctate but perinuclear pattern (Figure 4B, third column). Upon stimulation with forskolin, a cAMP-dependent protein kinase activator, wild-type AQP2 translocated to the apical surface of polarized MDCK cells (Figure 4B, second column). Along the z-axis, the perinuclear distribution of AQP2-F204V was clearly seen, and this distribution is not altered by forskolin (Figure 4B, bottom row, two right columns). The perinuclear distribution of AQP2-F204V is consistent with an ER compartmentalization. To test the idea that AQP2-F204V localizes to the ER, we co-stained cells transfected with Aqp2F204V (cDNA) for AQP2 and an ER marker, calnexin (Figure 4C). Colocalization of calnexin with AQP2 was investigated directly, and it was found that 80% of all AQP2-F204V protein colocalized with calnexin. The remaining 20% appeared at the periphery of the ER, representing AQP2-F204V that had potentially progressed beyond the ER. This “ER escape” was consistent with the small proportion of mature, complex glycosylated, AQP2-F204V in mutant kidneys (see Figure 3A).
Animals heterozygous for the Aqp2F204V mutation were not affected in their urine production or urine osmolality (see Figure 1C and 1D). It has also been shown that a recessive NDI allele, AQP2-R187C, does not interact with wild-type protein in oocytes [24], nor does it homo-oligomerize in MDCK cells [22]. Therefore, kidneys from heterozygous animals were examined for evidence of two populations of AQP2 protein. Surprisingly, immunohistochemical staining of kidney collecting ducts from Aqp2F204V/+ mice revealed a pattern remarkably similar to wild type (Figure 5A). AQP2 translocated completely to the apical cell surface upon dDAVP stimulation. This wild-type staining pattern may simply reflect the fact that decreasing the amount of mutant protein by half makes it undetectable by immunocytochemistry. Alternatively, the presence of wild-type protein may alter the localization of the mutant protein. Indeed, Hendriks et al. proposed a “piggy-back” mechanism to explain the transport of nonglycosylated subunits of AQP2 to the cell surface by glycosylated subunits [22]. It has also been shown that wild-type AQP2 protein can rescue a translocation-defective mutant protein, AQP2-P262L, when the two are coexpressed in MDCK cells [25].
Figure 5 AQP2-F204V Rescue in Heterozygous Mouse Collecting Ducts and in Cotransfected MDCK Cells
(A) In heterozygous animals, AQP2 localizes and responds to dDAVP normally. Immunohistochemistry was carried out on kidney sections from an Aqp2F204V/+ mouse, after injection with dDAVP. Kidney sections were sequentially immunostained for AQP2 (red) and the basolateral marker AQP3 (green).
(B) Mutant and wild-type AQP2 physically interact. MDCK cells stably expressing wild-type AQP2 were transiently transfected with GFP tagged wild-type AQP2, AQP2-F204V, or GFP alone. Solubilized membranes were immunoprecipitated with a GFP antibody. Total membranes and immunoprecipitates (GFP-IP) were Western blotted using an antibody against AQP2 (arrow) or AQP2-GFP fusions (arrowhead).
(C) Wild-type AQP2 rescues the localization defect of mutant AQP2. GFP fusions of either wild-type AQP2 (WT-GFP, top photomicrographs) or F204V AQP2 (F204V-GFP, bottom photomicrographs) were expressed in polarized MDCK stable cell lines expressing vector alone (vector, left photomicrographs) or AQP2-WT (right photomicrographs). Cells were stimulated with forskolin, processed for immunocytochemistry, and used to generate z-sectional images.
(D) Mutant AQP2 is present at the cell surface in cells coexpressing wild-type AQP2 (AQP2-WT). GFP fused to AQP2-F204V was expressed in MDCK cells expressing wild-type AQP2 or vector alone. Cells were stimulated with forskolin, and cell surface biotinylated proteins were precipitated then analyzed for the presence of wild-type AQP2 (arrow) and AQP2-F204V (arrowhead) by Western blot.
In the collecting ducts from Aqp2F204V/+ mice, the wild type may rescue the mutant protein as suggested by the subcellular distribution of AQP2 protein. To test this idea, we first looked for an interaction between mutant and wild-type proteins in transfected MDCK cells (Figure 5B). MDCK cells stably expressing wild-type AQP2 were transiently transfected with GFP expression constructs encoding GFP-tagged wild-type AQP2, AQP2-F204V, or GFP alone. Antibodies against GFP coimmunoprecipitated wild-type AQP2 when AQP2-GFP or AQP2-F204V-GFP was transiently transfected, but not when GFP by itself was transiently transfected into MDCK cells stably expressing wild-type AQP2 (Figure 5B, upper blots). Western blot of total membranes showed that wild-type AQP2 is equivalently expressed in all three cases (Figure 5B, lower blots).
If wild-type and mutant proteins are indeed interacting in the cell, is this interaction sufficient to rescue the localization of mutant protein? To answer this question, we used MDCK cells stably transfected with wild-type AQP2 expressing vector or with empty vector. On top of these, we transiently transfected AQP2-GFP or AQP2-F204V-GFP expression constructs. AQP2-GFP localized to the apical surface upon forskolin stimulation whether it was transiently transfected into vector-only cells (Figure 5C, upper left images) or into wild-type AQP2 cells (Figure 5C, upper right). AQP2-F204V-GFP, when expressed by transient transfection into vector only cells, showed a diffuse cytoplasmic distribution pattern (Figure 5C, lower left). When expressed in wild-type AQP2 cells, however, AQP2-F204V-GFP localized to the apical surface to varying degrees (Figure 5C, lower right images [i–iii]). The lower right images of Figure 5C shows three cells from a single transfection. The first is a nontransfected cell that shows the localization of the stably expressing wild-type AQP2, which is apical upon forskolin stimulation. The next two show expression of both the stable wild-type AQP2 and the transient AQP2-F204V-GFP. In cell (ii), localization of wild-type AQP2 was indistinguishable from AQP2-F204V-GFP; both were apical upon forskolin stimulation. Although the effect was subtle in cell (iii), AQP2-F204V-GFP was partly localized to the apical surface. Generally, the localization of AQP2-F204V-GFP was clearly more apical when wild-type AQP2 was also expressed.
To confirm these results biochemically, we transfected the same cell lines (wild-type AQP2 or vector) with F204V-GFP, biotinylated surface proteins after forskolin stimulation, and precipitated the biotinylated proteins (Figure 5D). AQP2-F204V-GFP is expressed approximately equally in both cell lines (Figure 5D, total cells), but is biotinylated only when wild-type AQP2 is also expressed (Figure 5D, surface biotinylated). Since only cell surface proteins are accessible to biotin, these results indicate that AQP2-F204V is transported to the cell surface when wild-type AQP2 is present, but not on its own.
Discussion
Aqp2F204/F204V mice are viable and grow and reproduce normally. They are, however, severely defective in their ability to concentrate urine, leading to increased urine output and water intake, thus making them the first mouse model of NDI to survive to maturity. In humans, NDI is caused by mutations in Avpr2 or Aqp2. Knockout of the X-linked Avpr2 gene in mice [15] gave an NDI-like phenotype in male, hemizygous neonates, but the phenotype could not be assessed in adults as the mice died within 1 wk of birth. The adult heterozygous females showed a mild tendency toward increased urinary output and water intake and decreased urine osmolality. Knockout of the mouse Aqp2 gene has not been reported. A knock-in of a human disease-causing mutation (T126M), however, has been made [18]. These mice have a severe urine-concentrating defect resulting in dehydration and death within 1 wk of birth. Curiously, AQP2-T126M does localize properly in at least a subset of cells. The grossly abnormal collecting duct morphology makes it impossible to pinpoint the molecular defect in these knock-in mice.
The T126M knock-in clearly shows that Aqp2 is an essential gene [18]. The fact that our mice survive shows either that AQP2-F204V possesses some residual water transporting ability or that there are AVP-independent pathways for water reabsorption. Residual activity of AQP2-F204V is likely, as mutant animals show some small response to dDAVP, although dDAVP-stimulated urine osmolality remains quite low. Immunostaining of kidney shows that AQP2-F204V does not efficiently transport water, because it fails to localize to the apical cell surface after dDAVP treatment. Some residual activity of AQP2 would imply that some small, undetectable portion of the mutant protein is getting to the cell surface. The surface biotinylation experiment (Figure 5D) suggests that no mutant protein gets to the surface, but this does not necessarily reflect the situation in vivo. While this small fraction of protein may not be detectable by immunofluorescence, Western blotting shows that some mutant protein does progress beyond the ER (35–45 kDa species in Figure 3A). Compared to wild-type, mutant protein is enriched in the high-mannose, core-glycosylated form (31 kDa) and deficient in nonglycosylated (29 kDa) and complex glycosylated (35–45 kDa) forms. The presence of a reduced but detectable amount of protein in the 35–45 kDa range indicates that mutant protein is transported out of the ER, but with greatly reduced efficiency. Colocalization of AQP2-F204V with the ER protein calnexin in transfected MDCK cells shows that, while most of the mutant protein is trapped in the ER, some does progress beyond the ER. Diminished response to dDAVP, diminished abundance of mature glycosylated protein in mutant animals, and the transport of a fraction of mutant protein beyond the ER in MDCK cells are all consistent with the notion that AQP2-F204V misfolding is limited and that it may retain some residual water transporting activity. Evidently this residual activity is sufficient for the viability and growth of mutant animals.
Reduced efficiency in exiting the ER may explain why AQP2-F204V is enriched in the 31 kDa high-mannose glycosylated form. The high-mannose core oligosaccharide is added in the ER and is later modified and elaborated in the Golgi apparatus [26]. The increase in the high-mannose glycosylated form of AQP2-F204V may simply reflect its prolonged presence in the ER and exposure to oligosaccharyl transferase.
While improper localization of AQP2 explains the phenotype of homozygous mutant mice, the complete lack of a phenotype in heterozygous mice is more difficult to explain. Physiologically, heterozygous mice have no symptoms (see Figure 1C), and they are indistinguishable from wild type on immunostaining of kidneys (Figure 5A). The presence of 50% of the normal amount of wild-type protein may explain the lack of symptoms, but it cannot explain the lack of any ER-retained mutant protein. Rather, the phenotype of the Aqp2F204V/+ animals suggests that the mutant protein is being rescued by the wild-type protein. Indeed, de Mattia et al. (26) have demonstrated that one recessive allele of Aqp2, P262L, does not properly translocate when expressed by itself in MDCK cells, but that in the presence of wild-type protein, it localizes normally. The same mechanism seems to apply in vivo with Aqp2F204V/+ mice. In support of this, AQP2-F204V can interact with wild-type AQP2 (Figure 5B), and when coexpressed with wild-type protein, AQP2-F204V can reach the cell surface (Figure 5C bottom right panel and 5D). Although it has been demonstrated that a recessive allele (encoding AQP2-R187C) of NDI fails to interact with wild-type AQP2 [6], here we show that AQP2-F204V does interact with the wild-type protein, presumably as part of heterotetramers, and represents a rescuable allele, both in vitro and in vivo.
Immunostaining the kidneys of homozygous Aqp2F204V/F204V mice shows that the mutant-expressing collecting duct cells can not mediate water reabsorption, because it fails to insert into the apical plasma membrane in response to dDAVP. This is this first in vivo proof of a long-standing hypothesis that comes from in vitro studies with recessive Aqp2 mutations. Transfection into MDCK cells of any of several Aqp2 mutations corresponding to recessive human alleles shows abnormal subcellular localization [25], [27] and failure to appropriately translocate to the plasma membrane. Thus, misfolding, retention in the ER, and failure to translocate in response to dDAVP were proposed as the mechanism for autosomal recessive NDI. Here we not only prove this hypothesis but also establish a useful model for human NDI. This mouse model of NDI based on an Aqp2 allele that can be rescued provides the opportunity to test therapies, including gene therapy, that may promote proper subcellular localization.
Materials and Methods
Generation of ENU mice and housing.
ENU mutagenized C57BL/6 mice were generated as described [19]. Mice were maintained by backcrossing affected animals to C57BL/6 and housed in the Genomics Institute of the Novartis Research Foundation Specific Pathogen Free animal facility (La Jolla, California, United States). All procedures were approved by the Genomics Institute of the Novartis Research Foundation Institutional Animal Care and Use Committee.
Constructs.
The complete coding sequence of mouse AQP2 from an IMAGE clone was digested from the pCMV⋅SPORT6 plasmid with EcoRI and NotI and ligated into pcDNA3.1 (Invitrogen, Carlsbad, California, United States). The F204V mutation was introduced by site-directed mutagenesis (Stratagene, La Jolla, California, United States), using the sense oligonucleotide 5′-GATGATCACTGGGTCGTCTGGATCGGACCCC-3′, and antisense oligonucleotide 5′-GGGGTCCGATCCAGACGACCCAGTGATCATC-3′. To generate GFP fusions of AQP2, the pCMV⋅SPORT6 AQP2 construct was used in a PCR reaction with the primers Sp6 and 5′-GACTGGATCCCGGCCTTGCTGCCGCGCGGCAG-3′ to remove the stop codon of AQP2. The product was digested with KpnI and BamHI and ligated into pEGFP-N2 (BD Biosciences, San Diego, California, United States). The F204V mutation was introduced using the same mutagenic oligonucleotides.
Cell culture and generation of stable cell lines.
MDCK cells (CCL-34; ATCC, Manassas, Virginia, United States) were cultured in DMEM (Sigma-Aldrich, St. Louis, Missouri, United States) supplemented with 10% FBS (Sigma-Aldrich), 100 U/ml of penicillin, and 100 μg/ml of streptomycin at 37 °C in 5% CO2. To generate stable MDCK cell lines, cells were transfected using Lipofectamine 2000 (Invitrogen) and the pcDNA3.1 expression constructs (containing wild-type AQP2, AQP2-F204V, or no insert) and selected with 900 μg/ml G418 (Sigma-Aldrich). Individual colonies were expanded 14 d later. For the duration of these experiments, the antibiotic was continually added to the media. Transient GFP transfections were carried out in subconfluent stable cells lines also using Lipofectamine 2000.
Sequencing of Aqp2 and genotyping of mice.
All exons of Aqp2 were amplified from mouse genomic DNA and sequenced. For genotyping, exon 4 was amplified using the primers 5′-TCAGAACTTGCCCACTAGCC-3′ and 5′-TGTAGAGGAGGGAACCGATG-3′.
Urine measurements.
Total urine output was measured by separately housing adult mice in Nalgene Metabolic Cages (Minimitter, Bend, Oregon, United States) for 2–3 d and collecting urine every 24 h period. Urine osmolalities were determined using an Osmometer (Osmette 5004; Precision Systems, Natick, Massachusetts, United States). Urine concentrating experiments were carried out by intraperitoneal injection of dDAVP (0.4 μg/kg). Mice were injected twice with dDAVP, once at time 0 and again at 30 min. Urine was collected at the start of the experiment and 30 min after the second injection.
Kidney membrane preparation.
Whole mouse kidneys were homogenized in 10 mM Tris (pH 7.4), 350 mM sucrose, and 5 mM EDTA containing protease inhibitors (Sigma-Aldrich, #P-8340) in a Potter-Elvehjem homogenizer. The homogenate was centrifuged at 2,000 g for 10 min and the supernatant was subjected to ultracentrifugation at 100,000 g for 1 h at 4 °C. Pelleted membranes were resuspended in the same buffer, and protein concentration was determined by Bradford assay.
Immunoblotting.
Kidney membrane fractions (60 μg) were resolved on a 12% SDS-polyacrylamide gel and transferred to a nitrocellulose membrane. Membranes were blocked in 5% nonfat milk in Tris-buffered saline with 0.05% Tween 20 (TBST), followed by an overnight incubation (at 4 °C) with AQP2 polyclonal antibody (Santa Cruz Biotechnology, Santa Cruz, California, United States; #sc-9882). Membranes were washed in TBST then incubated with HRP-conjugated donkey anti-goat antibody. Membranes were washed further in TBST and bands were visualized using ECL reagent (Amersham Biosciences, Little Chalfont, United Kingdom).
Endoglycosidase digestion.
Kidney membranes (60 μg) were incubated in 50 mM sodium phosphate (pH 5.5), 0.1% SDS, and 50 mM β-mercaptoethanol, heated to 100 °C for 5 min, then cooled. Endoglycosidase H (0.01 units; Sigma-Aldrich) was added and incubated at 37 °C for 2 h. The reaction was stopped by boiling the samples in Laemmli buffer. Total reactants were immunoblotted as described above.
Coimmunoprecipitation and biotinylation in MDCK cells.
MDCK cells stably expressing wild-type AQP2 (grown on 10-cm plates) were transfected with pEGFP-wild-type AQP2, pEGFP-AQP2-F204V, or vector alone. The cells were homogenized in 10 mM Tris (pH 7.4), 1 mM EDTA, and 250 mM sucrose 40 h later. The clarified supernatant was centrifuged at 200,000 g for 30 min. Pelleted membranes were resuspended in the same buffer but containing 4% sodium deoxycholate and incubated at 37 °C for 1 h. From the dissolved membranes, a 30 μl sample was removed and used as the total membrane fraction. The remaining membranes were diluted with 600 μl of the homogenization buffer, and incubated with 1 μl of GFP antisera (Invitrogen, #46–0092) and protein A/G sepharose. Following overnight incubation, the precipitated proteins were washed in RIPA buffer and finally boiled in 50 μl of Laemmli buffer. Half of the total membrane and the IP fractions were processed for immunoblotting.
Cell surface biotinylation was performed in a similar manner. However, pEGFP-AQP2-F204V, was transfected into MDCK cells stably expressing wild-type AQP2 and cells made stable with vector alone. Twenty-four hours post-transfection, cells were stimulated with forskolin, trypsinized, resuspended in 1 ml of PBS (2.5 × 106 cells/ml), and incubated with 0.5 mg of NHS-PEO4-biotin (Pierce Biotechnology) for 30 min at room temperature. Cells were washed once in 10 mM Tris (pH 8) and three times in PBS, after which membranes were purified and solubilized as described above. Solubilized membranes were incubated with 20 μl of immobilized streptavidin (Pierce Biotechnology) for 2 h at 4 °C. Finally the precipitated proteins were washed in RIPA buffer and boiled in 50 μl of Laemmli buffer. Total cells and the biotinylated precipitates were immunoblotting using an antibody to AQP2.
Kidney immunohistochemistry.
Whole mouse kidneys were fixed in 10% phosphate-buffered formalin for 24 h. Kidneys were embedded in paraffin, and 5-μm sections were prepared. Following antigen retrieval using 10 mM sodium citrate (pH 8) for 10 min at 98 °C, sections were sequentially probed, first for AQP3 and then for AQP2. Sections were incubated in 5% donkey serum and then in goat anti-AQP3 antibody (1:100; Santa Cruz Biotechnology; #sc-9885). Slides were washed with PBS and incubated with AlexaFluor 488-conjugated donkey anti-goat antibody (Molecular Probes, Eugene, Oregon, United States). The slides were subsequently blocked in 5% chicken serum, incubated with a rabbit anti-AQP2 antibody (1:250; USB, Cleveland, Ohio, United States; #A3000–06), which was detected with a AlexaFluor 594-conjugated chicken anti-rabbit antibody (1:500; Molecular Probes). The sections were stained with DAPI and mounted in Vectashield (Vector Labs, Burlingame, California, United States).
Immunocytochemistry on MDCK cells.
MDCK stable cell lines expressing vector alone, wild-type AQP2, or AQP2-F204V (and in some cases transiently expressing a GFP construct) were grown on fibronectin-coated coverslips until tight junctions formed. Cells were treated with or without 150 μM forskolin for 90 min, and fixed in methanol at −20 °C. Subsequently, cells were washed and permeabilized in 0.2% Triton X-100 for 5 min, and sequentially probed for AQP2 and organelle markers for either the PM or the ER. AQP2 was detected using goat anti-AQP2 (1:100; Santa Cruz Biotechnology; #sc-9882) and a 1:200 dilution of AlexaFluor 488-conjugated donkey anti-goat secondary antibody. The PM and ER were probed using mouse anti-Na+/K+-ATPase (Upstate, Waltham, Massachusetts, United States) or rabbit anti-calnexin (Stressgen Biotechnology, Victoria, British Columbia, Canada) antibodies and the secondary antibodies, Cy3-conjugated goat anti-mouse (1:200; Jackson ImmunoResearch, West Grove, Pennsylvania, United States) or AlexaFluor 594-conjugated chicken anti-rabbit (1:200) respectively. Cells were washed in PBS, counterstained with DAPI, and mounted in Vectashield. In experiments in which GFP fusions were used, AQP2 was probed using the antibody combination used for kidney immunohistochemistry in order to detect the AQP2 at 594 nm, to distinguish between the GFP fusion proteins.
Confocal microscopy.
Optical z-section images were collected on a BioRad (Hercules, California, United States) Rainbow Radiance 2100 Laser Scanning Confocal Microscope. Image stacks were flattened, or sectioned along the z-axis, then further processed using BioRad Laser Sharp 2000 software and Image J software (v. 1.32; National Institutes of Health). Colocalization was performed using the overlay coefficient of Image J software.
Supporting Information
Accession Numbers
The GenBank (http://www.ncbi.nlm.nih.gov/) accession number of Aqp2 is NM_009699. The IMAGE (http://image.llnl.gov) accession number of AQP2 is 4222942.
We thank Debby Stradley for all genotyping, Lacey Kischassey for breeding and care of mice, Karina Ayala and Sandy Bohan for phenotyping the study mice, Miah Gilmore for performing endoglycosidase H experiments, James Watson for sectioning tissue, and Dr. William Kiossis for collecting confocal images.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. DJL and NG conceived and designed the experiments. DJL performed the experiments. DJL, FWH, and NG analyzed the data. DJL and LMT contributed reagents/materials/analysis tools. DJL and NG wrote the paper.
Abbreviations
AQP[number]aquaporin-[number]
AVPR2AVP type 2 receptor
AVParginine vasopressin
dDAVP1-deamino-8-D-arginine vasopressin
ERendoplasmic reticulum
MDCKMadin-Darby canine kidney
NDInephrogenic diabetes insipidus
PMplasma membrane
==== Refs
References
Maghnie M 2003 Diabetes insipidus Horm Res 59 Suppl 1 42 54
van den Ouweland AM Dreesen JC Verdijk M Knoers NV Monnens LA 1992 Mutations in the vasopressin type 2 receptor gene (AVPR2) associated with nephrogenic diabetes insipidus Nat Genet 2 99 102 1303271
Deen PM Verdijk MA Knoers NV Wieringa B Monnens LA 1994 Requirement of human renal water channel aquaporin-2 for vasopressin-dependent concentration of urine Science 264 92 95 8140421
King LS Kozono D Agre P 2004 From structure to disease: The evolving tale of aquaporin biology Nat Rev Mol Cell Biol 5 687 698 15340377
Fushimi K Uchida S Hara Y Hirata Y Marumo F 1993 Cloning and expression of apical membrane water channel of rat kidney collecting tubule Nature 361 549 552 8429910
Kamsteeg EJ Bichet DG Konings IB Nivet H Lonergan M 2003 Reversed polarized delivery of an aquaporin-2 mutant causes dominant nephrogenic diabetes insipidus J Cell Biol 163 1099 1109 14662748
Marples D Knepper MA Christensen EI Nielsen S 1995 Redistribution of aquaporin-2 water channels induced by vasopressin in rat kidney inner medullary collecting duct Am J Physiol 269 C655 664 7573395
Nielsen S Chou CL Marples D Christensen EI Kishore BK 1995 Vasopressin increases water permeability of kidney collecting duct by inducing translocation of aquaporin-CD water channels to plasma membrane Proc Natl Acad Sci U S A 92 1013 1017 7532304
Fushimi K Sasaki S Marumo F 1997 Phosphorylation of serine 256 is required for cAMP-dependent regulatory exocytosis of the aquaporin-2 water channel J Biol Chem 272 14800 14804 9169447
van Lieburg AF Verdijk MA Knoers VV van Essen AJ Proesmans W 1994 Patients with autosomal nephrogenic diabetes insipidus homozygous for mutations in the aquaporin 2 water-channel gene Am J Hum Genet 55 648 652 7524315
Deen PM Croes H van Aubel RA Ginsel LA van Os CH 1995 Water channels encoded by mutant aquaporin-2 genes in nephrogenic diabetes insipidus are impaired in their cellular routing J Clin Invest 95 2291 2296 7537761
Asai T Kuwahara M Kurihara H Sakai T Terada Y 2003 Pathogenesis of nephrogenic diabetes insipidus by aquaporin-2 C-terminus mutations Kidney Int 64 2 10 12787389
Ma T Song Y Yang B Gillespie A Carlson EJ 2000 Nephrogenic diabetes insipidus in mice lacking aquaporin-3 water channels Proc Natl Acad Sci U S A 97 4386 4391 10737773
Ma T Yang B Gillespie A Carlson EJ Epstein CJ 1998 Severely impaired urinary concentrating ability in transgenic mice lacking aquaporin-1 water channels J Biol Chem 273 4296 4299 9468475
Yun J Schoneberg T Liu J Schulz A Ecelbarger CA 2000 Generation and phenotype of mice harboring a nonsense mutation in the V2 vasopressin receptor gene J Clin Invest 106 1361 1371 11104789
Yang B Ma T Verkman AS 2001 Erythrocyte water permeability and renal function in double knockout mice lacking aquaporin-1 and aquaporin-3 J Biol Chem 276 624 628 11035042
Matsumura Y Uchida S Kondo Y Miyazaki H Ko SB 1999 Overt nephrogenic diabetes insipidus in mice lacking the CLC-K1 chloride channel Nat Genet 21 95 98 9916798
Yang B Gillespie A Carlson EJ Epstein CJ Verkman AS 2001 Neonatal mortality in an aquaporin-2 knock-in mouse model of recessive nephrogenic diabetes insipidus J Biol Chem 276 2775 2779 11035038
Wen BG Pletcher MT Warashina M Choe SH Ziaee N 2004 Inositol (1,4,5) trisphosphate 3 kinase B controls positive selection of T cells and modulates Erk activity Proc Natl Acad Sci U S A 101 5604 5609 15064401
Nelms KA Goodnow CC 2001 Genome-wide ENU mutagenesis to reveal immune regulators Immunity 15 409 418 11567631
Knoers NV Deen PM 2001 Molecular and cellular defects in nephrogenic diabetes insipidus Pediatr Nephrol 16 1146 1152 11793119
Hendriks G Koudijs M van Balkom BW Oorschot V Klumperman J 2004 Glycosylation is important for cell surface expression of the water channel aquaporin-2 but is not essential for tetramerization in the endoplasmic reticulum J Biol Chem 279 2975 2983 14593099
Ishibashi K Sasaki S Fushimi K Uchida S Kuwahara M 1994 Molecular cloning and expression of a member of the aquaporin family with permeability to glycerol and urea in addition to water expressed at the basolateral membrane of kidney collecting duct cells Proc Natl Acad Sci U S A 91 6269 6273 7517548
Marr N Bichet DG Lonergan M Arthus MF Jeck N 2002 Heteroligomerization of an aquaporin-2 mutant with wild-type aquaporin-2 and their misrouting to late endosomes/lysosomes explains dominant nephrogenic diabetes insipidus Hum Mol Genet 11 779 789 11929850
de Mattia F Savelkoul PJ Bichet DG Kamsteeg EJ Konings IB 2004 A novel mechanism in recessive nephrogenic diabetes insipidus: Wild-type aquaporin-2 rescues the apical membrane expression of intracellularly retained AQP2-P262L Hum Mol Genet 13 3045 3056 15509592
Dempski RE Jr. Imperiali B 2002 Oligosaccharyl transferase: Gatekeeper to the secretory pathway Curr Opin Chem Biol 6 844 850 12470740
Tamarappoo BK Verkman AS 1998 Defective aquaporin-2 trafficking in nephrogenic diabetes insipidus and correction by chemical chaperones J Clin Invest 101 2257 2267 9593782
|
16121255
|
PMC1189073
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Aug 19; 1(2):e20
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010020
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1612125610.1371/journal.pgen.001002305-PLGE-RA-0058R2plge-01-02-07Research ArticleGenetics/Disease ModelsMus (Mouse)Medium-Chain Acyl-CoA Dehydrogenase Deficiency in Gene-Targeted Mice MCAD Gene Disruption in the MouseTolwani Ravi J 12Hamm Doug A 1Tian Liqun 1Sharer J. Daniel 1Vockley Jerry 34Rinaldo Piero 5Matern Dietrich 5Schoeb Trenton R 1Wood Philip A 1*1 Department of Genetics, University of Alabama, Birmingham, Alabama, United States of America
2 Department of Comparative Medicine, Stanford University, Stanford, California, United States of America
3 Department of Medical Genetics, Mayo Clinic College of Medicine, Rochester, Minnesota, United States of America
4 Division of Medical Genetics, Children's Hospital, University of Pittsburgh, Pittsburgh, Pennsylvania, United States of America
5 Laboratory Medicine and Pathology, Mayo Clinic College of Medicine, Rochester, Minnesota, United States of America
Valle David EditorJohns Hopkins Institute, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 19 8 2005 1 2 e2329 3 2005 1 7 2005 Copyright: © 2005 Tolwani et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Medium-chain acyl-CoA dehydrogenase (MCAD) deficiency is the most common inherited disorder of mitochondrial fatty acid β-oxidation in humans. To better understand the pathogenesis of this disease, we developed a mouse model for MCAD deficiency (MCAD−/−) by gene targeting in embryonic stem (ES) cells. The MCAD−/− mice developed an organic aciduria and fatty liver, and showed profound cold intolerance at 4 °C with prior fasting. The sporadic cardiac lesions seen in MCAD−/− mice have not been reported in human MCAD patients. There was significant neonatal mortality of MCAD−/− pups demonstrating similarities to patterns of clinical episodes and mortality in MCAD-deficient patients. The MCAD-deficient mouse reproduced important aspects of human MCAD deficiency and is a valuable model for further analysis of the roles of fatty acid oxidation and pathogenesis of human diseases involving fatty acid oxidation.
Synopsis
Medium-chain acyl-CoA dehydrogenase (MCAD) deficiency is one of the most common inherited disorders of metabolism. This defect in fatty acid oxidation can lead to severe and sometimes fatal disease, especially in young children because they are unable to tolerate a fasting episode. Metabolic complications include very low blood glucose concentrations and generation of toxic by-products. This disorder can result in sudden infant death. Using a process known as gene targeting in mouse embryonic stem cells, the authors have developed a mouse model with the same enzyme deficiency. This mouse model of MCAD deficiency develops many of the same disease characteristics found in affected children. The MCAD-deficient mouse model shows a high rate of newborn loss, intolerance to cold, and the characteristic biochemical changes in the blood, tissues, and urine that are very similar to those found in the human disease counterpart. The MCAD-deficient mouse model will allow researchers to better understand disease mechanisms so that new preventive measures or therapies can be developed.
Citation:Tolwani RJ, Hamm DA, Tian L, Sharer JD, Vockley J, et al. (2005) Medium-chain acyl-CoA dehydrogenase deficiency in gene-targeted mice. PLoS Genet 1(2): e23.
==== Body
Introduction
Mitochondrial β-oxidation of fatty acids provides energy, especially during fasting conditions. Fatty acid oxidation occurs in mitochondria and consists of a repeating circuit of four sequential steps. There are four straight-chain acyl-CoA dehydrogenases involved in the initial step. Medium-chain acyl-CoA dehydrogenase (MCAD) (the mouse gene is Acadm, whereas the protein is MCAD), specifically, is responsible for catalyzing the dehydrogenation of medium-chain length (C6–C12) fatty acid thioesters [1]. Acadm is transcribed in the nucleus, translated in the cytosol, and translocated into the mitochondrial matrix [2–4]. Once inside the mitochondrial matrix, the MCAD monomers are assembled into homotetramers to gain enzymatic activity [4].
MCAD activity is essential for complete fatty acid oxidation. Inherited MCAD deficiency exists in humans as an autosomal recessive disorder. MCAD deficiency was first described in 1982—1983 [5–7] and has been described in numerous patients [1,8–11]. The carrier frequency in the Caucasian population has been estimated to be between 1 in 50 to 80 with an incidence of clinical disease expected at around 1 in 15,000 [1,9,12].
MCAD-deficient patients exhibit clinical episodes often associated with fasting. Patients manifest disease usually during the first two years of life. Symptoms include hypoketotic hypoglycemia and Reye-like episodes [1]. It is estimated that approximately 59% of patients presenting clinically between 15 to 26 mo of age die during their first clinical episode [1].
The pathogenesis of the wide range of metabolic disturbances in MCAD deficiency is poorly understood and certain aspects of patient management are controversial. An animal model for MCAD deficiency is essential to better understand the pathogenesis of MCAD deficiency and to develop better management regimens for human patients. To gain further insight into the mechanisms of this disease, we developed a mouse model of MCAD deficiency by gene targeting in embryonic stem (ES) cells (for reviews [13,14]). The mutant mice had many relevant features characteristic of the disease found in human MCAD-deficient patients, along with some unexpected findings.
Results
Gene Targeting and Generation of MCAD-Deficient Mice
MCAD insertion vector (MCAD IV2) was designed to undergo gap repair of the 1.3-kb deleted region upon homologous recombination in 129P2 (129P2/OlaHsd) ES cells E14–1. Correct targeting of the MCAD locus resulted in a duplication of exons 8, 9, and 10 and integration of flanking plasmid and Neo sequences (Figure 1A). The insertion vector was designed to duplicate exon 8, 9, and 10 at the MCAD locus. Translation of the duplicated exon 8 region results in the formation of premature stop codons resulting in truncation of the MCAD monomer. Specifically, the first premature stop codon arises after translation of only seven amino acids from the duplicated exon 8. The resulting MCAD monomer is missing the C-terminal domain α-helixes that are responsible for making intersubunit contacts to generate the functional MCAD homotetramer.
Figure 1 Strategy for Disruption of the Mouse Acadm Gene
(A) The MCAD IV2 insertion targeting vector with a deleted 1.3-kb region encompassing exon 10 and flanking sequences. MCAD IV2 undergoes gap repair upon homologous recombination at the endogenous Acadm locus resulting in a duplication of exons 8, 9, and 10 at the disrupted allele.
(B) Southern blot analysis of EcoRI-digested genomic DNA from ES cells screened by PCR. Probe A, a DNA fragment consisting of a portion of exon 10 that is not present in the targeting vector, hybridizes to an endogenous 3.1-kb fragment and, upon homologous recombination, to a 13.2-kb fragment. Lane 1 represents a wild-type ES cell line, and Lane 2 and 3 represent targeted ES cell lines.
ES cell clones were screened by PCR (data not shown) and confirmed by Southern blot analysis. Southern blot analysis used an exon 10 probe (probe A), not present in the targeting vector, hybridized to a 13.2-kb band in addition to the 3.1-kb endogenous band indicating targeted insertion of the vector at the Acadm locus (Figure 1B). Correctly targeted ES cell clones were microinjected into B6 (C57BL/6NTac) blastocysts to generate chimeric mice. Chimeric mice were backcrossed to both 129P2 and B6 inbred mice to produce MCAD+/− and eventually MCAD−/− mice on a B6/129 mixed background. The studies described here were conducted exclusively on the B6/129 mixed background compared with littermate controls or B6/129 control groups maintained by intercrosses as were the mutants. Perpetuating this mutation as a congenic mutant line on the 129P2 background proved impractical. The 129P2 mice were poor breeders as wild-types, and when introduced, the Acadm mutation was nearly lost on this background because of the high rate of neonatal death. Because of the molecular structure of the targeted allele, it proved virtually impossible to distinguish all three potential genotypes. We could clearly detect the presence or absence of the targeted allele, however, whether a particular mouse was MCAD−/− or MCAD+/− could not be determined by Southern blot or PCR of genomic DNA. Ultimately MCAD−/− mice were ascertained by immunoblot analysis of offspring with subsequent perpetuation of MCAD−/− and MCAD+/+ mice as separate groups.
RNA Analysis
RT-PCR amplification from exon 7 to 11 from total heart RNA amplified the expected 600-base pair (bp) fragment in MCAD+/+ and MCAD+/− mice, and a 1.5-kb fragment in MCAD−/− mice (data not shown). Sequence analysis of the 1.5-kb fragment revealed that the amplified fragment consisted of exon 7 to exon 10 with 280 bp of pGEM plasmid sequence followed by exons 8–11. Some of the plasmid sequences, along with the pPGKNEOpA sequence, were deleted from this spliced mRNA.
Northern blot analysis revealed Acadm was normally expressed in all tissues analyzed from MCAD+/+ mice with the most robust expression occurring in brown fat, kidney, heart, skeletal muscle, and liver with minimal expression in the brain, white fat, and testes (Figure 2). Interestingly, although RT-PCR amplified an incorrectly spliced Acadm transcript, no Acadm transcripts were detected by northern blot analysis from MCAD−/− mice. These results strongly suggest that the mutant RNA is unstable and degraded rapidly or, alternatively, undergoes nonsense mediated RNA decay.
Figure 2 Northern Blot Analysis from MCAD−/− (n = 2) and MCAD+/+ (n = 2) Mice
Acadm message was detected from the heart, liver, brown fat, brain, kidney, and muscle (and white fat and testes, data not shown) of only MCAD+/+ mice. Most robust expression occurred in brown fat, kidney, heart, and skeletal muscle. MCAD−/− mice had no detectable message in all tissues examined.
Liver Enzyme Analyses
Immunoblot analyses of liver homogenates with anti-MCAD antisera demonstrated that the 42 kDa MCAD monomer was present in MCAD+/+ mice, but not in MCAD−/− mice (Figure 3). As a control analysis, anti–short-chain acyl-CoA dehydrogenase (SCAD) antisera revealed no differences in levels of expression of SCAD protein between MCAD+/+ and MCAD−/− mice (Figure 3).
Figure 3 Immunoblots of Liver Homogenates from MCAD+/+ and MCAD−/− Mice
These were probed with anti-MCAD antibody or anti-SCAD antibody. Homozygous MCAD−/− mice had no detectable MCAD protein. MCAD protein is only detectable under the MCAD protein–spiked (positive control) lane. As a control analysis, liver homogenates probed with anti-SCAD antibody revealed that SCAD protein was present in both MCAD+/+ and MCAD−/− mice. No MCAD positive-control protein is detected by anti-SCAD antibodies (MCAD lane). mw, molecular weight standards.
Enzyme activity was analyzed in mouse liver homogenates using the electron transport flavoprotein (ETF) reduction assay with octanoyl-CoA (C8:0) and palmitoyl-CoA (C16:0) as substrates. MCAD−/− mice had a significant reduction in ability to dehydrogenate octanoyl-CoA and a modest reduction in activity toward palmitoyl-CoA compared to MCAD+/+ mice (Table 1). Specifically, the dehydrogenation of octanoyl-CoA and palmitoyl-CoA substrates were reduced by 75% and by 30%, respectively, in MCAD−/− mice as compared to MCAD+/+ controls.
Table 1 Characteristics of MCAD-Deficient Mice
Values given are mean ± SD.
aMCAD+/+
n = 5 and MCAD−/−
n = 5.
bMCAD+/+
n = 8 litters and MCAD−/−
n = 10 litters.
cMCAD+/+
n = 5 and MCAD−/−
n = 6.
dMCAD+/+
n = 4 and MCAD−/−
n = 5.
eExpressed as a ratio relative to the internal standard.
Neonatal Deaths
Significant neonatal mortality was noted in MCAD−/− pups. Although equal numbers of pups were born from MCAD+/+ and MCAD−/− mice, only 41% of MCAD−/− mice survived to weaning as compared to 98% of MCAD+/+ mice (Table 1). The mechanism of neonatal loss remains undetermined. The MCAD−/− pups are abandoned more frequently than MCAD+/+ pups for unknown reasons. They are likely more prone to hypothermia than MCAD+/+. Because of the difficulty of distinguishing the MCAD−/− mutants from the MCAD+/− heterozygous pups by molecular analysis due to the insertion mutation, we could only compare MCAD+/+ × MCAD+/+ matings with MCAD−/− × MCAD−/− matings. Thus, we were unable to evaluate the pedigrees from heterozygous matings.
Fasting and Cold Intolerance
In order to examine the stress effects of fasting on MCAD-deficient mice, they were fasted for 24 h prior to analysis. MCAD−/− mice displayed lower serum glucose and elevated serum free fatty acid levels although neither result was significant, as compared to MCAD+/+ mice (Table 1).
To determine the effects of cold stress, mice were fasted for 18 h and placed in 4 °C environment for a 3-h period. The MCAD−/− mice were significantly compromised within this short period of cold stress, some severe enough to result in fatalities. After 1 h of the cold challenge, the average rectal temperature of MCAD−/− mice (n = 5) was 23.4 °C as compared with 35 °C for MCAD+/+ mice (n = 4). Rectal temperatures declined to unrecoverable temperatures of 16.7–19.2 °C in three of the five MCAD−/− mice. By the end of the 1.5-h mark, the two surviving MCAD−/− mice averaged 22.7 °C. In contrast, all four MCAD+/+ mice survived the 3-h cold stress, ending with an average rectal temperature of 33.6 °C.
Organic Acid and Acylcarnitine Analysis
Urine organic acid analysis revealed that MCAD−/− mice developed an organic acid profile similar to MCAD-deficient human patients. Specifically, when fasted for 18 h, MCAD−/− mice developed significantly elevated concentrations of adipic, suberic, and sebacic acids and hexanoylglycine as compared to MCAD+/+ controls, which showed trace to no detectable amounts of the same organic acids (Table 1). Adipic acid is not specific to MCAD deficiency. We also evaluated β-hydroxybutyric and acetoacetic concentrations and found no significant differences between MCAD genotypes (data not shown).
Comparison of MCAD+/+ and MCAD−/− mice revealed no significant differences in total serum carnitine concentrations between MCAD+/+ and MCAD−/− mice, but MCAD−/− mice had a 5- to 6-fold elevation of serum decenoylcarnitine evident in the acylcarnitine profile (Figure 4A). Bile acylcarnitine analysis revealed a similar acylcarnitine pattern as in serum (Figure 4B). However, the acylcarnitine profiles of the MCAD−/− mice are different from those of human MCAD-deficient patients (Figure 4C). Human patients present with elevated levels of C6, C8, and C10:1 acylcarnitines, as did the mutant mice; however, the predominant peak was C8 acylcarnitine in humans, whereas in the mouse it was C10:1 acylcarnitine.
Figure 4 Acylcarnitine Analyses
(A) Serum acylcarnitine analysis of MCAD+/+ (n = 4) and MCAD−/− mice (n = 4)
There are significant elevations in acylcarnitine species as indicated in MCAD−/− mice. Values shown are mean values ± standard deviation (SD). Asterisk indicates p < 0.002 and ‡ indicates p < 0.01.
(B) There are significant elevations in bile acylcarnitines of the same mice shown in (A) as indicated. Values shown are mean values ± SD. Asterisk indicates p < 0.001.
(C) Bile acylcarnitine profile of an MCAD−/− mouse compared to a human patient with MCAD deficiency. Internal standards are indicated by an asterisk.
Histopathology
Complete histopathologic examination of one group of mutant and MCAD+/+ control mice after a 24-h fast demonstrated diffuse microvesicular and macrovesicular hepatic steatosis in 6–8-wk-old MCAD−/− mice whereas MCAD+/+ mice had no histologic changes (Figure 5A and 5B). In another group of 4-wk-old MCAD+/+ and MCAD−/− mice fasted for 24-h, there were minimal to no abnormalities in all organs evaluated. Only the older MCAD−/− mice, therefore, demonstrated hepatic steatosis. In addition, we found sporadic cardiac lesions in multiple MCAD−/− mice.
Figure 5 Histopathology of MCAD+/+ and MCAD−/− Mice
(A) MCAD+/+ mice had no evidence of hepatic steatosis following a 24-h fast. Liver section with Oil-Red O stain.
(B) Hepatosteatosis in MCAD−/− mouse following a 24-h fast. Oil-Red O stained liver sections revealed severe and diffuse microvesicular and macrovesicular hepatic steatosis in MCAD−/− mice.
(C and D) Diffuse cardiomyopathy with chronic active multifocal myocyte degeneration and necrosis in MCAD−/− mice.
In one example, an MCAD−/− mouse had diffuse cardiomyopathy with chronic active multifocal myocyte degeneration and necrosis (Figure 5C and 5D). Changes in degenerating myocytes included swelling, pale staining, and, in portions of the sarcoplasm, replacement of myofibrils with finely granular eosinophilic material. Nuclei of affected myocytes were large, pale, and vesicular and had prominent nucleoli. In the most severely affected areas, there was loss of myocytes accompanied by fibrosis. In the wall of the aorta at the base of the heart there was multifocal degeneration of the elastic tissue, accompanied by multifocal collections of globular translucent yellow-brown pigment interpreted to be ceroid/lipofuscin. Similar deposits were scattered within adjacent adipose tissue.
Discussion
Successfully targeting Acadm produced a mouse model for MCAD deficiency with features that mimic the clinical, biochemical, and pathologic phenotype found in human patients. MCAD-deficient patients have abnormal plasma and urine metabolites associated with the medium chain–length enzyme specificity. MCAD-deficient patients [15] often display a characteristic urinary hexanoylglycine peak, as was seen in MCAD−/− mice. Acylcarnitine analysis indicated, however, mouse MCAD is more active toward longer chain substrates than the human MCAD enzyme. This finding is similar to that seen with very long-chain acyl-CoA dehydrogenase (VLCAD) where mouse VLCAD is most active toward C16 acyl-substrates as compared to human VLCAD with the most enzymatic activity toward C14 acyl-substrates [16].
ETF reduction assays of liver homogenates were performed to characterize the MCAD−/− mice at the enzymatic level. MCAD−/− mice had a significantly reduced ability to dehydrogenate C8-CoA, as is the case in human patients where MCAD activity is reduced to near zero with C8-CoA . This was less so with palmitoyl-CoA (C16:0). Because there was clearly no MCAD antigen detected in MCAD−/− mice, the residual dehydrogenase activity measured with these two substrates must represent the activity of other chain length–specific acyl-CoA dehydrogenases.
A high degree of neonatal mortality in MCAD−/− mice was a striking observation and appears to be analogous to the patterns of clinical episodes and mortality in MCAD-deficient patients. MCAD−/− mice exhibited significant neonatal mortality with approximately 60% of the MCAD−/− pups dying prior to weaning at 3 wk of age. Human patients present with hypoglycemia, hypoketonemia, and nonketotic organic aciduria precipitated by fasting, most frequently during the first 24 mo in life [1]. It is likely that neonatal MCAD−/− pups are manifesting sensitivity to fasting with decompensation in a short period of time if maternal milk is not ingested. In contrast, no mortality was noted in adult MCAD−/− mice unless challenged with cold stress and fasting. Under cold challenge conditions, however, MCAD−/− mice were unable to maintain body temperature. Brown fat is predominantly responsible for thermogenesis and normally expresses high levels of Acadm mRNA.
The microvesicular and macrovesicular hepatic steatosis seen in fasted MCAD−/− mice is consistent with the primary pathological finding seen in human MCAD patients with fasting stress. Sporadic cardiac lesions in MCAD−/− mice, however, were an interesting and unexpected finding. The diffuse cardiomyopathy with multifocal myocyte degeneration and necrosis observed in MCAD−/− mice has not been reported in human MCAD patients, however, cardiac arrhythmias and dysfunction have been reported in MCAD-deficient patients [17, 18]. Interestingly, cardiomyopathy has been observed in VLCAD deficiency [19] and other disorders of long chain fat metabolism such as severe CPT-1 and -2 deficiencies [1]. Thus it is tempting to relate the cardiac problems in the mouse to the apparent broader range of substrate utilization of mouse MCAD. The inconsistent liver and cardiac lesions in these mice is analogous with the significant inter- and intrafamilial phenotypic heterogeneity seen in MCAD deficiency in humans [1,20].
In comparisons with our experiences with the other mouse models for acyl-CoA dehydrogenase deficiencies, the overall phenotype of MCAD−/− mice is less severe than that found in LCAD−/− mice, yet more pronounced than the VLCAD−/− or SCAD−/− mouse models [16,21,22]. All of these mutants are cold intolerant and display varying degrees of fatty changes in liver, heart, and kidney. LCAD−/− mice show more spontaneous deaths and gestational losses than the other deficiencies [21]. The significant neonatal mortality in MCAD−/− mice is distinctive from these other mouse models suggesting a greater degree of sensitivity to fasting intolerance. The phenotypes of both the VLCAD−/− and SCAD−/− mice are relatively mild if the animals are not cold stressed [16,22]. The MCAD-deficient mouse offers new insights into the pathogenesis of mitochondrial β-oxidation deficiencies and will provide a robust tool to better understand the role of fatty acids in other relevant diseases.
Materials and Methods
Construction of MCAD targeting vector.
A neomycin resistance gene cassette [23] was subcloned into the SalI site of pGEM11Zf(+). The plasmid was digested with EcoRI and the overhangs were filled with Klenow enzyme. Subsequent ligation of the blunt ends recircularized the vector and destroyed the EcoRI site within the polylinker of the pGEMl1Zf(+) plasmid. Next, an 8-kb Acadm genomic fragment spanning exons 8, 9, and 10 and flanking intron sequences, originally obtained from a Lambda FIXII 129Sv mouse genomic library [24,25], was directionally cloned into the NotI and XhoI sites of pGEM11Zf(+). The vector was digested with BamHI and EcoRI to remove a 1.3-kb BamHI/EcoRI genomic fragment containing exon 10 and flanking intron sequences. The digested vector, without the 1.3-kb BamHI/EcoRI genomic fragment, was purified by gel purification and recircularized by ligation using three oligonucleotides: 5′-AATTGTCGACA-3′; 5′GATCGTCGACA-3′; and 5′-TCGATGTCGAC-3′. The recircularized vector, resulting from the ligation of the long arm to the short arm of homology, contained a unique SalI site where the 1.3-kb exon 10 region was deleted.
Generation of MCAD-deficient mice.
The Acadm insertion vector was linearized by SalI digestion, the site of the 1.3-kb genomic fragment deletion, and electroporated into E14–1 ES cells (a kind gift from R. Kuhn), derived from 129P2 mice. Correctly targeted Acadm insertion vector was designed to undergo gap repair of the 1.3-kb deletion upon double stranded–break repair [26] during homologous recombination. Southern blot analysis was conducted to confirm homologous recombination. Genomic DNA from G418 resistant ES colonies was digested with EcoRI, electrophoresed, blotted, and probed with an 850-bp probe (probe A) generated by PCR from Acadm exon 10 to intron 10. This DNA fragment is not present within Acadm insertion vector and was expected to hybridize to a 13.2-kb genomic DNA fragment upon homologous recombination. Correctly targeted ES cell clones were microinjected into B6 blastocysts to generate chimeric mice. Chimeric mice were subsequently backcrossed to B6 and 129P2 mice to produce gene-targeted mice AcadmtmUab1/+ (MCAD+/−) and eventually AcadmtmUab1/tm1Uab MCAD−/− (B6;129) mice. Mice were negative for mouse pathogens based on serological assays for ten different viruses, aerobic bacterial cultures of nasopharynx and cecum, examinations for endo- and ectoparasites, and histopathology of all major organs.
RNA analysis.
Total RNA was isolated from the heart of 30 day old MCAD+/+, MCAD+/−, and MCAD−/− mice by standard techniques using guanidinium thiocyanate method [27]. Reverse transcription was performed using random oligonucleotides as recommended by the manufacturer (Clontech, Mountain View, California, United States). PCR was subsequently performed using oligonucleotides specific to exon 7 and exon 11 of Acadm. PCR amplifications were performed as described above. PCR fragments were subsequently sequenced after subcloning into pGEM-T Easy vector (Promega, Madison, Wisconsin, United States).
In order to determine the extent of Acadm mRNA expressed from the MCAD−/− mice, northern blot analysis was performed. Total RNA was isolated from heart, liver, brown fat, brain, kidney, skeletal muscle, white fat, and testes of 3-mo-old MCAD+/+ and MCAD−/− mice using the Ultraspec RNA Isolation Kit (BIOTEX Laboratories, Inc., Houston, Texas, United States) as per manufacturer's protocol. Ten μg of total RNA from each sample was loaded onto a 0.8% agarose-formaldehyde gel, transferred to nitrocellulose filter (Hybond N; GE Healthcare Amersham Biosciences Corp., Piscataway, New Jersey, United States), and hybridized with 32P-radiolabeled full-length mouse Acadm cDNA probe using standard procedures [28]. Hybridizations were performed under highly stringent conditions (42 °C in 2× SSC, 50% formamide, 10% dextran sulfate, 5× Denhardt's reagent, 1% SDS, and salmon sperm DNA) for 18 h. The hybridized filter was washed two times in 4× SSC, 0.1% SDS and two times in 2× SSC; 0.1% SDS at 55 °C for 1 h. The filter was exposed to autoradiographic film (Hyperfilm MP; GE Healthcare Amersham Biosciences, Piscataway, New Jersey, United States). Replicate agarose-formaldehyde gels were stained by ethidium bromide to verify equal RNA loading.
Immunoblot analysis of MCAD protein.
To evaluate the quantity of MCAD protein in mouse tissue, liver samples from 6–8-wk-old MCAD+/+ (n = 1) and MCAD−/− (n = 3) mice were immediately frozen in liquid nitrogen and stored at −80 °C. For analysis, tissue was homogenized and lysed in RIPA buffer (1× PBS, 1% Nonidet P-40, 0.5% sodium deoxycholate, and 1% SDS) with 10% glycerol and Complete Protease Inhibitor, (Roche Diagnostics Corporation, Indianapolis, Indiana, United States), 1 mM of phenylmethylsulfonylfluoride, and 1 mM of sodium orthovanadate. Lysates were quantified by Bradford BCA protein assay (Bio-Rad, Hercules, California, United States). Protein lysates were denatured, separated in 8% SDS-PAGE, and transferred overnight onto a 0.45 μm nitrocellulose membrane (Schleicher and Schuell, Keene, New Hampshire, United States). After blocking with 5% nonfat milk in phosphate-buffered saline with 0.1% Tween-20, the membrane was immunoblotted overnight at 4 °C with 1:500 dilution of an anti-MCAD antibody. Blots were incubated in anti-mouse IgG HRP-conjugated secondary antibody for 2–4 h at room temperature using standard procedures and developed by chemiluminiscence (Renaissance, NEN Lifesciences Products, Boston, Massachusetts, United States).
Liver enzyme activity.
In order to evaluate MCAD activity in mice, liver homogenates were prepared from MCAD+/+ (n = 5) and MCAD−/− mice (n = 5). The sensitive and highly specific anaerobic ETF reduction assay was used on tissue extracts with octanoyl-CoA (C8) and palmitoyl-CoA (C16) as substrate as previously described [29].
Fasting and cold challenge.
Eight-wk-old MCAD+/+ and MCAD−/− mice were fasted for 18 h (cold tolerance experiments) or 24 h (serum glucose, free fatty acid, organic acid, and carnitine experiments) prior to analysis. Glucose concentration was measured in 10 μl sera using an Ektachem DT II system (Johnson and Johnson Clinical Diagnostics, Rochester, New York, United States). Total non-esterified fatty acids (NEFA) were measured by an enzymatic, colorimetric method (“NEFA-C” reagents, Wako Diagnostics, Richmond, Virginia, United States). The assay was modified to accommodate a reduced sample size (10 μl) and use of a microplate reader for measurement of optical density at 550 nm. Urine organic acid analyses were performed using gas chromatography-mass spectroscopy as previously described [22,30], except tetracosane (C24; Sigma, St. Louis, Missouri, United States) was used as the internal standard, and quantitative determinations were made based on comparison with synthetic calibration standards. Acylcarnitine analyses in serum and bile were conducted using electrospray tandem mass spectrometry [31,32].
Histopathology.
Twelve mice were examined for gross and histologic abnormalities, including one male and one female MCAD−/− mouse 18-mo-old, one male and one female MCAD+/+ mouse 6-mo-old, two male and two female MCAD−/− mice 4-wk-old, and two male and two female MCAD+/+ mice 4-wk-old. Kidney, spleen, pancreas, liver, brain, heart, testicles, ovaries, and skeletal muscle were fixed by immersion in buffered 10% formalin, processed routinely for paraffin sectioning, sectioned at 5 μm, and stained with hematoxylin and eosin. Frozen liver sections were prepared using standard methods and sections were stained with Oil-red-O. Slides were examined without knowledge of genotype or age. Other mice were examined because of sporadic clinical disease. Those with visible cardiac enlargement were evaluated for cardiac lesions.
Statistical analyses.
Results between groups were tested for statistical significance using Student's t-test. A p < 0.05 was set as significant.
Supporting Information
Accession Number
The GenBank (http://www.ncbi.nlm.nih.gov/Genbank) accession number for the mouse gene Acadm is U07159.
We thank Sushama Varma for technical assistance. This research was supported by the University of Alabama at Birmingham (UAB) Comprehensive Cancer Center (Oligonucleotide and Transgenic Animal Shared Facilities) grant P30-CA13148, UAB Musculoskeletal Disease and Arthritis Center (Gene Targeting Core Facility) grant P60-AR20614, UAB Clinical Nutrition Research Center grant P30-DK-56336, and by National Institutes of Health grants R01-RR02599 (PAW), T32-RR-07003 (RJT), K01-RR00129 (RJT), RO1-DK45482 (JV), and DK54936 (JV).
Competing interests. The authors have declared that no competing interests exist.
Author contributions. RJT, JV, PR, DM, and PAW conceived and designed the experiments. RJT, DAH, LT, JV, PR, DM, JDS, and PAW performed the experiments. RJT, DAH, LT, JV, JDS, PR, DM, TRS, and PAW analyzed the data and contributed reagents/materials/analysis tools. RJT, JV, PR, DM, TRS, JDS, and PAW wrote the paper.
Abbreviations
bpbase pair
ESembryonic stem
ETFelectron transport flavoprotein
MCADmedium-chain acyl-CoA dehydrogenase
SCADshort-chain acyl-CoA dehydrogenase
SDstandard deviation
VLCADvery long-chain acyl-CoA dehydrogenase
==== Refs
References
Rinaldo P Matern D Bennett MJ 2002 Fatty acid oxidation disorders Annu Rev Physiol 64 477 502 11826276
Matsubara Y Kraus JP Ozasa H Glassberg R Finocchiaro G 1987 Molecular cloning and nucleotide sequence of cDNA encoding the entire precursor of rat liver medium chain acyl coenzyme A dehydrogenase J Biol Chem 262 10104 10108 3611054
Kelly DP Kim JJ Billadello JJ Hainline BE Chu TW 1987 Nucleotide sequence of medium-chain acyl-CoA dehydrogenase mRNA and its expression in enzyme-deficient human tissue Proc Natl Acad Sci U S A 84 4068 4072 3035565
Ikeda Y Keese SM Fenton WA Tanaka K 1987 Biosynthesis of four rat liver mitochondrial acyl-CoA dehydrogenases: In vitro synthesis, import into mitochondria, and processing of their precursors in a cell-free system and in cultured cells Arch Biochem Biophys 252 662 674 3813556
Kolvraa S Gregersen N Christensen E Hobolth N 1982 In vitro fibroblast studies in a patient with C6-C10-dicarboxylic aciduria: Evidence for a defect in general acyl-CoA dehydrogenase Clin Chim Acta 126 53 67 7172449
Stanley CA Hale DE Coates PM Hall CL Corkey BE 1983 Medium-chain acyl-CoA dehydrogenase deficiency in children with non-ketotic hypoglycemia and low carnitine levels Ped Res 17 877 884
Rhead WJ Amendt BA Fritchman KS Felts SJ Dicarboxylic aciduria: Deficient [1-14 C] octonoate oxidation and medium-chain acyl-CoA dehydrogenase in fibroblasts Science 221 73 75 6857268
Coates PM 1992 Historical perspective of medium-chain acyl-CoA dehydrogenase deficiency: A decade of discovery Prog Clin Biol Res 375 409 423 1438385
Matsubara Y Narisawa K Tada K 1992 Medium-chain acyl-CoA dehydrogenase deficiency: Molecular aspects Eur J Pediatr 151 154 159 1601002
Ding JH Bross P Yang BZ Lafolla AK Millington DS 1992 Genetic heterogeneity in MCAD deficiency: Frequency of K329E allele and identification of three additional mutant alleles Coates PM Tanaka K New developments in fatty acid oxidation New York Wiley-Liss 479 488
Workshop on Molecular Aspects of MCAD Deficiency 1992 Mutations causing medium-chain acyl-CoA dehydrogenase deficiency: A collaborative compilation of the data from 172 patients Coates PM Tanaka K New developments in fatty acid oxidation New York Wiley-Liss 499 506
Matern D Rinaldo P
2003
Medium-chain acyl-coenzyme A dehydrogenase deficiency
Available:
http://www.geneclinics.org/servlet/access?db=geneclinics&site=gt&id=8888891&key=YpeEakmrXPZ3i&gry=&fcn=y&fw=nYYD&filename=/profiles/mcad/index.html.
Accessed 21 July 2005.
Capecchi MR 1989 Altering the genome by homologous recombination Science 244 1288 1292 2660260
Koller BH Smithies O 1992 Altering genes in animals by gene targeting Annu Rev Immunol 10 705 730 1591000
Rinaldo P O'Shea JJ Coates PM Hale DE Stanley CA 1988 Medium-chain acyl-CoA dehydrogenase deficiency. Diagnosis by stable-isotope dilution measurement of urinary n-hexanoylglycine and 3-phenylpropionylglycine N Engl J Med 319 1308 1313 3054550
Cox KB Hamm DA Millington DS Matern D Vockley J 2001 Gestational, pathologic and biochemical differences between very long-chain acyl-CoA dehydrogenase deficiency and long-chain acyl-CoA dehydrogenase deficiency in the mouse Hum Mol Genet 10 2069 2077 11590124
Feillet F Steinmann G Vianey-Saban C de Chillou C Sadoul N 2003 Adult presentation of MCAD deficiency revealed by coma and severe arrythmias Intensive Care Med 29 1594 1597 12897989
Maclean K Rasiah VS Kirk EP Carpenter K Cooper S 2005 Pulmonary haemorrhage and cardiac dysfunction in a neonate with medium-chain acyl-CoA dehydrogenase (MCAD) deficiency Acta Paediatr 94 114 116 15858970
Vianey-Saban C Divry P Brivet M Nada M Zabot MT 1998 Mitochondrial very-long-chain acyl-coenzyme A dehydrogenase deficiency: Clinical characteristics and diagnostic considerations in 30 patients Clin Chim Acta 269 43 62 9498103
Duran M Hofkamp M Rhead WJ Saudubray JM Wadman SK 1986 Sudden child death and ‘healthy' affected family members with medium-chain acyl-coenzyme A dehydrogenase deficiency Pediatrics 78 1052 1057 3786030
Kurtz DM Rinaldo P Rhead WJ Tian L Millington DS 1998 Targeted disruption of mouse long-chain acyl-CoA dehydrogenase gene reveals crucial roles for fatty acid oxidation Proc Natl Acad Sci U S A 95 15592 15597 9861014
Wood PA Amendt BA Rhead WJ Millington DS Inoue F 1989 Short-chain acyl-coenzyme A dehydrogenase deficiency in mice Pediatr Res 25 38 43 2919115
Soriano P Montgomery C Geske R Bradley A 1991 Targeted disruption of the c-src proto-oncogene leads to osteopetrosis in mice Cell 64 693 702 1997203
Tolwani RJ Farmer SC Wood PA 1994 Molecular cloning and characterization of the mouse medium-chain acyl-CoA dehydrogenase cDNA Genomics 23 247 249 7829081
Tolwani RJ Farmer SC Johnson KR Davisson MT Kurtz DM 1996 Structure and chromosomal location of the mouse medium-chain acyl-CoA dehydrogenase-encoding gene and its promoter Gene 170 165 171 8666240
Szostak JW Orr-Weaver TL Rothstein RJ Stahl FW 1983 The double-strand-break repair model for recombination Cell 33 25 35 6380756
Chomczynski P Sacchi N 1987 Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction Anal Biochem 162 156 159 2440339
Sambrook J Fritsch EF Maniatis T 1989 Molecular cloning: A laboratory manual, 2nd ed Cold Spring Harbor (New York) Cold Spring Harbor Laboratory 1,659 p.
Mohsen AW Vockley J 1995 Identification of the active site catalytic residue in human isovaleryl-CoA dehydrogenase Biochemistry 34 10146 10152 7640268
Wood PA Armstrong D Sauls D Davisson MT 1988 Screening mutant mice for inborn errors of metabolism Lab Anim Sci 38 15 19 3367620
Rashed MS Ozand PT Bucknall MP Little D 1995 Diagnosis of inborn errors of metabolism from blood spots by acylcarnitines and amino acids profiling using automated electrospray tandem mass spectrometry Pediatr Res 38 324 331 7494654
Rinaldo P Hahn SH Matern D 2005 Inborn errors of amino acid, organic acid, and fatty acid metabolism Burtis CA Ashwood ER Bruns D Tietz textbook of clinical chemistry, 4th ed. St. Louis (Missouri) Elsevier pp. 2207–2247.
|
16121256
|
PMC1189074
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Aug 19; 1(2):e23
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010023
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1612125710.1371/journal.pgen.001002505-PLGE-RA-0103R2plge-01-02-10Research ArticleBioinformatics - Computational BiologyGenetics/GenomicsGenetics/Comparative GenomicsGenetics/Gene ExpressionSaccharomycesLocal Regulatory Variation in Saccharomyces cerevisiae
Local Regulatory VariationRonald James 12Brem Rachel B 2Whittle Jacqueline 23¤Kruglyak Leonid 234*1 Department of Genome Sciences, University of Washington, Seattle, Washington, United States of America
2 Fred Hutchinson Cancer Research Center, Seattle, Washington, United States of America
3 Howard Hughes Medical Institute, Seattle, Washington, United States of America
4 Lewis–Sigler Institute for Integrative Genomics and Department of Ecology and Evolutionary Biology, Princeton University, Princeton, New Jersey, United States of America
Frankel Wayne EditorJackson Laboratory, United States of America¤Current address: Infectious Disease Research Institute, Seattle, Washington, United States of America
*To whom correspondence should be addressed. E-mail: [email protected] 2005 19 8 2005 1 2 e2513 5 2005 1 7 2005 Copyright: © 2005 Ronald et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Naturally occurring sequence variation that affects gene expression is an important source of phenotypic differences among individuals within a species. We and others have previously shown that such regulatory variation can occur both at the same locus as the gene whose expression it affects (local regulatory variation) and elsewhere in the genome at trans-acting factors. Here we present a detailed analysis of genome-wide local regulatory variation in Saccharomyces cerevisiae. We used genetic linkage analysis to show that nearly a quarter of all yeast genes contain local regulatory variation between two divergent strains. We measured allele-specific expression in a diploid hybrid of the two strains for 77 genes showing strong self-linkage and found that in 52%–78% of these genes, local regulatory variation acts directly in cis. We also experimentally confirmed one example in which local regulatory variation in the gene AMN1 acts in trans through a feedback loop. Genome-wide sequence analysis revealed that genes subject to local regulatory variation show increased polymorphism in the promoter regions, and that some but not all of this increase is due to polymorphisms in predicted transcription factor binding sites. Increased polymorphism was also found in the 3′ untranslated regions of these genes. These findings point to the importance of cis-acting variation, but also suggest that there is a diverse set of mechanisms through which local variation can affect gene expression levels.
Synopsis
Variation in DNA sequences in and around a gene can contribute to differences between individuals by affecting the gene's expression. The authors have used a variety of methods to characterize this local DNA sequence variation on a large scale in two strains of the budding yeast Saccharomyces cerevisiae. Their results suggest that the expression levels of a sizeable fraction of genes are affected by local sequence variation. Many local variants alter the expression of only one of two copies of a gene in diploid hybrid yeast, but other local variants can affect both copies equally. The authors also found that sequence variation in particular regions of DNA near genes, both upstream and downstream of coding sequences and especially in transcription factor binding sites, is most likely to affect gene expression. These results provide a detailed view of local sequence variation that affects the expression of nearby genes in S. cerevisiae.
Citation:Ronald J, Brem RB, Whittle J, Kruglyak L (2005) Local regulatory variation in Saccharomyces cerevisiae. PLoS Genet 1(2): e25.
==== Body
Introduction
Much effort has recently been devoted to understanding the genetic basis of natural variation in gene expression levels. Linkage mapping, in which gene expression levels are treated as quantitative traits in linkage analysis, has been used to characterize the heritability of these expression traits and to identify the loci that control them [1–7]. Analysis of allele-specific expression (ASE), in which the relative amount of each allele in a diploid is assayed, has been used to identify genes with variation in cis-acting regulatory elements and to distinguish between cis and trans control [8–12]. These two approaches, linkage mapping and ASE analysis, provide distinct and complementary axes of information: positional and mechanistic, respectively.
We previously performed linkage analyses on gene expression levels in haploid segregants from a cross between two Saccharomyces cerevisiae strains (BY4716 [BY], isogenic to S288C, and RM11-1a [RM], a wild vineyard strain) [1]. We identified two types of linkages: those in which the expression level of a gene is linked to its own locus in the genome (“self-linkages”), and those in which the expression level is linked to a distinct locus elsewhere in the genome. The latter linkage indicates that variation at a distant locus acts in trans to affect expression of a gene [13]. In contrast, although self-linkage implies that local variation in the vicinity of the gene affects the expression of that gene, the mechanism through which that variation acts may be either cis or trans, under the classical definitions of the terms. For example, polymorphisms in the promoter region that affect chromatin structure or transcription factor binding sites, or polymorphisms in the coding sequence or 3′ untranslated region that affect mRNA stability, would be expected to act in cis, altering the abundance of the transcript in an allele-specific manner in a diploid [11]. Alternatively, amino acid changes within the coding sequence that affect the activity of the gene product, or codon usage changes that affect the level of protein, may lead to a change in gene expression either directly through autoregulation of the gene by its protein product or indirectly through a pathway of intermediates. Such local variation affecting the protein product, although present in only one allele in a heterozygous diploid, would act in trans to alter the expression of both alleles.
Here we performed a hypothesis-driven linkage analysis to improve the sensitivity with which genes subject to local regulatory variation are identified. We then used ASE measurements to estimate the fraction of local variation that acts mechanistically in cis. The observed high proportion of cis-acting effects in the genes assayed for ASE prompted us to perform a global analysis of polymorphism in genes with local regulatory variation, with emphasis on non-coding regions, to identify the signature of functional sequence differences. We found that genes with local regulatory variation are concentrated in areas of the genome that are highly polymorphic between the parent strains in both the genic and intergenic regions. We also found that genes showing evidence of local regulatory variation have further enrichment of polymorphisms in their promoter regions, in their 3′ untranslated regions, and specifically in predicted transcription factor binding sites, underscoring that fine-scale sequence variation in these regions is likely to have functional consequences.
Results
Many Genes Show Local Regulatory Variation
We previously reported self-linkage of 578 genes in a genome-wide analysis of a smaller cross (consisting of 86 segregants) between these strains [13]. To more sensitively identify genes subject to local regulatory variation, we tested for linkage only at the single marker closest to the gene in question, using previously reported gene expression and genotype data for 112 segregants from the same cross [14]. This hypothesis-driven approach reduced the number of statistical tests performed for each expression trait from a whole-genome scan of approximately 3,000 markers to a single-marker test and therefore increased the power to detect self-linkages. A total of 1,428 transcript levels (25% of the 5,727 transcripts tested) showed significant linkage at a permutation-based false discovery rate less than 0.05 (corresponding to a nominal p-value of 0.012). Multipoint linkage analysis showed that the gene encoding the linking transcript fell within the 1 logarithm of odds (LOD) support interval of the linkage peak in 92% of cases, and within the 2 LOD interval in 97% of cases. Based on genome-wide linkage results, we estimate that approximately 6% of true self-linkages may be due to polymorphisms in distinct trans-acting regulatory genes located close to their targets by chance (see Materials and Methods). Thus, a larger fraction of all genes (20%–25%) than previously reported contains local regulatory variation. The genes showing self-linkage, their effect size estimates, and their LOD support intervals are shown in Table S1.
ASE of Genes with Local Regulatory Variation
In order to directly test whether local regulatory variation acts in cis, we assayed 77 genes showing self-linkage for the presence of ASE in a diploid hybrid of the two parent strains, BY and RM. These genes were chosen on the basis of showing highly significant self-linkage (p < 10−8) and at least a 1.2-fold difference in expression between segregants bearing the BY and RM alleles, such that no false positives and only one chance trans linkage due to a nearby gene were expected (see Materials and Methods). Of the 77 assayed genes, 44 (57%) showed ASE at a nominal p-value of less than 0.05 (Table S2). In only two of the 44 cases, ASE favored the allele associated by linkage analysis with lower expression. For comparison, we tested ASE in a control set of 16 genes that were selected because they showed heritable variation of equivalent effect size, with transcript levels linked to other loci in the genome, but without evidence of significant self-linkage. In this set of 16, we observed only two results with a nominal p-value of less than 0.05, a rate slightly higher but not significantly different from that expected by chance (Table S2).
We next sought an estimate of the total fraction of assayed genes with ASE, correcting for the fact that some true cases of ASE may not have reached a nominal p-value less than 0.05 in our experiment. To obtain this estimate, we used the method of Storey and Tibshirani [15], which considers the complete distribution of p-values to estimate the rate of true alternative hypotheses in a large set of statistical tests. This procedure estimated a true rate of ASE of 78% in the 77 genes tested. Such a high proportion of ASE is consistent with the results of Doss et al. [16], who showed that 18 of 28 self-linkages in a cross between two mouse strains had allele-specific effects, and those of Wittkopp et al. [11], who found that ASE was common in an F1 hybrid of Drosophila melanogaster and D. simulans among 29 genes with interspecific expression differences.
Our results also suggest that trans-acting local variation is likely to be responsible for a minority of the self-linkages tested. Indeed, a number of genes with self-linkage showed nearly equal expression of the two alleles in a diploid hybrid (Table S2). Although it has been argued that self-linkage without ASE is most likely due to a closely linked gene that happens to regulate the gene in question in trans [16], our linkage analyses suggest that such nearby regulators may not account for all local trans-acting effects in S. cerevisiae (see Materials and Methods). Instead, we believe that in some cases local trans-acting effects are best explained by a polymorphism in the gene itself that acts in trans through a feedback loop. For example, expression of the regulatory gene AMN1 [17] showed strong self-linkage but weak ASE. Segregants that carry the BY allele of AMN1 show a 2.2-fold increase in its expression relative to segregants that carry the RM allele, but in the diploid hybrid, the ratio of expression of the BY allele to expression of the RM allele is 1.12 ( p = 0.067; 95% confidence interval 0.99–1.27). We previously hypothesized that the functional polymorphism in AMN1 is a single nucleotide substitution that leads to a missense amino acid change in the BY coding sequence at residue 368, replacing a highly conserved aspartic acid with valine [13]. The Amn1 protein has been proposed to indirectly negatively regulate itself as well as the daughter-specific transcriptional program, which includes the genes DSE1 and DSE2 [17]. DSE1 and DSE2 are upregulated 15.2- and 20.4-fold, respectively, in segregants bearing the BY allele at AMN1, consistent with the hypothesis that the negative regulator function of Amn1 is impaired in the BY strain. To determine whether the D368V amino acid change is the polymorphism that causes AMN1 to show self-linkage, we engineered a BY strain carrying aspartic acid at residue 368 and measured gene expression levels using microarrays. We observed a 2.3-fold upregulation in the expression of AMN1 in the original BY strain carrying the valine, relative to the engineered BY strain carrying aspartic acid at position 368. This result confirms that the coding mutation D368V is the predominant factor responsible for variation in expression of AMN1. In addition, we found that the original BY strain showed 9.7- and 15.3-fold upregulation of DSE1 and DSE2, respectively, relative to the engineered strain carrying the aspartic acid; this further suggests that an aspartic acid at position 368 is sufficient to restore trans-regulatory function to Amn1 in the BY strain. This example directly illustrates that a change in protein sequence can lead to a difference in the encoding gene's expression, and that such a trans-acting mechanism affects both alleles equally through a feedback loop.
Both cis-acting and trans-acting local polymorphisms could in principle affect expression of the same gene, so we next analyzed whether the expression changes in genes with cis-acting regulatory variation could be attributed primarily to this variation. Although the linkage study measured expression with microarrays and the ASE measurements were carried out by quantitative PCR, we noted that there was reasonable agreement between linkage results and ASE results in the fold-change estimates for the 44 nominally significant genes (Figure 1). Thus, for many of these genes, the linkage signal can be accounted for entirely by the polymorphisms producing ASE. Results for the 33 genes that were not significant for ASE, as well as the 16 controls, are shown in Figure S1.
Figure 1 Comparison of Linkage and ASE Fold-Change Estimates
Points represent the fold-change estimates from linkage analysis (horizontal axis) and from ASE experiments (vertical axis) for the 44 genes with nominally significant ASE (p < 0.05). Horizontal and vertical bars represent 95% confidence intervals. The solid line (y = x) represents equal fold-change estimates in the two experiments. The dashed line (y = 0.85x + 0.09, R
2 = 0.68) is the best fit, excluding one outlier and the two genes showing ASE favoring the opposite allele than that expected from linkage analysis (open circles).
Although we assayed a subset of yeast genes for ASE, our results may provide insight into the prevalence of ASE genome-wide. The distribution of self-linkage effect sizes of the 77 assayed genes was not different from that of all 446 genes with effect size greater than 1.2 (Kolmogorov–Smirnov test, p = 0.77), suggesting that the prevalence of ASE among the latter may be well represented by estimates from the assayed set. Indeed, the estimates may be appropriate for all genes with self-linkage irrespective of linkage effect size, as the prevalence of ASE among the genes assayed was not a function of this quantity: 21 of 39 genes with linkage effect size less than 1.34 and 23 of 38 genes with linkage effect size greater than 1.34 showed ASE at p < 0.05. In addition, the distribution of linkage effect sizes was indistinguishable among the 44 genes showing ASE at p < 0.05 and the remaining 33 genes (Kolmogorov–Smirnov test, p = 0.62). We therefore hypothesize that the subset of genes assayed here may be representative of self-linkages across the genome, and that a substantial fraction of all 1,428 self-linking genes is likely to show ASE due to the presence of cis-acting local regulatory variation.
Genes with Local Regulatory Variation Map to Regions with Increased Polymorphism
Because the ASE experiments in a selected set of genes led us to hypothesize that a substantial amount of local regulatory variation is due to cis-acting polymorphisms, we sought to analyze such variation genome-wide. We carried out a sequence comparison between the BY and RM strains for regions containing 5,182 genes with high-quality alignments between the genomes of the two strains. These 5,182 genes included 1,233 of the 1,428 genes showing self-linkage. Because the divergence in non-coding regions between BY and RM is approximately 0.005 (five polymorphisms per 1,000 bases) and most intergenic regions in S. cerevisiae are smaller than 1 kb, polymorphisms between these strains in non-coding regulatory regions are sufficiently infrequent that ascertainment of genes based on self-linkage should produce a detectable increase in divergence due to the presence of causative regulatory polymorphisms. We indeed observed a greater rate of polymorphisms in the upstream regions of the 1,233 genes showing self-linkage (0.0071, 95% confidence interval 0.0067–0.0075) compared to upstream regions of 3,949 genes without self-linkage (0.0040, 95% confidence interval 0.0038–0.0042). This increase was not limited to the region most likely to contain regulatory elements, but rather extended for at least 1 kb upstream (Figure S2). In fact, the polymorphism rate was also higher in the coding and downstream regions of genes showing self-linkage (0.0044 versus 0.0029 and 0.0069 versus 0.0040, respectively). These effects appear to be due in part to the correlation in the level of divergence over both short and long distances (Figure 2), with highly polymorphic regions tending to show a high density of genes with local regulatory variation (Figure 3).
Figure 2 Autocorrelation in Divergence as a Function of Distance between Genes
Each point indicates the correlation in the rate of substitution polymorphisms in the coding sequences of genes separated by at least x − 2.5 kb and at most x kb, for x = 2.5 kb, 5 kb, 7.5 kb, …, 100 kb. Correlations were similar for rates of substitution polymorphisms in intergenic regions (data not shown). ORFs, open reading frames.
Figure 3 Rates of Substitution Polymorphisms between BY and RM
Chromosome numbers are indicated on the left; large black circles represent centromere locations. Small points indicate gene locations: red, genes showing self-linkage; green, genes not showing self-linkage. The black jagged line represents the rate of substitution polymorphisms per 1,000 bp, with a maximum of 25 polymorphisms per 1,000 bases. The highly diverged Chromosome 2 region and relatively non-diverged Chromosome 7 region described in the text are indicated by boxes.
For example, gross inspection of Figure 3 reveals a large region on Chromosome 7 that shows a low rate of polymorphism (approximately 0.6 polymorphisms per 1,000 bases over 200 kb) and, as a result, a low rate of genes showing self-linkage (three genes with self-linkage and 88 without). In contrast, an extended region on Chromosome 2 shows an elevated rate of polymorphism (approximately eight polymorphisms per 1,000 bases over 200 kb) and a large number of genes showing self-linkage (46 genes with self-linkage and 52 without). We found that the Chromosome 2 region showed a much higher polymorphism rate than the Chromosome 7 region in a comparison of BY and YJM789 [18] (a wild strain approximately as divergent from BY as RM), and a comparison of YJM789 and RM yielded a similar result, but essentially no difference in divergence was found between S. cerevisiae and S. paradoxus (data not shown). Thus, the heterogeneous pattern of divergence in these two regions apparently occurred in the S. cerevisiae lineage, but is not unique to a single strain. This is consistent with previous work showing extended regions of higher or lower diversity between yeast strains [19], presumably as a result of the complex, stochastic interbreeding and recombination histories of yeast strains that have led to differences in time to the most recent common ancestor for different chromosomal regions. The effects of such forces appear to be absent when more distantly related yeast genomes separated by a species barrier, such as S. cerevisiae and S. paradoxus, are compared [20].
A higher rate of local regulatory variation (and hence a higher rate of genes showing self-linkage) is expected in more polymorphic regions of the genome. Such a result was also suggested by Hubner et al. [7] and observed by Doss et al. [16]. However, we were interested in whether we could use an enrichment for polymorphisms in specific regions of the genes (for example, likely regulatory regions) to identify those regions most likely to carry causative regulatory variants. Therefore, we sought to correct for the observed correlation in divergence and the resulting elevated rate of polymorphisms across entire genes, including upstream and downstream intergenic regions (Figure S2). We employed two approaches to do so. In both, we counted the number of substitution single nucleotide polymorphisms (SNPs) in each 100-bp bin from 1,000 bp upstream to translation start, treating each coding sequence as a single bin, and in each 100-bp bin from translation stop to 1,000 bp downstream. In the first approach, we performed logistic regression using all bins simultaneously in the model, estimating the significance of each bin conditional on all others; any overall, nonindependent elevation of polymorphism is factored out by this procedure (Figure S2). In the second approach, we directly matched genes by location, analyzing only those 1,125 genes with self-linkage that could be paired with exactly one unique gene without self-linkage located as close as possible but no more that 5,000 bp away. Both approaches showed significantly increased polymorphism, both upstream of translation start and downstream of translation stop, in genes showing self-linkage (Figure 4). The strongest enrichment for polymorphisms was found in the upstream region from 101 to 200 bp upstream of start (Figure 4). The enrichment in the downstream 3′ untranslated regions was not due to overlap with upstream regions of adjacent genes (see Materials and Methods; Figure S3). Thus, the increased polymorphism in genes with self-linkage in upstream and downstream regions is likely to reflect different classes of functional regulatory variation rather than local differences in the level of divergence.
Figure 4 Increased Divergence in the Promoter Region and 3′ Untranslated Region in Genes Showing Self-Linkage
The numbers of substitution SNPs are counted over each 100-bp bin centered at the position from translation start (or from translation stop for downstream bins) indicated on the x-axis, with the exception of the coding sequence, which is treated as a single bin. Solid line: the difference in inter-strain divergence between 1,125 position-matched pairs of genes, calculated by subtracting the average divergence in genes without self-linkage from the average divergence in genes with self-linkage (the left y-axis indicates difference in divergence in substitutions per basepair). Dotted line: −log10
p-values from logistic regression of self-linkage status on SNP rate in each bin independently, for the 1,125 pairs of genes (the right y-axis indicates −log10[p]); the dashed line shows p = 0.0024 (p = 0.05 after a Bonferroni correction for the 21 bins tested).
Enrichment of Polymorphisms in Transcription Factor Binding Sites
Our finding that ascertainment based on self-linkage most strongly enriches for polymorphisms from 101 to 200 bp upstream of translation start is consistent with the results of Cliften et al. [21], who found that intergenic conservation across six Saccharomyces species is highest in this region, and with the results of Harbison et al. [22], who found that transcription factor binding site occupancy is highest in this region. We therefore sought to analyze whether polymorphisms in this critical upstream region in genes showing self-linkage were more likely to be located in predicted transcription factor binding sites based on the dataset of Harbison et al. [22]. In order to control for both the increased rate of polymorphism in the upstream regions of genes with self-linkage and the decreased rate of polymorphism in predicted transcription factor binding sites due to negative selection on these functional sites, we performed chi-squared tests comparing the number of substitution SNPs in transcription factor binding sites versus non-sites in the upstream regions of genes with self-linkage versus genes without self-linkage.
This analysis showed that predicted transcription factor binding sites in the upstream regions of genes with self-linkage were modestly enriched for polymorphisms relative to surrounding bases (Table 1). When all genes showing self-linkage were analyzed, there was a trend for the enrichment to be greater at predicted sites that were conserved between species, and greatest at predicted sites that are bound by transcription factors in vivo [22]. This trend grew stronger when we compared those genes that showed self-linkage with fold-changes greater than 1.2 (approximately the 75th percentile of fold-change estimates) against genes without self-linkage. In this comparison, sites conserved in at least two additional species showed a 1.87-fold enrichment in polymorphisms (p = 0.021), while those with the strongest evidence of transcription factor binding showed a 3.73-fold enrichment (p = 0.00041). These results suggest that polymorphisms in conserved or bound sites tend to lead to larger changes in transcript abundance, but that polymorphisms in nonconserved sites may also contribute to variation in gene expression levels. In support of this finding, a functional role for nonconserved transcription factor binding sites was also shown by Doniger et al [23].
Table 1 Polymorphisms in Transcription Factor Binding Sites
Each of the middle four columns represents one category of bases in the region from 101 to 200 bp upstream of the start of translation; categories are defined by the self-linkage status of the gene and whether or not each base from 101 to 200 bp upstream belongs to a predicted transcription factor binding site. Each row represents one set of criteria for predicting such sites. The numerator in each cell gives the number of substitution SNPs that occur at bases in predicted transcription factor binding sites (“sites”) or at all other bases (“non-sites”) across all genes in the category. The denominator in each cell gives the total number of bases in predicted sites or non-sites across all genes in the category. Odds ratios and p-values from chi-squared tests were calculated on the number of SNPs in the four categories. Linkage: 1,233 genes with self-linkage versus 3,949 genes without self-linkage. Linkage and large effect: 330 genes with self-linkage that had a greater than 1.2-fold expression effect versus 3,949 genes without self-linkage. Conservation (in S. paradoxus, S. mikatae, and S. bayanus) and binding data used to predict transcription factor binding sites are from Harbison et al. [22].
If some sites that do not bind transcription factors contain other functional regulatory sequences, these sites may also show increased polymorphism in genes with self-linkage. Thus, our comparison of transcription factor binding sites to the surrounding sequence may provide a conservative estimate of the importance of variation in transcription factor binding sites relative to truly neutral sequence. Therefore, we also compared the rate of polymorphisms in transcription factor binding sites to the rate of synonymous polymorphisms in the coding sequence. This analysis showed a more pronounced enrichment of polymorphisms in transcription factor binding sites in genes showing self-linkage, and also showed a tendency for increased polymorphism at sites 101–200 bp upstream that are not predicted to fall within transcription factor binding sites (Table S3).
Discussion
We have found that nearly a quarter of yeast genes are affected by local regulatory variation between two strains. In our efforts to characterize fine-scale genetic variation that affects the expression level of nearby genes, we found that among a set of 77 genes showing strong self-linkage, most of these genetic changes act in cis, implying that they directly affect message abundance through changes in the rate of transcription or post-transcriptional regulation. A similar finding in the mouse [16] suggests that this observation is likely to be general. A high rate of cis-regulatory variation among all self-linking genes is supported by the observation of a highly significant increase in polymorphism in the promoter regions of genes with local regulatory variation, and a further increase in polymorphism in predicted transcription factor binding sites. The enrichment of polymorphisms in motifs discovered by cross-species conservation and binding site occupancy [22] suggests that variation in these sites is associated with changes in gene expression and, conversely, that discovery of regulatory polymorphisms may aid in the annotation of non-coding regulatory regions.
Although we found a global enrichment of polymorphisms in predicted transcription factor binding sites of genes with local regulatory variation, the effect was modest, and on a fine scale the pattern was complex. Even within the critical region 101–200 bp upstream thought to contain the highest density of transcription factor binding sites, genes with local regulatory variation showed an enrichment for polymorphisms at positions that are not predicted sites even by liberal criteria. Several explanations may account for this. First, we noticed that the rate of polymorphism between the two strains was non-uniform across the genome, with correlation in the level of divergence over both short and long distances. Thus, even in the analyses conditioned on local divergence, it is possible that some enrichment of polymorphism is due to an increased level of divergence rather than ascertainment based on functional significance. Second, it is likely that additional transcription factor binding sites in these genes exist, but that these sites have escaped detection because of their poor conservation across species, their low occupancy, or their noncanonical sequences. Finally, it is also possible that the increased level of polymorphism in the upstream region of genes with local regulatory variation may represent the signature of functionally important sites that participate in transcriptional control but are not directly involved in transcription factor binding. Indeed, Doniger et al. [23] argued that less than half of the functional constraint on Saccharomyces intergenic sequence could be attributed to predicted transcription factor binding sites based on known motifs, further suggesting that additional regulatory sequences remain to be discovered.
We also found increased polymorphism in the 3′ untranslated regions of genes with local regulatory variation that could not be explained by overlap with the promoter regions of adjacent genes, suggesting that sequence variants in this region can alter expression. Complex but non-random patterns of sequence conservation and composition have been observed in the 3′ untranslated regions of yeast genes, with significantly lower conservation in the 30 bases immediately downstream of translation stop and increased conservation in the subsequent 70 bases [24]. In addition, precision in the genome-wide control of mRNA half-life [25], which may involve the binding of Puf proteins to the 3′ UTR of mRNAs [26], further suggests that sequence signals besides those involved in transcriptional initiation play active roles in regulating transcript abundance.
A sizable minority of the genes with local regulatory variation that were assayed for ASE failed to show evidence of cis-acting variants. We also noted that genes showing self-linkage tended to have more polymorphisms in their coding sequences than genes not showing self-linkage. Although this increased polymorphism was not elevated above the baseline increase in divergence in genes showing self-linkage, more focused analyses of coding sequence polymorphisms may reveal changes in the gene product, which acts in trans to influence expression through autoregulation or feedback control, possibly indirectly through a pathway of mediators. One concrete example of such trans-acting local variation is the D368V amino acid substitution in Amn1. Another possible source of trans-acting variation in genes showing strong self-linkage is polymorphism in a different nearby gene that regulates the gene in question [16]. Our analyses suggest that this source can account for only a minority of self-linkages in our data.
Our results suggest that polymorphisms in the vicinity of a gene can affect its transcription level through a variety of mechanisms, with alteration of transcription factor binding sites being only one. Although this underscores challenges both in determining functional polymorphisms and in characterizing gene regulation in S. cerevisiae, unbiased identification of local regulatory variation through linkage analysis of expression levels will help to refine and validate currently proposed sets of regulatory motifs and will prompt exploration of novel classes of regulatory elements.
Materials and Methods
Linkage analysis and effect size estimates.
Linkage analysis and permutation tests were done as described [1] with genotypes and phenotypes from Brem and Kruglyak [14], except that only a single marker per transcript, the one closest to the gene's start site, was tested. A p-value of 0.012 corresponded to a false discovery rate less than 0.05 for the 5,727 transcripts tested. Thus, of the 1,428 transcripts showing significant self-linkage, 1,357 were expected to be true positives. The effect size of each self-linkage was computed as
, where
represents the mean expression level of the self-linking gene across all segregants bearing the BY allele at the marker closest to the gene, and
, represents the analogous mean taken with the RM allele. We obtained confidence intervals for the effect sizes by bootstrap resampling from the 112 segregants and taking the middle 95 of 100 fold-change estimates based on these resampled datasets.
The expression level of a transcript may show significant linkage to the location of its encoding gene for at least two reasons. The linking transcript level may vary because of a mutation in the coding sequence or regulatory region of its encoding gene. Alternatively, the linking transcript level may vary because of polymorphism in a neighboring regulatory gene acting in trans. We addressed the distinction between these scenarios in two ways. First, we performed nonparametric multipoint linkage analysis at 5-cM intervals using R/qtl [27] to define the LOD support interval for the highest peak on the respective chromosomes of each of the 1,428 transcripts showing self-linkage. For 1,313 and 1,380 of these transcripts, the encoding genes fell within the 1 LOD and 2 LOD support intervals, respectively. The observed data are in reasonable agreement with the theoretical expectations that the causative underlying polymorphism should be contained within the 1 LOD and 2 LOD support intervals approximately 90% and 99% of the time, respectively [28]. Next, we sought a direct estimate of the proportion of transcripts whose self-linkage was caused by a neighboring regulator. This proportion is related to the frequency of trans-acting regulators of transcript levels in the yeast genome. Therefore, we estimated the number of linkages across all transcripts that are due to trans-acting regulators. We selected a single marker for each transcript at random but not within 100 kb of the start site of the gene in question, and determined the number of such marker–transcript pairs that showed linkage at p < 0.012. We observed 160 significant linkages across all transcripts, 73 of which are expected to be false positives based on permutation tests. This suggests that testing a randomly chosen marker for each transcript across all transcripts is expected to yield 87 true positive linkages due to the presence of polymorphic trans-acting regulators near the markers. If such trans-acting regulators are distributed uniformly throughout the genome with respect to their targets, it follows that 87 of the 1,357 expected true positive self-linking transcripts in our data are due to polymorphisms in trans-acting regulators near the genes encoding the transcripts. In contrast, the remaining 94% of true positive self-linkages (1,270 of 1,357) are due to causative polymorphisms in the genes encoding the transcripts. Thus, 1,270 of 5,727 genes (22%) are expected to represent true self-linkages due to polymorphisms in the encoding genes. For the 77 genes with effect sizes greater than 1.2 chosen for direct tests of cis-acting variation by ASE, the significance level was more stringent (p < 10−8), such that no false positives were expected among the set of significant linkages. With these criteria (fold-change > 1.2 and p < 10−8), 224 self-linkages were identified with the single marker closest to the start site, while only three linkages were identified when the single marker was chosen at random as above. Thus, we expect 1.3% of self-linkages identified at this significance level (three out of 224), or one of the 77 tested, to be due to polymorphisms in nearby trans-acting regulators distinct from the gene in question.
The above analysis assumes that trans-acting regulators are distributed uniformly throughout the genome with respect to their targets. In the yeast genome, there is little evidence for strong deviation from this model. Approximately 19% of adjacent gene pairs show common function as opposed to 14% of random pairs [29], and less than 10% of adjacent pairs show correlated gene expression [30]. Thus, our estimate for the prevalence of trans-acting regulators falling near genes with self-linking transcript levels is likely to be reasonably accurate, and such neighboring factors are unlikely to be responsible for the majority of self-linkages we studied here.
ASE measurements.
We used the TaqMan (Applied Biosystems, Foster City, California, United States) genotyping system in real-time quantitative PCR experiments to assay for the presence of ASE in diploid hybrids of BY and RM. Primer and probe sequences for the TaqMan assays are available on request. Genomic DNA and mRNA were extracted from four independent diploid cultures. In addition, we made two technical replicates each of 2:1, 1.5:1, 1.2:1, 1:1.2, 1:1.5, and 1:2 mixtures of each allele from the same extraction of BY and RM genomic DNA. This standard curve allowed us to identify outliers and assess the performance of each assay and to estimate the fold-change. It also suggested that we could reliably detect a 1.2-fold difference in the expression of each allele. For each sample, the logarithm of the fold-change in the amount of each allele present was estimated by the difference in cycle number at which the FAM and VIC dye intensities crossed the threshold intensity. We tested for the presence of ASE using a t-test to compare the diploid cDNA to the diploid genomic DNA and used the t statistic to form confidence intervals for the fold-change. In order to avoid underestimating the extent of ASE in these 77 genes because of the possibility of low statistical power in our four-sample versus four-sample t-test, we estimated 1 − π0, the rate of true alternative hypotheses, by the method of Storey and Tibshirani [15]. The estimate of 1 − π0 converged to a stable value (0.78) for maximum values of the tuning parameter ranging from 0.4 to 0.75. We observed 44 significant tests out of the expected 60 truly alternative tests estimated by 1 − π0, suggesting that our experiments had reasonable power (approximately 70%) to detect ASE at a p-value of less than 0.05.
Effects of Amn1 polymorphism.
Linkage results for DSE1/YER124C and DSE2/YHR143W were from Brem and Kruglyak [14], and effect sizes of AMN1 polymorphism were calculated as above. To test the effect of variation at amino acid 368, the D368 variant of AMN1 was engineered into the S288c derivative JW2 (MATα, ura3Δ0, clonNAT+ downstream of GPA1) by the two-step gene replacement method [31]. Briefly, an integrating URA3 plasmid carrying the D368 allele was introduced at the endogenous AMN1 locus, resulting in a duplication of AMN1. Selection with 5FOA resulted in colonies that had lost the URA3 marker and one of the copies of AMN1. The allele was determined by sequencing. Expression arrays were performed as by Yvert et al. [13], except that the reference sample was a 1:1 mixture of RNA from the BY and RM strains. Transcriptional effects given in the text represent the ratio between the wild-type control and allele-replaced strains.
Sequence analysis.
The BY sequence was obtained from the Saccharomyces Genome Database (http://www.yeastgenome.org). The RM and YJM789 (version 2) whole-genome assemblies were obtained from the Broad Institute (http://www.broad.mit.edu/annotation/fungi/fgi/) and the Stanford Genome Technology Center (http://med.stanford.edu/sgtc/research/yjm789.html), respectively. We used CROSSMATCH (http://bozeman.mbt.washington.edu/phrap.docs/phrap.html) to identify a stringent set of ORFs in RM and in BY that were reciprocal best matches with at least 98% identity. We then aligned these sequences using CLUSTALW [32] and annotated the coding sequence and predicted transcription factor binding sites using data from the Saccharomyces Genome Database and supplementary data from Harbison et al. [22] (http://jura.wi.mit.edu/fraenkel/download/). Of the 5,727 ORFs in the linkage analyses, 5,182 met the criteria for high-confidence sequence alignment, 1,233 of which showed self-linkage.
Statistical analyses.
All statistical analyses were performed using R [33]. Logistic regression analyses were performed using the number of substitution SNPs in each 100-bp bin as independent variables and the self-linkage status of each gene (zero if the gene did not show self-linkage, one if the gene showed self-linkage) as the dependent variable. The coding sequence was treated as a single bin. We set a ceiling of ten substitution SNPs per 100-bp bin to control for outliers due to low-quality subregions of alignments. For each bin, this ceiling affected no more than five of 2,250 genes total. The p-value for each bin was obtained separately by performing logistic regression to estimate the predictive value of a model containing that bin only. The p-value for each bin in the conditional logistic regression was obtained by estimating the additional predictive value of a model including all 21 bins relative to a null model containing the remaining 20 bins.
Corrections for overlapping intergenic regions.
Because intergenic regulatory regions overlap between neighboring genes in the compact S. cerevisiae genome, we tested whether the apparently increased rate of polymorphism in the 3′ untranslated region could be explained by polymorphisms located in the promoter region of neighboring genes with self-linkage. We used two related approaches, one based on simulation and the other based on an expectation maximization (EM) algorithm, to estimate the rate of polymorphisms attributable to each bin and the rate attributable to overlapping bins of adjacent genes, conditional on the self-linkage status of the gene and its neighbors.
For the first correction method, we simulated the expected number of SNPs in each bin according to the specified underlying divergence for that bin and according to the divergence of overlapping bins as follows:
where
is the length of the quantity in parentheses and d is the divergence (per-base probability of substitution polymorphism) for that bin, given that it belongs to a gene showing self-linkage or a gene not showing self-linkage. We then minimized the difference between the expected number of SNPs per bin and the observed numbers of SNPs in each bin in the actual data over the parameters dij, where
We chose initial values for dij such that all bins had the same divergence both in genes showing self-linkage and genes not showing self-linkage, but the values of dij that minimized the difference between the expected and observed data were similar over several different sets of initial values.
For the second correction method, we used an approach based on EM. For each SNP in the observed data, we assigned a fractional count to each bin into which the SNP fell as follows:
where d is the divergence for the bin. Initially, the dij values were taken to be the observed divergence in each bin for genes showing self-linkage and genes not showing self-linkage. After each iteration, the dij values were updated by calculating new values based on the fractional counts for each SNP. The choice of starting values had negligible impact on the final dij obtained at convergence. Note that this approach differs somewhat from the simulation-based method in that, rather than specifying an underlying pattern of divergence and determining what distribution of SNPs it would produce, the EM approach takes the observed distribution of SNPs and estimates what underlying pattern of divergence would be most likely to produce it.
Both of the above procedures suggested that the spacing between adjacent genes is variable enough that no specific, artifactual enrichment of polymorphisms is produced in any single bin (Figure S3). As a final test, we repeated the logistic regression analysis across four separate subsets of the 2,250 genes: only those 1,057 genes whose start was more than 1,000 bases from the nearest start of any other gene, 1,065 genes whose start was more than 1,000 bases from the nearest stop, 925 genes whose stop was more than 1,000 bases from the nearest start, and 981 genes whose stop was more than 1,000 bases from the nearest stop of another gene. In spite of the reductions in power incurred by discarding much of the data for these analyses, the region from 101 to 200 bp upstream of translation start and the region from 1 to 100 bp downstream of translation stop continued to show significantly increased divergence in all analyses (Figure S3).
Supporting information
Figure S1 Comparison of Linkage and ASE Fold-Change Estimates
Points represent the fold-change estimates from linkage analysis (horizontal axis) and from ASE experiments (vertical axis). Horizontal and vertical lines on each point give the 95% confidence intervals. The solid diagonal line (y = x) represents equal fold-change estimates in the two experiments.
(A) Strong self-linkage is shown in 33 genes with ASE p > 0.05. The dashed line gives the line of best fit (y = 0.48x − 0.036; p-value for slope = 10−8). Note that the slope and the estimated π0 from the method of Storey and Tibshirani [15] suggest that a sizeable fraction of these 33 genes show ASE.
(B) Self-linkage p > 0.012 in 16 genes. The confidence intervals for several of the genes tested failed to overlap zero, suggesting that these genes may show weak self-linkage that did not meet our experiment-wide criterion of p < 0.012. The dashed line indicates the best fit (y = 0.33x − 0.008, p-value for slope = 0.37).
(16 KB EPS)
Click here for additional data file.
Figure S2 Increased Divergence across Extended Regions in Genes Showing Self-Linkage
The y-axis represents −log10(p) from logistic regression of self-linkage status on SNP rate, for each 100-bp bin at the distance from translation start (or from translation stop for downstream bins) indicated on the x-axis. The coding sequence is treated as a single bin. Open circles connected by dashed lines: analysis of each bin separately, across all genes. Filled circles connected by solid lines: analysis of each bin conditional on SNP rates in all other bins, across all genes. Open triangles connected by dotted lines: analysis of each bin separately between 1,125 position-matched pairs of genes with and without self-linkage (see Figure 4). The dashed horizontal line indicates p = 0.0024 (p = 0.05 after a Bonferroni correction for the 21 bins tested).
(8 KB EPS)
Click here for additional data file.
Figure S3 Enrichment in Genes with Self-Linkage for Non-Coding Polymorphism Is Not an Artifact of Overlap between Intergenic Regions
The y-axis indicates the difference in inter-strain divergence (substitutions per basepair) between 1,125 position-matched pairs of genes with and without self-linkage. Each point represents the divergence averaged over a 100-bp bin centered at the distance from translation start (or from translation stop for downstream bins) indicated on the x-axis. The coding sequence is treated as a single bin. Solid circles: difference in divergence estimated directly from observed data (see Figure 4). Triangles: difference in divergence corrected for overlap by the simulation-based approach (see Materials and Methods). Plus symbols: difference in divergence corrected for overlap by the EM-based approach (see Materials and Methods). Shading indicates regions with significantly increased divergence (p < 0.05) in genes showing self-linkage, across genes spaced at least 1,000 bp from one another.
(6 KB EPS)
Click here for additional data file.
Table S1 Linkage Analysis Results
(235 KB TXT)
Click here for additional data file.
Table S2 ASE Results
(7 KB TXT)
Click here for additional data file.
Table S3 Comparison of Promoter Polymorphisms with Synonymous Polymorphisms
The first column lists selection criteria for transcription factor binding sites predicted by Harbison et al. [22]. The next two columns list the numbers of substitution polymorphisms in transcription factor binding sites (numbers of substitution polymorphisms in all other bases not belonging to predicted sites, labeled as non-sites, are in parentheses) in the region from 101 to 200 bp upstream of translation start in genes showing self-linkage and genes not showing self-linkage, respectively. The next two columns show the numbers of substitution polymorphisms in synonymous sites in the coding sequences of genes showing self-linkage and genes not showing self-linkage, respectively. Boundaries of the coding sequences were determined by Saccharomyces Genome Database annotations. If a gap or premature stop codon was encountered in the RM sequence, all subsequent codons were ignored. Each row in the table represents one set of criteria for transcription factor binding site prediction (as in Table 1). In the last column, p-values are from a chi-squared test comparing the number of substitutions in “sites” and “non-sites” in genes with or without self-linkage to the number of synonymous substitutions in genes with or without self-linkage. Linkage: 1,233 genes with self-linkage versus 3,949 genes without self-linkage. Linkage and large effect: 330 genes with self-linkage that had a greater than 1.2-fold expression effect versus 3,949 genes without self-linkage.
(43 KB DOC)
Click here for additional data file.
The authors wish to thank E. Smith, D. Spencer, and J. Akey for helpful discussions. J. Akey contributed advice on analyses and made critical readings of the manuscript. We also thank E. Foss for plasmids. The research was supported in part by the Howard Hughes Medical Institute and National Institutes of Mental Health grant R37 MH59520 to LK. LK is a James S. McDonnell Centennial Fellow. JR is supported by the University of Washington Medical Scientist Training Program. RBB is supported by a Burroughs-Wellcome Career Award at the Scientific Interface.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. JR, RBB, and LK conceived and designed the experiments. JR and JW performed the experiments. JR, RBB, and LK analyzed the data and wrote the paper.
Abbreviations
ASEallele-specific expression
BYBY4716
EMexpectation maximization
LODlogarithm of odds
RMRM11-1a
SNPsingle nucleotide polymorphism
==== Refs
References
Brem RB Yvert G Clinton R Kruglyak L 2002 Genetic dissection of transcriptional regulation in budding yeast Science 296 752 755 11923494
Schadt EE Monks SA Drake TA Lusis AJ Che N 2003 Genetics of gene expression surveyed in maize, mouse and man Nature 422 297 302 12646919
Morley M Molony CM Weber TM Devlin JL Ewens KG 2004 Genetic analysis of genome-wide variation in human gene expression Nature 430 743 747 15269782
Monks SA Leonardson A Zhu H Cundiff P Pietrusiak P 2004 Genetic inheritance of gene expression in human cell lines Am J Hum Genet 75 1094 1105 15514893
Bystrykh L Weersing E Dontje B Sutton S Pletcher MT 2005 Uncovering regulatory pathways that affect hematopoietic stem cell function using ‘genetical genomics' Nat Genet 37 225 232 15711547
Chesler EJ Lu L Shou W Qu Y Gu J 2005 Complex trait analysis of gene expression uncovers polygenic and pleiotropic networks that modulate nervous system function Nat Genet 37 233 242 15711545
Hubner N Wallace CA Zimdahl H Petretto E Schulz H 2005 Integrated transcriptional profiling and linkage analysis for identification of genes underlying disease Nat Genet 37 243 253 15711544
Cowles CR Hirschhorn JN Altshuler D Lander ES 2002 Detection of regulatory variation in mouse genes Nat Genet 32 432 437 12410233
Lo HS Wang Z Hu Y Yang HH Gere S 2003 Allelic variation in gene expression is common in the human genome Genome Res 13 1855 1862 12902379
Knight JC 2004 Allele-specific gene expression uncovered Trends Genet 20 113 116 15049300
Wittkopp PJ Haerum BK Clark AG 2004 Evolutionary changes in cis and trans gene regulation Nature 430 85 88 15229602
Ronald J Akey JM Whittle J Smith EN Yvert G 2005 Simultaneous genotyping, gene-expression measurement, and detection of allele-specific expression with oligonucleotide arrays Genome Res 15 284 291 15687292
Yvert G Brem RB Whittle J Akey JM Foss E 2003
Trans -acting regulatory variation in Saccharomyces cerevisiae and the role of transcription factors Nat Genet 35 57 64
Brem RB Kruglyak L 2005 The landscape of genetic complexity across 5,700 gene expression traits in yeast Proc Nat Acad Sci U S A 102 1572 1577
Storey JD Tibshirani R 2003 Statistical significance for genome-wide studies Proc Natl Acad Sci U S A 100 9440 9445 12883005
Doss S Schadt EE Drake TA Lusis AJ 2005
Cis -acting expression quantitative trait loci in mice Genome Res 15 681 691 15837804
Wang Y Shirogane T Liu D Harper JW Elledge SJ 2003 Exit from exit: Resetting the cell cycle through Amn1 inhibition of G protein signaling Cell 112 697 709 12628189
Gu Z David L Petrov D Jones T Davis RW 2005 Elevated evolutionary rates in the laboratory strain of Saccharomyces cerevisiae
Proc Natl Acad Sci U S A 102 1092 1097 15647350
Winzeler EA Castillo-Davis CI Oshiro G Liang D 2003 Genetic diversity in yeast assessed with whole-genome oligonucleotide arrays Genetics 163 79 89 12586698
Chin CS Chuang JH Li H 2005 Genome-wide regulatory complexity in yeast promoters: Separation of functionally conserved and neutral sequence Genome Res 15 205 213 15653830
Cliften P Sudarsanam P Desikan A Fulton L Fulton B 2003 Finding functional features in Saccharomyces genomes by phylogenetic footprinting Science 301 71 76 12775844
Harbison CT Gordon DB Lee TI Rinaldi NJ Macisaac KD 2004 Transcriptional regulatory code of a eukaryotic genome Nature 431 99 104 15343339
Doniger SW Huh J Fay JC 2005 Identification of functional transcription factor binding sites using closely related Saccharomyces species Genome Res 15 701 709 15837806
Shabalina SA Ogurtsov AY Rogozin IB Koonin EV Lipman DJ 2004 Comparative analysis of orthologous eukaryotic mRNAs: Potential hidden functional signals Nucleic Acids Res 32 1774 1782 15031317
Wang Y Liu CL Storey JD Tibshirani RJ Herschlag D 2002 Precision and functional specificity in mRNA decay Proc Nat Acad Sci U S A 99 5860 5865
Gerber AP Herschlag D Brown PO 2004 Extensive association of functionally and cytotopically related mRNAs with Puf family RNA-binding proteins in yeast PLoS Biol 2 e79. DOI: 10.1371/journal.pbio.0020079 15024427
Broman KW Wu H Sen S Churchill GA 2003 R/qtl: QTL mapping in experimental crosses Bioinformatics 19 889 890 12724300
Kruglyak L Lander ES 1995 High-resolution genetic mapping of complex traits Am J Hum Genet 56 1212 1223 7726179
Cohen BA Mitra RD Hughes JD Church GM 2000 A computational analysis of whole-genome expression data reveals chromosomal domains of gene expression Nat Genet 26 183 186 11017073
Kruglyak S Tang H 2000 Regulation of adjacent yeast genes Trends Genet 16 109 111 10689350
Scherer S Davis RW 1979 Replacement of chromosome segments with altered DNA sequences constructed in vitro Proc Natl Acad Sci U S A 76 4951 4955 388424
Thompson JD Higgins DG Gibson TJ 1994 CLUSTAL W: Improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice Nucleic Acids Res 22 4673 4680 7984417
R Development Core Team 2004 R: A language and environment for statistical computing Vienna (Austria) R Foundation for Statistical Computing
|
16121257
|
PMC1189075
|
CC BY
|
2021-01-05 08:00:26
|
no
|
PLoS Genet. 2005 Aug 19; 1(2):e25
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010025
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1612125810.1371/journal.pgen.001002605-PLGE-RA-0104R2plge-01-02-06Research ArticlePlant ScienceStatisticsSystems BiologyPlantsArabidopsisA Global Survey of Gene Regulation during Cold Acclimation in Arabidopsis
thaliana
Cold regulated genes in Arabidopsis
Hannah Matthew A 1*Heyer Arnd G 2Hincha Dirk K 1Kim Stuart Editor1 Max-Planck-Institut für Molekulare Pflanzenphysiologie, Potsdam, Germany
2 Biologisches Institut, Abteilung Botanik, Universität Stuttgart, Stuttgart, Germany
Stanford University School of Medicine, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 19 8 2005 1 2 e2617 5 2005 8 7 2005 Copyright: © 2005 Hannah et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.Many temperate plant species such as Arabidopsis thaliana are able to increase their freezing tolerance when exposed to low, nonfreezing temperatures in a process called cold acclimation. This process is accompanied by complex changes in gene expression. Previous studies have investigated these changes but have mainly focused on individual or small groups of genes. We present a comprehensive statistical analysis of the genome-wide changes of gene expression in response to 14 d of cold acclimation in Arabidopsis, and provide a large-scale validation of these data by comparing datasets obtained for the Affymetrix ATH1 Genechip and MWG 50-mer oligonucleotide whole-genome microarrays. We combine these datasets with existing published and publicly available data investigating Arabidopsis gene expression in response to low temperature. All data are integrated into a database detailing the cold responsiveness of 22,043 genes as a function of time of exposure at low temperature. We concentrate our functional analysis on global changes marking relevant pathways or functional groups of genes. These analyses provide a statistical basis for many previously reported changes, identify so far unreported changes, and show which processes predominate during different times of cold acclimation. This approach offers the fullest characterization of global changes in gene expression in response to low temperature available to date.
Synopsis
Freezing tolerance is an important determinant of geographical distribution of plant species, and freezing damage in crop plants leads to severe losses in agriculture. Many temperate plants increase their freezing tolerance during exposure to low, but nonfreezing temperatures, a process known as cold acclimation. Freezing tolerance and cold acclimation are complex, quantitative genetic traits. The number and functional roles of the responsible genes are not known for any plant species. Using the model plant Arabidopsis thaliana, which is moderately freezing tolerant and able to cold acclimate, the global regulation of gene expression during exposure to 4 °C for 14 d was analyzed by microarray hybridization. For validation of gene expression data, triplicate biological samples were hybridized to two different oligonucleotide arrays. Results from the two platforms showed good agreement, indicating the reliability of the measurements. The authors combined their data with all publicly available data on cold-regulated gene expression in A. thaliana to compile a database detailing the cold responsiveness of 22,043 genes as a function of exposure time. In addition, thorough statistical analysis was used to identify metabolic pathways and physiological processes that are predominantly involved in the plant cold-acclimation process.
Citation:Hannah MA, Heyer AG, Hincha DK (2005) A global survey of gene regulation during cold acclimation in Arabidopsis thaliana. PLoS Genet 1(2): e26.
==== Body
Introduction
Cold has major influences on crop production, limiting geographical distribution and growing season and affecting quality and yield. Considerable effort has therefore been directed toward understanding how plants respond and adapt to low temperature. Arabidopsis thaliana, like many plants, increases its freezing tolerance when exposed to low nonfreezing temperatures (reviewed in [1]). This process of cold acclimation is a multigenic and quantitative trait that is associated with complex physiological and biochemical changes. These changes are extensive and affect growth and water balance, the accumulation of compatible solutes, membrane and cell wall composition, antioxidant production (increased), and cold-regulated (COR) gene expression and protein levels [1–3]. Traditional approaches have identified around 200 cold-responsive genes, but more recently this list has been extended by several hundred using expression profiling technologies [4–9]. Many of the expression changes can be related to the well-documented biochemical changes listed above, while others provide new information. The characterization of genes that respond to cold in Arabidopsis is important to understand the response of plants to low temperature and the processes involved in plant cold acclimation. Such information will help in the development of strategies for the improvement of freezing tolerance in crop plants.
The regulation of cold-responsive gene expression has received much attention and has been recently reviewed [3,10]. The C-repeat binding factors (CBFs)-1, -2, and -3 (also known as dehydration-responsive element binding1 (DREB1)-b, -c, and -a, respectively), have been a focus of research, including the identification of target genes involved in the cold response [11,12]. The overexpression of the CBF genes has been shown to have large effects on the cold-responsive transcriptome and metabolome of Arabidopsis, and their activities may be functionally redundant [6,11,13]. However, analysis of a mutant in which the CBF2 gene was disrupted revealed that this gene negatively regulates CBF1 and CBF3 expression [14]. Other transcription factors have been shown to act as positive [4] or negative [15] regulators of, or in addition to [16], the CBF pathway. Recently, ZAT12 was shown to down-regulate the expression of the CBF genes and to have a cold-responsive regulon that partially overlapped with CBF2 [9]. Clearly, the transcriptional regulation of cold-responsive gene expression is complex. The identification of cold-responsive genes that are potentially under the control of these transcription factors will be useful to decipher their function and relative importance.
Despite the wealth of published data and the increasing availability of public datasets (e.g., NASCArrays, AtGenExpress), there is currently no consensus on the number and identity of cold-responsive genes in Arabidopsis; reports and estimates vary from less than 100 to about 1,000 [5–8]. This variation is related to the diversity of growth conditions and experimental treatments of the plants and of the profiling technologies used. In addition, the development of standards for microarray experiments and data analysis has left many of the initial studies lacking sufficient replication or a thorough statistical evaluation of the data. Moreover, in common with many other treatments that have large effects on gene expression, there has been little discussion of the assumptions underlying normalization for microarrays and the confirmation of observed changes. Finally, although results from expression profiling have been discussed in relation to previously reported cold-responsive genes, there has been no large-scale validation of the observed changes using an independent platform. The discussions of these results have focused mainly on individual or small groups of genes, and functional classifications have not been related to, or statistically tested against, those of all genes.
Here, we present replicated statistical comparisons from independent microarray platforms of the changes in gene expression after 14 d of cold acclimation. A comparison of these data with published and publicly available datasets allowed us to compile consensus lists of the most consistent changes for short- (12 h or less), medium- (24–48 h), and long-term (more than 48 h) cold-responsive genes in Arabidopsis. Using improved annotation, visualization, and statistical testing, we focused on global changes and patterns to provide new insights into which processes are most important during exposure to low temperature.
Results
Data Quality Assessment
As a necessary prerequisite for data interpretation, the first priority should be the assessment of data quality. This can be separated into two basic considerations: the technical quality of the arrays, and the validity of the expression estimates. The former approach is straightforward, with all our arrays passing both the manufacturers guidelines and additional data quality assessment using the bioconductor (http://www.bioconductor.org/) and R software. The second approach is often addressed by confirming a small number of candidate genes using RNA blots or quantitative or semi-quantitative RT-PCR with one or two housekeeping genes as controls. This is not practicable when confirmation of many changes is required. Therefore, we used two independent microarray platforms: the Affymetrix ATH1 genome array (subsequently referred to as ATH1) [17] and the MWG Biotech 25k Arabidopsis 50-mer oligonucleotide array (MWG).
The possibility of a global change in cellular mRNA abundance needs to be considered, as this may confound the interpretation of data from expression profiling due to bias introduced during normalization (see Discussion). To our knowledge, this has not been discussed in relation to plant studies. For both platforms, we compared signal intensities for a large number of common housekeeping genes across treatments. Additionally, we compared our results to changes reported using conventional non-microarray approaches. The data for housekeeping genes show good linear correlations (Figure 1). The mean absolute log2 fold change in gene expression during cold acclimation for the housekeeping genes on the ATH1 and MWG arrays are 0.35 and 0.41, respectively. These are lower than the 0.39 and 0.62 obtained using permutations of the same number of randomly selected genes of similar intensities. We compared our data to previous reports of genes responding to low temperatures in the long term (more than 48 h) that did not use profiling technologies. For genes mentioned in these reports and present on either of the arrays, signal changes matched for 15/16 and 15/17 down-regulated genes, and 23/29 and 26/29 up-regulated genes for the ATH1 and MWG arrays, respectively. These signal changes are significant in at least one dataset for 13/18 down-regulated and 16/29 up-regulated genes.
Figure 1 Comparison of Housekeeping Gene log2 Signal between Nonacclimated and 14 d Cold Acclimated Arabidopsis
Data for the ATH1 (A) and MWG (B) arrays are shown. Genes annotated as actins (ACT), elongation factors (EF), DNA-dependent RNA polymerases (RP), tubulins (TUB), poly-ubiquitins (UBQ), and cytosolic cyclophilins (CYC) were used. Genes used are listed in Table S7.
Comparison of ATH1 and MWG Arrays
Recently, there has been considerable interest in cross-platform reproducibility of profiling experiments [18]. This has mainly focused on animal studies, with only one plant study in which multiple array technologies were used to confirm expression changes [12]. Arrays can be compared on the basis of absolute expression or the change between treatments. Using data for all 20,159 AGI locus codes that are represented on both arrays, there is a correlation of 0.7 for both nonacclimated log2 absolute expression and log2 change during cold acclimation. However, this gives a distorted view of the reproducibility of these parameters due to differences in the distribution of their values. Gene expression changes during cold acclimation are clustered around zero but extend as a linear distribution, while the log2 nonacclimated expression comparison is linear only at higher intensities (Figure 2). At low intensities, the MWG expression estimates are more compressed than those from the ATH1 array.
Figure 2 Comparison of log2 Signal and Signal Changes between the ATH1 and MWG Arrays
Shown here is a comparison of log2 signal for nonacclimated plants (A) and log2 signal change during cold acclimation (B), between the ATH1 and MWG arrays. Data are for the 20,159 genes that are represented by probes on both the ATH1 and the MWG arrays. Genes are colored on the basis of their detection as significantly (FDR p < 0.05) differentially expressed in response to 14 d cold acclimation using the two-array datasets. Black, both datasets; green, MWG; red, ATH1; yellow, neither.
It is now becoming accepted that p-values from statistical testing are better criteria for defining differential expression than arbitrary fold-change cutoffs. However, the validity of p-values arising from statistical tests of differential expression has been the subject of debate, and it is necessary to discuss this issue before interpreting results based on these values. The distribution of unadjusted p-values resulting from a test should show an expected distribution. This distribution can be expressed as a mixed model of the p-values arising from the null hypothesis, which are distributed uniformly on the interval (0,1), and the p-values of the alternative hypothesis, which have a higher density of small p-values [19]. The p-values from the test for differential expression using the MWG data fit this model perfectly, and those from the ATH1 data fit very closely (very slight overabundance of high p-values). This indicates that the chosen test is valid and that a model can be used to estimate true and false negatives and positives [19]. We used a linear step-up false discovery rate (FDR) correction of p-values for multiple testing [20], which has been shown to provide conservative FDR control for microarray data [21]. An FDR-corrected p-value of less than 0.05 means that less than 5% of the total number of significant values (genes) below this threshold are false positives. This differs from the uncorrected p-value, which gives the proportion of all tests (total genes on the array) that falsely declare a significant difference.
The MWG data analysis identifies more genes that are differentially expressed in response to 14 d of cold treatment than the ATH1 data analysis at any given significance threshold. This effect is greater than can be explained by the larger size of the MWG array. At the FDR p-value of less than 0.05, approximately 900 more genes are significantly regulated according to the MWG data than the ATH1 data. This is most likely related to differences in the statistical analysis of the data necessitated by the technical replication of the MWG array hybridizations. Among genes present on both arrays, the ATH1 data reveal 2,165 that change significantly at the FDR threshold of 0.05, of which 58% are also significant at the same threshold using the MWG data (Figure 3). Of the 1,263 genes that have an FDR p-value of less than 0.05 for both arrays, only seven have opposite ratios (see Figure 2B). As there are high false-negative rates associated with these p-value thresholds (see below), it is not sufficient only to compare numbers of significantly changed gene expression levels at a given threshold. The overall consistency of the data is evident from the fact that only 6.5% of the 2,643 genes showing a significant change in expression on the ATH1 arrays show an opposite direction of change using the MWG arrays (see Figure 2B). Of these 141 genes, the majority (64%) have FDR corrected p-values greater than 0.5 in the MWG data, indicating that these ratios are not significant. In contrast, the FDR p-values of the remaining genes, which agree in direction between the two platforms, are heavily biased toward the MWG data also giving a low FDR p-value.
Figure 3 Comparison of Significant Expression Changes between the ATH1 and MWG Arrays
Venn diagrams showing the overlap between genes detected as significantly (FDR p < 0.05) up- or down-regulated using data from the ATH1 and the MWG arrays. Only data for the 20,159 genes that are represented by probes on both arrays are shown.
Comparison with Previous Studies
We have already shown the consistency of our data in comparison to previous studies using nonprofiling techniques. As we have compiled a database containing the expression changes for cold-responsive genes from other expression profiling experiments, we can easily compare between studies. Studies that had multiple observations for a single time class were averaged. As different data analysis methods were used, and for simplicity, we express the similarity between two studies as the proportion of the common differentially expressed genes that agree in their direction of change. Comparisons between all studies indicate that the main effects leading to higher agreement are the duration of the cold exposure, the array type, and the replication/data analysis (Table S1). ATH1 data analyzed using the Microarray Suite 5 (MAS5) software form the most obvious group. Datasets that have only one biological and/or technical replication and lower cutoffs have lower agreement. The classification of the studies according to time classes of short, medium, and long term is supported by the comparisons between the studies. The minimum agreement between our data and other long-term studies was 0.83 and 0.76 for the ATH1 and MWG datasets, respectively, obtained in comparison to Vogel-P 7d (Table S1). There was lower agreement with the medium-term datasets, and agreement was lowest in comparison to the short-term datasets.
Cold-Responsive Gene Expression
There have been previous estimates of how many genes respond to low temperature, but these have tended to be conservative. In the long term, with an FDR p-value threshold of less than 0.05, we detect 2,297 and 3,379 cold-responsive genes using the ATH1 and MWG arrays, respectively. These are underestimates due to the high false-negative rates. If we model the p-value distributions [19] using either of the datasets, we find the true positive rate is likely to be in excess of 10,000 genes. In the short and medium term, there are no suitable datasets to estimate the true number of cold-responsive genes. However, we made comparisons using available data to obtain an approximation of the relative numbers of cold-responsive genes. These comparisons indicate that at 24 h there are either more genes changing or that the magnitude of changes is greater in comparison to the long term. Of these changes visible at 24 h, it is likely that 20%–30% are seen within 3 h of cold treatment, and this could have increased to 75% after 12 h.
Obviously, some of these changes depend on the treatment and technology used, with the changes of most interest being those that are consistent across independent studies or technologies. We have used the available data to generate lists of the genes most consistently responding to low temperature in the short, medium, or long term (Table S2). Genes mentioned in these lists have been reported in at least two biologically and/or technically independent experiments. Specifically, for each independent study within a time class a gene was given a score of +1 for up-regulation and −1 for down-regulation. Those genes scoring 2 or higher or −2 or below were defined as cold responsive. These criteria allowed us to define 808, 1,224, and 672 genes as consistently up-regulated, and 240, 1,364, and 915 genes as down-regulated in response to cold in the short, medium, and long term, respectively (Figure 4). Of the up-regulated genes, 114 are common to all lists, while the number of common down-regulated genes is 42. Of the short-term cold responsive genes, about 55% and 17% are similarly changed in the medium and long term, respectively. Approximately 30% of the long-term cold-responsive genes are similarly changed in the medium term.
Figure 4 Comparison between Short-, Medium-, and Long-Term Cold Responsive Genes
Venn diagrams showing the overlap between genes classified as being up- or down-regulated in response to short-, medium-, or long-term cold treatments. Full details are given in Materials and Methods and Table S2.
Annotation-Based Analysis Reveals Key Changes for Cold-Responsive Gene Expression
It is clearly not possible to describe all changes reported for cold-responsive gene expression, and the reader is encouraged to explore the data using the Mapman plots [22,23] (S1–11, M1–11, and L1–11 in Dataset S1), the analysis of Mapman functional groups (Table 1), and that of The Arabidopsis Information Resource (TAIR) AraCyc pathways (Table S3). Our analysis did not focus on individual genes, as these can be identified directly (S1–11, M1–11, and L1–11 in Dataset S1). We also did not directly compare the number of up- or down-regulated genes, as the differences in the numbers of regulated genes (Figure 4) would make some of the analysis trivial. Instead, we took a global view, where distributions of cold-responsive genes among different functional groups were tested for significant deviation from that expected for all genes within the database. Mapman functional groups and AraCyc pathways that were significantly under- or overrepresented in this analysis were identified using Fisher exact tests. Using the Mapman functional groups, unknown genes were significantly underrepresented among many of the gene lists (Table 1). The statistical analysis was repeated with these genes excluded, as their underrepresentation could increase the proportion of genes in functionally defined groups without pertaining to biological effects. The highlighting of p-values in Table 1 is based on their meeting a given cutoff in both analyses.
Table 1 Statistical Analysis Showing Over- and Under-Representation of Mapman Functional Groups
+/−, fold change of the proportion of cold-responsive genes in a group versus the proportion of all genes cold responsive; p-Value, p-value from Fisher exact test; p < 0.05, p < 0.01, and p < 0.001 are highlighted in purple, green, and red, respectively.
The most affected group is that of stress, which is significantly (p < 0.001) overrepresented among cold up-regulated genes at all times (Table 1). Stress related genes up-regulated include those known to respond to temperature, drought, and salt (S1, M1, and L1 in Dataset S1). Other groups that have significantly more up-regulated genes than expected at most times are polyamine and secondary metabolism. The AraCyc pathway analysis reveals that flavonoid, lignin, phytoalexin, and suberin biosynthesis, and the initial phenylpropanoid reactions, which share many common genes, are overrepresented in the short to medium term (Table S3). Mapman visualization indicates that the up-regulated genes of secondary metabolism are involved mainly in the metabolism of dihydroflavonols, chalcones, and anthocyanins (S2, M2, and L2 in Dataset S1). The hormone functional group is the only one with significantly more down-regulated genes than expected at most times. This is mainly due to the down-regulation of auxin-induced or -responsive genes, although genes from the jasmonic acid and ethylene groups are also abundant (S3, M3, and L3 in Dataset S1). In the long term, signaling/response to gibberellin also has many down-regulated members. The only hormone biosynthetic pathway significantly down-regulated is brassinosteroid biosynthesis in the short and long term (Table S3A).
Short-term–specific responses are the significant overrepresentation of transport among up-regulated genes, and of nucleotide metabolism and redox regulation among down-regulated genes. The AraCyc pathway analysis also identifies purine biosynthesis from nucleotide metabolism as significantly overrepresented among down-regulated genes (Table S3A). In addition, the cell wall group is significantly overrepresented among down-regulated genes in the short to medium term (Table 1). Up-regulated genes involved in transport are thought to have roles as ATP binding cassette, sugar, and phosphate transporters and ATPases (S4 in Dataset S1). Changes in the transport group are generally similar in the medium to long term, although sugar transporters become predominantly down-regulated (M4 and L4 in Dataset S1).
In the short term, it may be expected that many of the observed changes are involved in the signaling and regulation of the responses to low temperature. However, the proportion of these genes up- or down-regulated is not significantly different from that expected (Table 1). Genes that do change are those involved in calcium, G-protein, and light signaling and those annotated as receptor or mitogen-activated protein kinases (S3 in Dataset S1). The significant overrepresentation of redox regulation among down-regulated genes indicates a possible importance in the initial response to low temperature.
The statistical analyses reveal that there are no functional groups that are changed specifically in the medium term (Table 1). The miscellaneous group contains significantly more down-regulated genes than expected, but the p-values indicate this is a medium- to long-term response. The peroxidase subgroup appears to be the most affected within this group, with eight of 72 members down-regulated, while in the long term group, ten of the 82 GDSL-motif lipases are down-regulated (M5 in Dataset S1). The other group that is overrepresented among down-regulated genes in the medium to long term is photosynthesis (Table 1). The down-regulation of photosynthesis is general, with the light reactions, Calvin cycle, and photorespiration all down-regulated in the medium to long term. Tetrapyrrole synthesis is significantly (p < 0.01) overrepresented among down-regulated genes in the long term (M6, L6 in Database S1). Genes from fermentation are up-regulated in both the medium and long term, including lactate dehydrogenase, pyruvate decarboxylase, and alcohol dehydrogenase.
The highest number of significantly affected groups is seen in the long term. In addition to the shared responses already listed, there are a number of long-term–specific changes. Primary metabolism is strongly affected, with the functional groups of minor and major carbohydrate metabolism and glycolysis all containing significantly more up-regulated genes than expected (Table 1). In addition, there could be a similar response for the tricarboxylic acid cycle (TCA) and amino acid groups, which also have low p-values. Major carbohydrate and lipid metabolism are significantly overrepresented among down-regulated genes. Many of these long-term changes are already observed in the medium term, as indicated by the proportion changes of the medium-term genes and the proportion changes and p-values of the genes that are common to both the medium and long term (Table 1). Sucrose synthesis and the conversion of UDP-glucose are both significantly overrepresented among long-term up-regulated genes, while starch degradation is significantly (p < 0.01) overrepresented among down-regulated genes (Table S3A). The functional groups of RNA, protein, and signaling all contain significantly fewer down-regulated genes than expected (Table 1).
Analysis of COR Genes
It has been previously reported that many COR/late embryogenesis abundant (LEA) genes are cold induced [6]. These genes encode hydrophilic proteins that are thought to be important for freezing tolerance [3]. The hydrophilic nature of proteins encoded by cold-responsive genes of unknown function was used to designate them as putative COR genes [6]. An earlier comparative study on water stress in prokaryotes and eukaryotes had defined a subgroup of LEA proteins (including two COR proteins), called hydrophilins, that had a mean hydrophilicity of over 1 and a glycine content of over 0.08 [24]. Another group of LEA proteins, with hydrophilicity indices between 0 and 0.5, were not included in this classification.
Genes encoding highly hydrophilic proteins (hydrophilicity of about 1) are significantly more abundant than expected among short-, medium-, and long-term cold up-regulated genes in comparison to all genes (Table S4). This increased abundance is higher for genes that are up-regulated at multiple times. In contrast, genes down-regulated in response to cold encode proteins with hydrophilicity indices more similar to those expected by chance. Among long-term down-regulated genes, there are actually significantly fewer genes coding for highly hydrophilic proteins than expected (Table S4).
Many of the COR proteins have glycine contents of more than about 0.08 (Table 2). The hydrophilicity indices of these proteins range from 0.29 to 1.19 (Table 2). The proportions of genes coding for proteins with glycine content of more than 0.08 among those up- and down-regulated in response to cold are significantly higher than expected (Table S4). Among up-regulated genes there are many that belong to the abiotic stress group that contribute to this increase, while for down-regulated genes the increase comes from proteins in the photosynthesis, cell wall, and hormone functional groups. High-glycine proteins with roles in abiotic stress are mainly hydrophilic, with the notable exceptions of the highly hydrophobic RCI2A and RCI2B, known to be induced by abiotic stress [25]. Among hydrophobic proteins, enzymes of primary metabolism and transport proteins also contribute to the increased abundance of proteins with high glycine content. The genes up-regulated by cold include 58 encoding unknown proteins with high glycine content, 11 of which are induced at multiple times (Table S5).
Table 2 Mean Hydrophilicity, Glycine Content, Expression Changes, and the Frequencies of Cis-Elements for Known COR Proteins
aPutative signal peptides removed prior to calculations.
bAGI locus codes can be found at the Institute for Genomic Research Web site (http://www.tigr.org).
The predicted proteins from 12 short-, 20 medium-, and 13 long-term cold up-regulated genes meet the criteria of being hydrophilins. The proportions of these genes that are up-regulated are significantly higher than expected at all times, but highest in the medium to long term (Table S4). The majority of these proteins that are annotated interact with RNA, with roles as transcription factors, splicing factors, and ribosomal proteins. The COR proteins RAB18, XERO2, RD29A, RD29B, COR15a, and COR8.5 [6] are also hydrophilins (Table 2). There are 11 up-regulated genes encoding unknown proteins that are possible hydrophilins, most of which have not been previously reported (Table S5). In addition, another unreported protein meeting the hydrophilin criteria is annotated as an abscisic acid (ABA)-responsive protein similar to KIN1 from Brassica napus.
Transcriptional Regulation
Transcriptional regulation of the response to low temperature is the subject of intensive research. Our analysis focuses on identifying families that are overrepresented among cold-responsive transcription factors. These families may be important during cold acclimation. The complexity of the overall regulation is clear, with 111, 141, and 58 known and putative transcription factors up-regulated, and 25, 133, and 84 down-regulated in the short, medium, and long term, respectively. In the short term there is a significant (p < 1 × 10−5) overrepresentation of the APETALA2/ethylene response element binding protein (AP2/EREBP) family among cold up-regulated transcription factors (Table S6). This family contains 17 up-regulated members, including CBF1–CBF3, DREB2B, and RAV1 (S7 in Dataset S1). Although many members continue to be up-regulated, the total number is not significantly different to that expected in the medium term and less significant in the long term. The CONSTANS (CO)-like family is also significantly (p < 1 × 10−6) overrepresented among short-term up-regulated transcription factors, but this overrepresentation continues to be significant (p < 1 × 10−3) in the medium to long term (Table S6). In the medium term the type B response regulator, basic leucine zipper, heat shock transcription factor (HSF), and TCP-domain families are significantly overrepresented among up-regulated transcription factors. The p-values indicate that the HSF and TCP-domain families are also overrepresented in the short term. In the long term the auxin response factor (ARF) family is also likely to be overrepresented (p = 0.051) (Table S6). There are 13 transcription factor genes that are up-regulated at all times, and these include the AP2/EREBP members RAP2.1, RAV1, and CBF2, and four from the CO-like family (S7, M7, and L7 in Dataset S1). There are too few short-term down-regulated transcription factors to draw any conclusions. In the medium term there is a significant overrepresentation of members from the MYB-related and Aux/IAA (p = 0.057) families among down-regulated transcription factors (Table S6). In the long term, the basic helix-loop-helix (bHLH) and Aux/IAA families are significantly overrepresented (M7 and L7 in Dataset S1).
Promoter Analysis of Cold-Responsive Genes
The CBF/DREB transcription factors specifically bind the dehydration-responsive element (DRE)/C-repeat cis-acting element that is present in the promoter regions of many COR genes [26,27]. The DRE core motif is A/GCCGAC, although recently the consensus motif for the binding of CBF3/DREB1A has been further characterized as A/GCCGACNT [12]. This study also found the ABA response element (ABRE) ACGTGG/T to be overrepresented in the promoters of cold up-regulated genes. We compared the frequencies of the DRE core, the A/GCCGACNT motif, and the ABRE motif in the 500- and 500–1,000-bp promoter regions of our consensus cold-responsive genes. Statistical comparisons were made against the frequencies found for all genes in our database. The ABRE motif was significantly (p < 1 × 10−6) overrepresented in the 500-bp promoter regions among all cold up-regulated genes, but was significant only in the short term for motifs found in the 500–1,000-bp regions (Table 3). The A/GCCGACNT motif was significantly (p < 1 × 10−8) overrepresented in the 500-bp promoter regions among all cold up-regulated genes (Table 3). In addition, the A/GCCGAC and A/GCCGACNT motifs were generally overrepresented in the 500-bp and 500–1,000-bp promoter regions, respectively, although these differences were only significant at the p < 0.01 level for medium-term up-regulated genes (Table 3). The increase in the frequency of DRE elements is greatest for the A/GCCGACNT motif, and increases further among genes induced at multiple times (Table 3). Of the genes up-regulated in the short, medium, and long term we identified, respectively, 80, 118, and 63 genes (165 unique) containing the A/GCCGACNT motif within their 500-bp promoter regions. There were 83 additional genes with the A/GCCGACNT motif within 500–1,000 bp upstream and 230 additional genes with the A/GCCGAC motif within 500 bp upstream of their start codons. In total, 478 of the 2,055 up-regulated genes meet these criteria and are therefore potential targets for CBF transcription factors. This gives a proportion of 0.23, which is significantly higher than the 0.16 found among all genes in our database. However, our statistical analysis revealed that the A/GCCGACNT motif within the 500-bp promoter region is the most specific to cold-responsive genes, occurring in the promoters of twice the number of expected genes.
Table 3 Enrichment of DRE and ABRE Cis-Elements among Consensus Cold-Responsive Genes
aExcludes those already identified in the A/GCCGACNT motif.
+/−, fold change of the proportion of cold-responsive genes containing the cis-element versus the proportion of all genes containing that element; p-Value, p-value from Fisher exact test; p-values less than 0.05, 0.01 and 0.001 are highlighted in purple, green and red respectively.
Discussion
There are now a large number of studies investigating the changes in gene expression in response to low temperature (for example, see the Web sites at http://affymetrix.arabidopsis.info/ and http://www.arabidopsis.org/info/expression/ATGenExpress.jsp) [4–9]. Although some of these studies have used time courses, for individual time points biological variability has been sampled only in duplicate at most. This has led to the use of arbitrary cutoffs to define differentially expressed genes. In addition, only small numbers of candidate genes have been confirmed by independent means. In this study, we report a genome-wide validation of expression changes using independent microarray platforms for Arabidopsis, and through statistical analysis we used these data to determine changes in gene expression after 14 d of cold acclimation. We integrated these data with comprehensive analyses of published and publicly available data to define consensus lists of short-, medium-, and long-term cold-responsive genes.
The Analysis and Validation of Expression Profiling Data
Despite the large number of expression profiling studies on plants, there is often little discussion of the data analysis and validity of the measured expression estimates. In almost all microarray studies there is the usually unmentioned assumption that either the vast majority of transcripts do not change, or that any changes that do occur are balanced [28]. This is because the expression levels of all genes are used for normalization. Even the most basic normalizations, such as a mean or median scaling, are based on the assumption that the average numbers and abundances of mRNAs are similar between samples. Other normalizations, such as quantile or loess, additionally assume similar distributions of mRNA abundance by fitting a linear relationship between all samples. This means that the expression estimate of a single mRNA should be considered as relative to the total combined expression of all transcripts (i.e., total mRNA). The assumption that most genes do not change is probably violated in many studies, but is often not evident, because stringent but arbitrary cutoffs are used that reveal relatively few changes. If the majority of changes occur in one direction, or for transcripts of a certain intensity (e.g., highly expressed genes), then the total amount or distribution of mRNA will change and any internal normalization will lead to bias due to the reasons highlighted above. As normalization is necessary to remove technical variability, and without it most patterns would be obscured by noise, these issues must be considered before data can be confidently interpreted.
To investigate the possibility of bias due to normalization, we compared the expression of a large number of commonly used housekeeping genes between treatments. Genes annotated as tubulin, actin, polyubiquitin, DNA-dependent RNA polymerase, cytosolic cyclophilins, and elongation factors were selected. It seems likely that if there were changes in the total amount or distribution of mRNA, then diverse housekeeping genes such as these would show a bias in their correlation between treatments. Similar to previous reports [7,29], there are effects on the expression of individual genes but overall there is a close correlation, giving no indication that any global mRNA changes occurred (see Figure 1). Of 44 selected housekeeping genes, only two are changed in both of our array datasets, indicating only slight changes in gene expression. Additionally, our data show 93% agreement with long-term cold-responsive genes reported using nonprofiling techniques (see Table S1). Together, these approaches offer good evidence that global mRNA changes did not occur in our study.
Previous studies in the field of cold-responsive gene expression have not considered global mRNA changes. Using our database of cold-responsive genes, we looked at the expression of the selected housekeeping genes across all studies (Table S7). In the long term, there are the two up-regulated genes from our study, plus one additional down-regulated gene. In the short term there is just one up-regulated gene. However, in the medium term there are no up-regulated genes, but six of the 44 genes are down-regulated. This has a high probability of not being due to chance (p = 0.052), which could be due to a number of reasons. First, the two elongation factors, three DNA-dependent RNA polymerases, and Actin 11 could be cold-responsive isoforms of these gene families. Second, the high expression of housekeeping genes means that they are more likely to be detected as “present” using the Affymetrix MAS5 software, and as only genes “present” before cold treatment can be down-regulated, this would contribute to their increased abundance compared to all genes. Finally, the possibility of a global change in total mRNA needs to be considered. In the AtGenExpress time series, there are about 9-, 3-, and 2-fold more up-regulated than down-regulated genes at 1, 3, and 6 h respectively, while at 12 h these changes are balanced (see Table S2). If a sufficient proportion of transcripts increased in abundance, an increase in the total amount of mRNA within the cell could occur, and, following normalization, genes that are not changed could be detected as down-regulated. However, the high magnitude of the housekeeping genes down-regulation, and the low proportion of early unbalanced changes, indicate that the first two reasons are the more likely.
Comparison of ATH1 and MWG Arrays
The agreement between expression changes identified using independent Arabidopsis microarray platforms has not received much attention. This leaves some doubt over the universal validity of results obtained using a single technology, even if several of the larger changes were confirmed by independent methods. In our study, we find a good correlation between the results obtained using the ATH1 and MWG arrays. Although transcripts with higher abundances should give higher intensities, differences in probe sequence could lead to discrepancies. This can be due to the binding efficiency of the probe to the target (G/C content and interference from labeled nucleotides) and the labeling potential of the target sequence. Considering these factors, the correlation of the actual intensities between the two technologies is better than expected (see Figure 2A). This comparison of absolute expression values suggests that the ATH1 array has a greater dynamic range at lower intensities, while the arrays behave more similarly at higher intensities. This could reflect the decreasing influence of the mismatch probes at higher intensities. To increase dynamic range, MWG combines measurements using different photomultiplier tube gain settings, and this may also contribute to the nonlinear relationship. The correlation of 0.7 between absolute signal for the MWG and ATH1 arrays is very similar to the recent report of 0.72 between the Agilent 60-mer oligonucleotide and ATH1 arrays [30]. However, this correlation was obtained for filtered data and no details of the distribution were given.
If statistical testing is used to detect differential gene expression, it is necessary to assess the suitability of the chosen test. The unmodified t-statistic can lead to false positives from genes with low variances but very small expression changes. Various methods have been proposed to overcome this limitation [31]. One of the most recent and versatile developments, used in the present study, is the empirical Bayes moderation of t-statistics implemented in the Limma Bioconductor package [32]. Once a statistical test is used, the validity of the p-values and their correction for multiple testing at an accurate threshold requires them to fit an expected distribution [19]. The ATH1 and MWG data both meet these criteria, allowing us to compare significant changes in expression across the two platforms.
It would be expected that the signal changes are more stable between technologies than are absolute expression levels, as sequence-specific interference will have less effect. Signal changes show a linear correlation between both array datasets, although the magnitudes of changes are smaller for the MWG than the ATH1 data (see Figure 2B). As with the absolute expression, this could indicate a lower dynamic range for the MWG array. Alternatively, the linear correlation could indicate a difference in scale between the two expression measures. Keeping these small differences in mind, there is close agreement between the significant changes detected by the two technologies (see Figures 2B and 3), and this agreement would likely be higher were it not for the high numbers of false negatives using both technologies. The low proportion of opposite changes and the general agreement of p-values also indicate close agreement. Taking these factors into consideration, it seems likely that the vast majority of significant changes are due to biological rather than technical variation, and that data from either microarray can be interpreted with confidence. Similar conclusions have been made recently for the comparison of mammalian Affymetrix and spotted oligonucleotide arrays [18,33].
Comparison of Previous Studies
All of the comparisons made between previous studies were based on genes that changed in both studies. The proportion of all changes that were common to both studies was extremely variable (see Table S1B). The impact of the high false-negative rates on comparisons has already been discussed. The different cutoffs used in previous studies are likely to lead to increased false-negative rates. Therefore, the massive effect of using cutoffs on the change/no change classification of a gene, and the different genes represented on each array, allow no interpretation of the proportion of common genes. These factors have little effect on the comparison of expression changes for genes that are identified in both studies, so this represents the best method of comparison. In general, considering the diverse nature of growing conditions, treatments, and profiling technologies, the agreement between all studies was good. The close agreement between the ATH1 datasets is obviously due to the use of the same array, normalization, and analysis. The agreements within short-, medium-, and long-term data support our separation of studies into these three classes. Across all time classes, the lowest agreement with our datasets was found in comparison to poorly replicated studies using lower cutoffs. The use of a single biological sample hybridized to either one or multiple chips or duplicate biological samples hybridized to a single chip will result in an increased false-positive rate caused by the underestimation of biological or technical variability.
Cold-Responsive Gene Expression
In studies using the approximately 8,000-gene Arabidopsis Genome Array (AtGenome) array, the numbers of genes reported to respond to low temperature are about 300 [6] and 1,100 [7]. These data suggest that 4%–14% of the Arabidopsis genome is cold responsive. These studies focused on controlling the false-positive rate and, with no possibility of statistical analysis, there was little discussion of false negatives. Our data indicate that the true number of cold-responsive transcripts is likely to be approximately 45% of those represented on the array. A comparison of the variability of all datasets indicates that, at 24 h, either more genes change or the magnitude of changes is greater than for other times, and that of these genes, significant proportions change in the short term. Such large numbers of cold-responsive genes are clearly higher than most expectations. These estimates will include even those genes that show small changes in expression. Small changes of gene expression of 10%–50%, although generally not considered, could be important for many responses, including that to low temperature. These changes may be more significant if they are coordinated within functional groups or specific pathways. Improved gene ontology and visualization of significant changes will be important if so many changes are to be meaningfully interpreted.
Integration of Data from Diverse Experiments into a Single Database
In the present study, samples harvested after 14 d of cold acclimation were used to reveal the stable changes in gene expression required to maintain the cold-acclimated state. This is before any new leaves develop at 4 °C. We chose to use soil-grown plants grown under a photoperiod, as these conditions are more similar to natural conditions. Obviously, under any experimental design, not all changes are relevant to cold acclimation, as many could be specific to the conditions used. Genes that are consistently cold responsive under diverse experimental conditions are likely to be of more general importance in the response of plants to low temperature. Therefore, we have incorporated all available data into consensus lists of short-, medium-, and long-term cold-responsive genes. These include data from plants of different ages grown on agar plates, in hydroponics and soil, and under continuous light or different photoperiods.
We identified 808, 1,224, and 672 up-regulated genes, and 240, 1,364, and 915 down-regulated genes in response to cold in the short, medium, and long term, respectively (Figure 4). These consensus lists offer the most comprehensive overview of cold-responsive gene expression presently available, to our knowledge. Moreover, the complete database offers an “at-a-glance” summary of the cold responsiveness of 22,043 genes (see Table S2). This will be a valuable tool for single- and multi-gene characterizations.
Annotation Based Analysis Reveals the Most Prevalent Cold Responsive Expression Changes
The Mapman software offers improved ontology and the ability to visualize gene expression overlaid onto biochemical pathways or diagrams [22,23]. In addition, the TAIR AraCyc pathways detail genes involved in various biological processes. The complete datasets are available in Dataset S1 and Table S3, allowing the reader to obtain detailed overviews of cold-responsive gene expression. As there have been previous studies on the changes in gene expression in response to cold, many of the observed changes have been discussed [5–8]. These studies have mainly focused on individual or small groups of changes. Moreover, discussion of functional classification has not been related to, or statistically tested against, that of all genes. The knowledge that the number of genes changing expression within a functional group is significantly different from that expected by chance is important information that has been largely neglected in plant studies.
The overrepresentation of the stress group among up-regulated genes is expected and clearly related to the literature reports incorporated into our database that have also been used to provide gene ontology. The up-regulation of genes involved in temperature, drought, and salt stresses is known and reflects the cross-talk between signaling pathways [8]. Among short-term up-regulated genes there is a significant overrepresentation of transport proteins, including ATP binding cassette, sugar and phosphate transporters, and ATPases, at the plasma, vacuolar, and plastid membranes. This indicates that, in addition to changes in their metabolism, the transport of these substances into, out of, and within the cell may be an essential component of the initial response to low temperature. The transport of sugars appears to be a short-term–specific response, with the majority of sugar transporters becoming down-regulated in the long term. The redistribution of sugars may become less important as rearrangement of carbohydrate metabolism becomes established. The significant overrepresentation of carbohydrate metabolism among long-term up-regulated genes asserts that such rearrangement is occurring.
The accumulation of many compatible solutes and metabolites is known to occur in response to low temperature [13,34,35]. The well-documented increases in proline, and raffinose involve the increased expression of the P5CS2 and AtGolS3 genes, respectively [13]. The increased abundance of enzymes responsible for raffinose synthesis may be more general, as in addition to AtGolS3, genes encoding three other galactinol synthases (Institute for Genomic Research [http://www.tigr.org] Arabidopsis Genome Initiative [AGI] locus codes At1g56600, At2g47180, At3g28340), a raffinose synthase (At5g40390) [36] and enzymes responsible for myo-inositol and sucrose synthesis are all up-regulated in response to cold. The recently reported increases in pyruvate, alpha-oxoglutarate, succinate, and fumarate [13,34], are accompanied by increased expression of at least one enzyme responsible for their synthesis in the medium or long term, indicating a general increase in TCA cycle activity. The discussed overrepresentation of enzymes of carbohydrate metabolism includes also those involved in the production of glucose-6-phosphate, fructose-6-phosphate, glucose, maltose, fructose, sucrose, and trehalose, all of which accumulate in response to cold [13,34]. Expression changes can also explain many other metabolite changes, but caution is needed, because so many metabolite and expression changes occur, that patterns can always be found. Good annotation and statistical testing provide a more robust basis for conclusions based on global changes of pathways or functional groups. The Mapman and AraCyc analyses reveal coordinated changes within primary metabolism (S8–S10, M8–M10, and L8–L10 in Dataset S1). In the short term, genes encoding enzymes for sucrose and starch degradation are up-regulated, while in the medium to long term, sucrose degradation by invertase and starch degradation become down-regulated [35]. Sucrose degradation by sucrose synthase continues to be up-regulated, and enzymes for the conversion of the resulting UDP-glucose and fructose are also up-regulated. These could then enter into glycolysis and subsequently into the TCA cycle or fermentation, all of which are overrepresented among long-term up-regulated genes. This broad up-regulation of primary metabolism could explain the large number of metabolites that increase in response to cold acclimation [13,34].
There is a significant overrepresentation of down-regulated genes of lipid metabolism in the long term. Cellular membranes are a primary site of freezing damage, and changes in their composition are thought to be important for cold acclimation [1]. There is an increase in the proportion of di-unsaturated fatty acids which can reduce freezing-induced membrane damage in the plasma membrane [37]. The up-regulation of desaturases has been reported and these might be important in the conversion of existing fatty acids [38,39]. The down-regulated subgroups of lipid metabolism are those involved in fatty acid synthesis and elongation, synthesis of phospholipids and steroids/squalene, and degradation by lipases and lysophospholipases. Such a general reduction of lipid metabolism is probably related to lower demand due to the reduced growth at low temperatures. However, it could also indicate a key role for transcriptional down-regulation in changing enzyme activities necessary to maintain the altered lipid composition seen after cold acclimation [37]. Cell wall modification also occurs during cold acclimation [1] and, in common with lipid metabolism, our analysis shows that down-regulation may be important. In contrast, this down-regulation occurs in the short to medium term, indicating a possible role in the initial reduction in growth rate in response to low temperature.
Recent work on the response of carp to low temperature also identified genes encoding TCA cycle and mitochondrial electron transport components and transport as prominent functional groups among up-regulated genes [40], indicating possible functional similarities to Arabidopsis. These authors also made comparisons with earlier work on yeast, and although some genes such as delta-9-desaturase and translation initiation factors also have homologs up-regulated in Arabidopsis, the numbers are not sufficient to draw conclusions on conservation among functional groups.
The Down-Regulation of Photosynthesis and Coordinated Changes in Other Pathways as Significant Response to Low Temperature
In Arabidopsis, 3 d of cold acclimation at 5 °C inhibited light-saturated rates of photosynthesis by approximately 75% [41]. The down-regulation of photosynthetic genes in response to cold has been previously reported [41,42], yet their significant overrepresentation among medium- to long-term down-regulated genes has not. These changes are massive, with 3- and 5-fold the expected number of genes down-regulated in the medium and long term, respectively. This appears to be a general response involving all aspects of photosynthesis. Tetrapyrrole synthesis is also significantly overrepresented among down-regulated genes. Light signaling contains higher proportions of down-regulated genes in the medium to long term (S3, M3, and L3 in Dataset S1), but in contrast, is mainly up-regulated in the short term, probably resulting from a transient “shock” prior to adaptation [6].
In cereals there is a short-term phosphate limitation of photosynthesis and a reduction in the supply of ATP, while in Arabidopsis ATP and ATP:ADP ratio actually increase [41]. However, in Arabidopsis it was concluded that low phosphate still plays an important role in triggering cold acclimation [42]. Such limitation is supported by the short-term down-regulation of a putative phosphate-induced gene (AGI locus code At4g08950). Our functional group and pathway analyses also revealed other potentially related changes during exposure to low temperature. The overrepresentation of transporters in the short-term includes ATPases and phosphate transporters which could be important in the response to short-term phosphate limitation. In nucleotide metabolism there are significantly more genes down-regulated in the short-term than expected indicating a possible role in changing cellular energy status in response to cold. The down-regulation of apyrase APY2 and an adenylate kinase (AGI locus code At5g47840) could both contribute to increase the ATP:ADP ratio. The significant overrepresentation of up-regulated genes of secondary metabolism at all times is mainly due to those involved in the synthesis of flavonoids (S2, M2, and L2 in Dataset S1). Flavonoids are known to accumulate during plant stress and they are thought to protect cells from photo-oxidative stress by absorbing UV light [43]. This can prevent damage to photosystem II by increased excitation due to the reduced electron consumption at low temperatures [44].
As discussed, exposure to low-temperatures can lead to oxidative damage, particularly under high light, so reduced light intensities are generally used to minimize these effects. Most studies incorporated into the consensus lists of cold-responsive genes used reduced intensities, and as previously suggested [6], many of the changes of photosynthesis and related processes could represent general responses to low light conditions. Therefore, although the occurrence of these changes is not in doubt, their specificity and importance for plant cold acclimation remain to be fully established.
COR/LEA Proteins and Their Importance for Plant Freezing Tolerance
One group of proteins that has received much attention is the COR/LEA proteins, some of which have protective effects during freezing [45,46]. In order to better characterize these proteins, we calculated mean hydrophilicity and glycine content of the proteins encoded by our consensus cold-responsive genes. The involvement of hydrophilic proteins in the response to cold [3] is supported, as hydrophilic proteins are significantly more abundant than expected among cold up-regulated genes. Of the established COR proteins, COR15a, COR15b, XERO2, RD29A, RD29B, RAB18, COR47, ERD10, and ERD14 have mean hydrophilicity greater than about 1 (see Table 2). Proteins with high glycine content are also significantly more abundant among cold-responsive genes. However, this increased abundance is due only to genes from the stress functional group for up-regulated genes. The COR proteins XERO2, RD29A, RD29B, RAB18, D113, M17, COR15a, COR15b, Di21, KIN1, and COR6.6 all have glycine contents of over about 0.08 (see Table 2). The hydrophobic cold-responsive proteins RCl2A and RCl2B [25] also have high glycine content. In fact, of the established COR proteins, only COR47, ERD10, and ERD14 have glycine contents of less than 0.08, and these proteins are all very hydrophilic with mean hydrophilicity greater than 1.2 (see Table 2). The cold-responsive genes encoding high-glycine proteins include 58 of unknown function, many of which may be similar to those already described (see Table S5).
The COR/LEA classification is broad, and their importance is often unclear. In order to improve their classification, we calculated the mean hydrophilicity and glycine content for COR/LEA proteins from various species shown to have effects that could relate to improved freezing tolerance (Table 4). COR15a has been shown to have protective effects on membranes during freezing [46], and conformational changes of maize DHN1 during binding to vesicles also suggest a role in the stabilization of membranes [47]. WCS120 from wheat and COR85 from spinach have been shown to increase enzyme stability during freezing [48,49], and PsLEAm stabilizes enzymes during drying [50]. Also, during drying a D-7 LEA protein was shown to stabilize sucrose glasses in vitro [51]. At the whole plant level, the overexpression of HVA1 was shown to improve salt and drought tolerance in rice [52], while in Arabidopsis the overexpression of the dehydrins RAB18 and COR47 or ERD10 and XERO2 enhanced freezing tolerance [45]. Our analysis reveals that all of these proteins are highly hydrophilic, with values of over about 1 and, with the exceptions of COR85, COR47, and ERD10, all have glycine contents over 0.08 (Table 4). It is likely that other cold-responsive proteins with similar properties are also involved in plant freezing tolerance. Therefore, we propose that the existing COR/LEA genes be classified based on these properties. Group 1 represents COR/LEA proteins generally meeting the definition of hydrophilins [24] with hydrophilicity over 1 and glycine contents over 0.08. Group 2 represents highly hydrophilic proteins that have glycine contents less than about 0.07 (see Table 2). There are eight proteins from the first group and three from the second that have evidence to support a functional role in plant freezing tolerance. Group 3 hydrophilic proteins, with hydrophilicity less than 1 and glycine contents over about 0.07 could also be retained, as it contains many named COR/LEA proteins. However, no members of this group have been shown to be important for plant freezing tolerance. Of the named unknown proteins reported to be unusually hydrophilic COR8.5, COR28, and COR8.6 fit into groups 1, 2, and 3 respectively [6]. Two additional named unknown proteins, COR12 (± signal peptide) and COR18 and eight of the 11 transiently expressed genes encoding COR/LEA-like hydrophilic proteins [6] do not fit into these groups. The LEA proteins designated by AGI locus codes At1g01470 and At1g16850 and reported as being cold-responsive [12,53] are also not members of these groups.
Table 4 Mean Hydrophilicity and Glycine Content of COR/LEA Proteins Shown to Have Effects that May Be Important for Plant Freezing Tolerance
aPutative signal peptides removed prior to calculations.
N/A, not available.
We have separated the known COR proteins (see Table 2) and cold induced unknown proteins (see Table S5) into these groups. A cutoff of 0.3 was used as a lower limit for the hydrophilicity of the unknown proteins, as this is close to the mean of all genes (0.32) and to the minimum for a named COR gene (0.29). Most of the established COR genes are highly up-regulated at multiple time points and have high frequencies of DRE and ABRE elements in their promoter regions (see Table 2). Only COR28, D113, and M17 are not up-regulated at multiple time points. We identify 11, 30, and 46 cold up-regulated unknown proteins that are members of group 1, 2, and 3 respectively (see Table S5). Obviously, not all of these genes are COR genes; some genes, such as those designated by AGI locus codes At1g13650 and At1g69160 are down-regulated at other times, while others, such as At1g73350 and At4g17410, are only weakly up-regulated (see Table S5). However, there are many others that provide interesting new candidate COR genes that are up-regulated at multiple time points. Some, but not all, contain DRE elements in their 1,000 bp upstream regions (see Table S5). In support of their involvement in protection against freezing stress, hydrophilins are the most significantly over-abundant group of proteins encoded by cold up-regulated genes (see Table S4).
The Continued Importance of Short-Term Cold-Responsive Transcription Factors in Regulating Long-Term Changes in Gene Expression
In all our datasets of cold induced genes, there is a significant overabundance of the DRE promoter element that is bound by the CBF/DREB transcription factors [26,27]. The further characterization of the consensus motif for CBF3/DREB1A as A/GCCGACNT [12] is supported by our study, as this motif is significantly overrepresented in the 500-bp promoter regions of all our cold up-regulated gene lists (see Table S4). The A/GCCGACNT motif in the 500–1,000-bp region and the A/GCCGAC motif in the 500-bp region are also overrepresented, but to a lesser extent. The occurrence of the A/GCCGAC motif in the 500–1,000-bp region is not higher than expected. Our data extend the number of cold-responsive genes that are members of the CBF regulon from 38 [12] to a possible 478 genes, representing about 23% of those up-regulated. Included among these genes are 41 known and putative transcription factors, which are mainly from the AP2/EREBP, CO-like, and C2H2 families (S7, M7, and L7 [DRE] in Dataset S1). However, not all of these genes are likely to be under the control of CBF transcription factors, as the frequency of these motifs among all genes is 0.16 and among down-regulated genes 0.145, indicating that the proportion of up-regulated genes in the CBF regulon may be closer to 0.1. Previous work has shown that cold-responsive genes in the CBF regulon respond within 12 h and continue to be up-regulated after 24 h [12], and that many of the genes are stably up-regulated in the long term [6]. We find that the frequency of the DRE element increases for genes up-regulated at multiple time points and is similar among lists for short to medium and medium to long term up-regulated genes.
Overexpression of the CBF transcription factors revealed overlapping and redundant functionality [11]. However, analysis of a mutant in which the CBF2 gene was disrupted indicated that this gene negatively regulates CBF1 and CBF3 and may ensure their expression is transient and tightly controlled [14]. CBF2 is up-regulated approximately 4-fold in the long term. Although CBF1 and CBF3 are not on our long term consensus list, it is still likely they are up-regulated. They are both about 2-fold up-regulated in our ATH1 dataset, and in the other large long-term study [9], they are up-regulated approximately 2- and 8-fold in plants grown on plates and soil, respectively, but are excluded from the final lists due to the MAS5 filtering criteria. These expression levels are much lower than their initial induction, but are still significantly higher than before cold treatment. In addition, 12 other short-term–responsive transcription factors continue to be up-regulated at all times during long-term cold acclimation. These include RAP2.1, RAV1, ZAT10, and COL1, all of which contain the A/GCCGACNT motif in their 1,000-bp promoter regions. This indicates that some of the regulators that are responsible for the initial large changes of expression may continue to regulate the smaller but stable changes of gene expression that are required to maintain the cold acclimated state.
Global Analysis Reveals That Many Transcription Factor Families Show Coordinate Regulation
The transcriptional regulation of cold-responsive gene expression is well studied and much is now known. The AP2/EREBP family, with 17 members up-regulated in the short term, has received most attention. The key role of the CBF/DREB transcription factors is well established [6,10,13]. Other members, such as RAP2.1, RAP2.6, and RAV1 have also attracted interest [6]. The significant role of this family in the short-term is supported by our statistical analysis (see Table S6). Our analysis reveals other families that are significantly overrepresented among up-regulated transcription factors that may also be important for cold acclimation. The CO-like family is the most significantly overrepresented family at all times. CO controls the photoperiodic regulation of flowering. CONSTANS-LIKE 1 and 2 (COL1 and 2) are also circadian controlled and COL1 overexpression influences circadian rhythms and may be light sensitive [54]. The overlap between stress- and circadian-regulated gene expression has been highlighted in previous studies [7]. The overrepresentation of CO-like transcription factors may indicate an important role for these transcription factors in the integration of cold-responsive gene expression with light and circadian regulation.
As has been previously reported, there are also transcription factors down-regulated in the short term [6]. With large numbers of transcripts down-regulated in the medium to long term, it seems likely that some of the 25 short-term and 133 medium-term down-regulated transcription factors also have key roles in the response of plants to low temperature. The significant overrepresentation of transcription factor families among down-regulated transcription factors has not been previously reported and indicates there may be functional overlap in their response to low temperature. The Aux/IAA family is the most affected in the long term and is also significantly overrepresented in the medium term. The bHLH family is significantly overrepresented in the long term. The affected members of these families could provide new insights into processes important in the long-term response to low temperature.
Changes in Hormone Processes May Be Central to the Regulation of Growth during Cold Acclimation
The hormone functional group is the only group overrepresented among down-regulated genes at all time points, indicating a functional relevance of altered hormonal regulation during the process of plant cold acclimation. Auxin-induced genes are most abundant, although the metabolism, signaling, and response of other hormones are also affected. It has previously been shown that there is cross-talk between cold- and ABA-regulated gene expression, but that the accumulation of ABA is transient, peaking at 24 h in the cold [3]. Based on the timing of expression changes, it was concluded that there may also be cross-talk with ethylene responses, but that little ethylene was likely to be produced [6]. These responses both involve short-term up-regulation of gene expression, while our analysis reveals the importance of down-regulation. This is most likely related to the decreased growth observed in response to low temperature, but whether the down-regulated genes are causal in this relationship is uncertain. As many metabolites increase in response to low temperature [13,34,35], resource limitation of growth is unlikely. This leads to the question of which mechanisms are responsible for regulating growth. Changes in the expression of enzymes of hormone metabolism and coordinate changes among hormone responsive transcription factor families support the role of hormones in this regulation.
A phylogenetic analysis based on bHLH domain amino acid sequences revealed 21 sub-families of bHLH transcription factors [55]. The sub-family 18 is the largest, containing 16 of the 117 genes from our database belonging to these families including BEE1, BEE2, and BEE3, which are redundant brassinosteroid (BR)-responsive genes that are also repressed by ABA [56]. This family contains two of the nine medium-term (AGI locus codes At3g07340 and At3g23690) and three of the nine long-term (BEE2 and AGI locus codes At5g39860 and At5g50915) bHLH transcription factors down-regulated in response to cold. Our data indicate that BEE1, BEE3, and AGI locus codes At2g18300 and At3g23690 are also long-term down-regulated, although these differences are significant for only one of our two datasets. Additionally, analyses of raw ATH1 data indicate that many of these changes are likely to begin at 24 h (unpublished data). A survey of the AtGenExpress hormone data series reveals that genes designated by AGI locus codes At2g18300 and At5g39860 are strongly down-regulated by ABA and At5g50915 is up-regulated by brassinolide after 3 h [57].
Members of the Aux/IAA family of transcription factors, which are up-regulated by auxin, repress the interaction of ARF family proteins with target genes, leading to either activation or repression of transcription [58]. Among our long-term cold-responsive genes, the Aux/IAA family is coordinately down-regulated and the ARF family is up-regulated. The up-regulation of Aux/IAA and down-regulation of ARF genes have been reported in response to BR treatment [59]. Moreover, it has been suggested that the Aux/IAA transcription factors may be important cross-talk points in BR, auxin, light, and other signaling pathways [60].
The cross-talk between auxin and BR may be more extensive; a recent study reported a 25% overlap between the auxin- and BR-induced transcriptomes [59]. There are 332 genes represented in our database that are reported to be up-regulated in response to BR [59]; of these, 46 and 38 are significantly down-regulated in the medium and long term, while the corresponding up-regulated lists contain only 19 and 11 of these genes, respectively. Of a similar number of auxin up-regulated genes there are 10 and 28 long-term up- and down-regulated in response to cold; however, 11 of the down-regulated genes also respond to BR [59]. The number of BR up-regulated genes down-regulated by cold is about 2.8-fold the number expected by chance, and includes ten genes annotated as auxin-responsive proteins. Some of these are members of the Aux/IAA transcription factor family. Two genes are annotated as being involved in gibberellin processes and three are GDSL-lipases. An alternative study reported less overlap between IAA- and BR-induced genes, but of the 31 genes up-regulated by BR and auxin [60], six are down-regulated in the medium term in response to cold. The AtGenExpress hormone data indicate that many of the genes down-regulated by cold and up-regulated in response to auxin and BR are also negatively regulated by ABA [57]. As previously suggested [59], these shared genes may represent a common growth signature, and their down-regulation could also be correlated with reduced growth in response to low temperatures.
The analysis of hormone biosynthetic pathways indicates that any reduction in auxin levels does not appear to be controlled at the transcriptional level, as the only change for genes involved in auxin biosynthesis is the medium-term up-regulation of two enzymes (see Table S3B). There are 16 annotated genes encoding enzymes of the brassinolide biosynthesis pathway, of which five are significantly down-regulated in the long term (see Table S3B). Of these, DWF1 is also down-regulated in the medium term, and STE1 in both the short and medium term. Additionally, the cytochrome P450 gene CYP90A1 is down-regulated in both the short and medium term, and the recently reported CYP90C1 [61] is medium-term down-regulated. These changes are obviously responsible for the significant overrepresentation of BR biosynthesis among short- and long-term down-regulated genes, and they contribute to the observed overrepresentation of lipid metabolism among long-term down-regulated genes (see Tables 1 and S3A). Of the five annotated enzymes of ABA synthesis, the one designated AGI locus code At3g14440 is highly induced in both the short and medium term.
These findings indicate that ABA may have a significant role in controlling longer-term responses to low temperature. The down-regulation of many BR-responsive genes, the known effect of BR on cell division and expansion [62], and the significant down-regulation of genes involved in BR synthesis indicate that reduced BR signaling may have a central role in reducing growth at low temperatures. The Aux/IAA and bHLH families of transcription factors contain a number of candidates for the regulation of these responses.
In conclusion, the data discussed in this paper provide the most comprehensive overview of cold-responsive gene expression available. Our analyses provide a statistical basis for many previously reported changes and, moreover, identify which processes predominate during cold acclimation. We believe that this approach offers the fullest characterization possible using expression profiling of a single genotype exposed to low temperature. The next significant advance will be to more fully integrate changes of gene expression with changes detected using metabolic profiling. However, approaches using Arabidopsis lines with quantitative differences in freezing tolerance due to transgenic or natural genetic variation will be crucial to identify the processes that are central to plant freezing tolerance.
Materials and Methods
Plant material.
The A. thaliana accession Columbia (Col-0-G1) [63] was grown in soil in a greenhouse with supplementary light providing a 16 h photoperiod at a minimum of 200 μmol m−2 s−1, and at a day/night temperature of 20 °C/18 °C until bolting (43–46 days after germination). For cold acclimation, plants were transferred to a 4 °C growth chamber with a 16-h photoperiod at 90 μmol m−2 s−1 for an additional 14 d [64]. Whole rosettes pooled from ten plants were harvested after approximately 8 h of light. The experiment was repeated to give three independent biological replicates.
Expression analysis.
Extraction, labeling, hybridization, and scanning for the ATH1 arrays was as described previously [17]. Each biological replica was hybridized to a single ATH1 array at the RZPD Deutsches Resourcenzentrum für Genomforschung, Berlin, Germany (http://www.rzpd.de). Aliquots from the same RNA extracts were processed and hybridized to the MWG arrays as described (http://www.THE-MWG.com). All samples were labeled with Cy3 and single-channel hybridized to two MWG arrays to give two technical replicates. All data are available in the TAIR database (http://www.arabidopsis.org) under the submission number ME00369.
ATH1 data preprocessing.
Data were analyzed using the bioconductor software project [65]. The starting point for all analyses was the “.CEL” files from the MAS5 software. Data quality was assessed using functions in the affy [66] and affyPLM packages. The GCRMA algorithm (ver. 1.1.0) was used to obtain expression estimates [67]. The array element mappings of Affymetrix probe set IDs to AGI locus codes by TAIR (01/06/04) were used. Only probe sets from the ATH1 array that matched either a single locus code or two with clearly matching annotations were included in further analyses (21,581 probe sets). Only probe sets from the AtGenome that matched a single locus code were used (7,097 probe sets) when compiling the database.
MWG data preprocessing.
Data were obtained from MWG hybridization service (http://www.THE-MWG.com). Briefly, MWG arrays were scanned at multiple photomultiplier amplification settings (PMT gain) and Imagene intensities were combined and log scale normalized using the MAVI Pro software. Spots flagged as low quality and those not matching to an AGI locus code were excluded from further analyses. Data quality of raw and normalized data was assessed using standard functions of the R software (http://www.r-project.org).
Statistical testing for differential expression.
The two datasets were analyzed separately using the Limma bioconductor package. A linear model was fitted for each gene with the “NA” expression acting as an intercept and the contrast “ACC-NA” as the coefficient of interest. The “duplicateCorrelation” function was used for the MWG technical replication. An empirical Bayes approach was used to moderate t-statistics [32], and an FDR correction for multiple testing was used [20].
Database construction.
A database of previously reported cold-responsive genes in Arabidopsis was compiled from the literature and publicly available datasets (http://affymetrix.arabidopsis.info/; http://www.arabidopsis.org/info/expression/ATGenExpress.jsp). Transcripts reported in the literature to be up- or down-regulated in response to low temperature were classified as responding in the short (12 h or less), medium (24–48 h), or long term (over 48 h). Studies using expression profiling were analyzed separately (Table S8). Unless more complete data were available, cutoffs suggested by the authors were used. Updated mappings of AGI locus codes were used where possible and any duplicate values were averaged. ATH1 data were analyzed with the comparison mode of the Affymetrix MAS5 software. All possible pairwise comparisons between cold-treated samples and the appropriate controls were made. Probe sets meeting the Affymetrix suggested criteria (Increase/Decrease, appropriate “present” call, 2-fold change) in all comparisons were considered cold responsive.
Data analysis.
All protein and nucleotide sequences used were obtained from TAIR. The Web-based TAIR Patmatch (http://www.arabidopsis.org/cgi-bin/patmatch/nph-patmatch.pl) was used to search for cis-elements in promoter regions. Mean hydrophilicity using a sliding window of 12 and glycine content were calculated as described [24] for all Arabidopsis genes (TAIR file, ATH1_pep_cm_20040228) using the “seqinr” package and a custom script in the R software. Statistical comparisons of functional groups for cold-responsive genes were made against those for all genes in the database using a Fisher exact test in the R software. Microsoft Excel, Access, and the R software were used to perform general analyses. Unless an alternative value is stated the term significant is used to denote an FDR-corrected p < 0.05 for differential expression and an uncorrected p < 0.05 in all other cases.
Supporting Information
Dataset S1 Mapman Visualization of Cold-Responsive Gene Expression
See readme.txt_ in ZIP file for Mapman software details. The Short (S), Medium (M), Long (L), SM, ML, and SML files contain the mean log2 expression changes for the consensus genes regulated at these times. The files with the prefix DRE indicate which of these genes contain either the A/GCCGACNT motif in their 1,000-bp promoter or the A/GCCGAC motif in their 500-bp promoter. The All_AGI file includes all 22,043 genes in our database. The suffixed numbers in our manuscript refer to different pathways that can be used for visualization; 1, Cellular response; 2, Secondary Metabolism; 3, Regulation; 4, Transport; 5, Large enzyme families; 6, Photosynthesis; 7, Transcription; 8, Sucrose-Starch; 9, Glycolysis; 10, TCA; and 11, Metabolism.
(3.6 MB ZIP)
Click here for additional data file.
Table S1 Comparisons between All Independent Datasets Incorporated into Table S2
Comparisons shown are of the agreement (A), the proportion of common genes (B), and the number of common genes (C) between all independent datasets incorporated into Table S2. Data from grouped datasets were averaged. The agreement is the proportion of common cold responsive genes that change in the same direction in both studies. Comparisons within each time class with greater than 0.7, 0.8, and 0.9 agreement are highlighted yellow, orange, and red, respectively. Studies with fewer than ten common genes are in italics and not highlighted.
(59 KB XLS)
Click here for additional data file.
Table S2 Database of Cold-Responsive Genes in Arabidopsis
“AGI” contains a single entry for each AGI locus code, while “AGI-2” shows additional matches for ATH1 probe sets. “Probe set ID” shows the ATH1 probe set IDs that match each locus code, and “Probe set ID-2” shows additional matches where present. “AtGenome” and “MWG” indicate genes represented on these arrays; Y, AGI match; S, AGI-2 match; U, no match to ATH1 array; a blank indicates that the gene is not represented. The “Lit” columns contain numbers that correspond to supporting papers (Table S9), while the other columns contain the log2 change of expression for genes considered to be cold responsive in each study. The log2 nonacclimated signal, log2 change, unadjusted, and FDR adjusted p-values are provided for both our datasets. Datasets are split into the time classes of short, medium, and long term, and highlighted in yellow, blue, and green respectively. The cold responsiveness of each gene was calculated by scoring 1 for up-regulation and −1 for down-regulation in biologically or technically independent studies in each time class. The average log2 fold change within each time class for genes scoring 2 or above or −2 or below is given in the “_up” and “_down” columns, respectively. The mean hydrophilicity, glycine content, and frequencies of the DRE and ABRE promoter elements are also included. Details of the expression profiling studies used can be found in Table S8.
(13 MB XLS)
Click here for additional data file.
Table S3 Statistical Analysis of TAIR AraCyc Pathways and the Table S2 Data for Each Gene
Statistical analysis is presented for (A) TAIR AraCyc pathways and (B) the corresponding entries from Table S2 for each gene. All genes from the AraCyc pathways included in aracyc_dump_20040520 (ftp://ftp.arabidopsis.org/home/tair/Pathways/) were used to extract values for cold responsive genes from Table S2. Enzymes without known AGI locus code matches or duplicate codes within a single pathway were excluded. The numbers of genes within each pathway up- or down-regulated in the short, medium, and long term were counted. The proportion of genes within a pathway that are cold responsive was compared to the proportion of all genes that are cold responsive. The fold change of this comparison (+/−) and a p-value (p-val) from a Fisher exact test are given. Fold changes of 0 and p-values of 1 are not shown, and p-values less than 0.05, 0.01, and 0.001 are highlighted in yellow, orange, and red, respectively.
(977 KB XLS)
Click here for additional data file.
Table S4 Statistical Analysis of Hydrophilicity and Glycine Content of Proteins Encoded by Consensus Cold-Responsive Genes
Hydrophilicity and glycine content were calculated as described [24] and cross-referenced against Table S2 for the all Arabidopsis proteins (ATH1_pep_cm_20040228; ftp://tairpub:[email protected]/home/tair/Sequences/blast_datasets/). Where multiple gene models were predicted only the first was used. The proportion of each group cold responsive was compared to the proportion of all genes within our database that are cold responsive. The fold change of this comparison (+/−) and a p-value (p-val) from a Fisher exact test are given; p-values less than 0.05, 0.01, and 0.001 are highlighted in yellow, orange, and red, respectively. Hydrophilins are defined as having hydrophilicity values over 1 and a glycine content over 0.08 [24].
(21 KB XLS)
Click here for additional data file.
Table S5 Mean Hydrophilicity and Glycine Content of Unknown Proteins that May Be COR/LEA Proteins
Genes annotated as expressed or hypothetical proteins up-regulated in the short, medium, or long term were selected. The mean expression changes during short, medium, and long term cold treatment and the frequencies of the DRE and ABRE cis-elements are also shown. Mean hydrophilicity was calculated using a sliding window of 12 residues as described [24]. Proteins are grouped by their hydrophilicity and glycine content and sorted by AGI code. AGI locus code At2g36220 has been previously reported as COR/LEA-like [6]. Although not unknown, At3g02480 is also included as it has not been previously reported.
(30 KB XLS)
Click here for additional data file.
Table S6 Statistical Analysis of the Cold Regulation of Selected Transcription Factor Families
The proportion of each family that is cold responsive was compared to the proportion of all transcription factors within our database that are cold responsive. The fold change of this comparison (+/−) and a p-value (p-val) from a Fisher exact test are given; p-values less than 0.05, 0.01, and 0.001 are highlighted in yellow, orange, and red, respectively. The numbers of genes in each family and the total numbers of transcription factors regulated are also shown.
(21 KB XLS)
Click here for additional data file.
Table S7
Table S2 Extract for Housekeeping Genes Used in this Study
This table includes all actins, elongation factors, DNA-dependent RNA polymerases and poly-ubiquitins, tubulins not reported as cold responsive [29], and annotated cytosolic cyclophilins. Only one AGI locus code entry was retained for ATH1 probe sets that cross-hybridize to close family members. See Table S2 for more details.
(45 KB XLS)
Click here for additional data file.
Table S8 Details of the Studies Used to Compile Table S2
Datasets are listed in the first row with experimental, analysis, and reference details below. Biologically or technically independent datasets within each time class are grouped. “Growth” indicates the growing medium and “Age” the age or growth stage of plants prior to cold treatment. “Time,” “Photoperiod,” “Light intensity,” and “Temperature” refer to the cold treatment conditions. Details regarding “Array” and “Analysis” can be found in Materials and Methods. In “Replication,” T, technical; B, biological; and −, “hybridized with...”. “Data” provides a link to the dataset used, and “Reference” the relevant publication.
(31 KB XLS)
Click here for additional data file.
Table S9 Publications Detailing Experimental Evidence for Cold-Responsive Gene Expression
The numbers in the “Down” and “Up” columns correspond to those included in the “Lit” columns of Table S2, and a supporting publication is given in the “Reference” column. Genes were identified by AGI locus code, accession number, gene name, or a high-scoring BLAST match. In a few cases where the probes used would cross hybridize, very similar family members were included.
(40 KB XLS)
Click here for additional data file.
Accession Numbers
The Institute for Genomic Research (http://www.tigr.org) AGI locus codes for the genes and proteins discussed in this paper are: ABA-responsive protein similar to KIN1 from Brassica napus (At3g02480), alcohol dehydrogenase (At1g77120), APY2 (At5g18280), AtGolS3 (At1g09350), BEE1 (At1g18400), BEE2 (At4g36540), BEE3 (At1g73830), CBF1 (At4g25490), CBF2 (At4g25470), CBF3 (At4g25480), COL1 (At5g15850), COL2 (At3g02380), COR12 (At4g33550), COR18 (At2g43060), CYP90A1 (At5g05690), CYP90C1 (At4g36380), DREB2B (At3g11020), DWF1 (At3g19820), lactate dehydrogenase (At4g17260), P5CS2 (At3g55610), pyruvate decarboxylase (At4g33070), RAP2.1 (At1g46768), RAP2.6 (At1g43160), RAV1 (At1g13260), RCI2A (At3g05880), RCI2B (At3g05890), STE1 (At3g02580), ZAT10 (At1g27730), and ZAT12 (At5g59820).
We would like to thank Susanne Freund for excellent technical assistance, the RZPD for ATH1 hybridizations, the R and Bioconductor community for software, Jan Hummel for writing the R script to calculate hydrophilicity, and the Mapman team for generous help. This work was supported by the Max-Planck-Institut für Molekulare Pflanzenphysiologie.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. MAH, AGH, and DKH conceived and designed the experiments. MAH analyzed the data. MAH, AGH, and DKH wrote the paper.
Abbreviations
ABAabscisic acid
ABREabscisic acid response element
AGI
Arabidopsis Genome Initiative
AP2/EREBPAPETALA2/ethylene response element binding protein
ARFauxin response factor
AtGenome
Arabidopsis Genome Array
ATH1Affymetrix ATH1 genome array
bHLHbasic helix-loop-helix
BRbrassino-steroid
CBFC-repeat binding factor
COCONSTANS
COLCONSTANS-LIKE
CORcold-regulated
DREBdehydration-responsive element binding
FDRfalse discovery rate
HSFheat shock transcription factor
LEAlate embryogenesis abundant
MAS5Microarray Suite 5
MWGMWG Biotech 25k Arabidopsis 50-mer oligonucleotide array
TAIRThe Arabidopsis Information Resource
TCAtricarboxylic acid
==== Refs
References
Xin Z Browse J 2000 Cold comfort farm: The acclimation of plants to freezing temperatures Plant Cell Environ 23 893 902
Pearce RS 1999 Molecular analysis of acclimation to cold Plant Growth Regul 29 47 76
Thomashow MF 1999 Plant cold acclimation: Freezing tolerance genes and regulatory mechanisms Annu Rev Plant Physiol Plant Mol Biol 50 571 599 15012220
Chinnusamy V Ohta M Kanrar S Lee BH Hong XH 2003 ICE1: A regulator of cold-induced transcriptome and freezing tolerance in Arabidopsis
Gene Dev 17 1043 1054 12672693
Jung SH Lee JY Lee DH 2003 Use of SAGE technology to reveal changes in gene expression in Arabidopsis leaves undergoing cold stress Plant Mol Biol 52 553 567 12956526
Fowler S Thomashow MF 2002
Arabidopsis transcriptome profiling indicates that multiple regulatory pathways are activated during cold acclimation in addition to the CBF cold response pathway Plant Cell 14 1675 1690 12172015
Kreps JA Wu YJ Chang HS Zhu T Wang X 2002 Transcriptome changes for Arabidopsis in response to salt, osmotic, and cold stress Plant Physiol 130 2129 2141 12481097
Seki M Narusaka M Ishida J Nanjo T Fujita M 2002 Monitoring the expression profiles of 7000 Arabidopsis genes under drought, cold and high-salinity stresses using a full-length cDNA microarray Plant J 31 279 292 12164808
Vogel JT Zarka DG Van Buskirk HA Fowler SG Thomashow MF 2005 Roles of the CBF2 and ZAT12 transcription factors in configuring the low temperature transcriptome of Arabidopsis
Plant J 41 195 211 15634197
Shinozaki K Yamaguchi-Shinozaki K Seki M 2003 Regulatory network of gene expression in the drought and cold stress responses Curr Opin Plant Biol 6 410 417 12972040
Gilmour SJ Fowler SG Thomashow MF 2004
Arabidopsis transcriptional activators CBF1, CBF2, and CBF3 have matching functional activities Plant Mol Biol 54 767 781 15356394
Maruyama K Sakuma Y Kasuga M Ito Y Seki M 2004 Identification of cold-inducible downstream genes of the Arabidopsis DREB1A/CBF3 transcriptional factor using two microarray systems Plant J 38 982 993 15165189
Cook D Fowler S Fiehn O Thomashow MF 2004 A prominent role for the CBF cold response pathway in configuring the low-temperature metabolome of Arabidopsis
Proc Natl Acad Sci USA 101 15243 15248 15383661
Novillo F Alonso JM Ecker JR Salinas J 2004 CBF2/DREB1C is a negative regulator of CBF1/DREB1B and CBF3/DREB1A expression and plays a central role in stress tolerance in Arabidopsis
Proc Natl Acad Sci USA 101 3985 3990 15004278
Lee H Xiong L Gong Z Ishitani M Stevenson B 2001 The Arabidopsis HOS1 gene negatively regulates cold signal transduction and encodes a RING finger protein that displays cold-regulated nucleo-cytoplasmic partitioning Genes Dev 15 912 924 11297514
Zhu J Shi H Lee BH Damsz B Cheng S 2004 An Arabidopsis homeodomain transcription factor gene, HOS9 , mediates cold tolerance through a CBF-independent pathway Proc Natl Acad Sci USA 101 9873 9878 15205481
Redman JC Haas BJ Tanimoto G Town CD 2004 Development and evaluation of an Arabidopsis whole genome Affymetrix probe array Plant J 38 545 561 15086809
Yauk CL Berndt ML Williams A Douglas GR 2004 Comprehensive comparison of six microarray technologies Nucl Acids Res 32 e124 15333675
Pounds S Morris SW 2003 Estimating the occurrence of false positives and false negatives in microarray studies by approximating and partitioning the empirical distribution of p-values Bioinformatics 19 1236 1242 12835267
Benjamini Y Hochberg Y 1995 Controlling the false discovery rate—A practical and powerful approach to multiple testing J Roy Stat Soc B Met 57 289 300
Reiner A Yekutieli D Benjamini Y 2003 Identifying differentially expressed genes using false discovery rate controlling procedures Bioinformatics 19 368 375 12584122
Thimm O Blasing O Gibon Y Nagel A Meyer S 2004 MAPMAN: A user-driven tool to display genomics data sets onto diagrams of metabolic pathways and other biological processes Plant J 37 914 939 14996223
Usadel B Nagel A Thimm O Redestig H Blasing OE 2005 Extension of the visualisation tool MapMan to allow statistical analysis of arrays, display of co-responding genes and comparison with known responses Plant Physiol 138 1195 1204 16009995
Garay-Arroyo A Colmenero-Flores JM Garciarrubio A Covarrubias AA 2000 Highly hydrophilic proteins in prokaryotes and eukaryotes are common during conditions of water deficit J Biol Chem 275 5668 5674 10681550
Capel J Jarillo JA Salinas J Martinez-Zapater JM 1997 Two homologous low-temperature-inducible genes from Arabidopsis encode highly hydrophobic proteins Plant Physiol 115 569 576 9342870
Liu Q Kasuga M Sakuma Y Abe H Miura S 1998 Two transcription factors, DREB1 and DREB2, with an EREBP/AP2 DNA binding domain separate two cellular signal transduction pathways in drought- and low-temperature-responsive gene expression, respectively, in Arabidopsis
Plant Cell 10 1391 1406 9707537
Stockinger EJ Gilmour SJ Thomashow MF 1997
Arabidopsis thaliana CBF1 encodes an AP2 domain-containing transcriptional activator that binds to the C-repeat/DRE, a cis-acting DNA regulatory element that stimulates transcription in response to low temperature and water deficit Proc Natl Acad Sci U S A 94 1035 1040 9023378
van de Peppel J Kemmeren P van Bakel H Radonjic M van Leenen D 2003 Monitoring global messenger RNA changes in externally controlled microarray experiments EMBO Rep 4 387 393 12671682
Chu B Snustad DP Carter JV 1993 Alteration of [beta]-Tubulin gene expression during low-temperature exposure in leaves of Arabidopsis thaliana
Plant Physiol 103 371 377 12231943
Allemeersch J Durinck S Vanderhaeghen R Alard P Maes R 2005 Benchmarking the CATMA Microarray: A novel tool for Arabidopsis transcriptome analysis Plant Physiol 137 588 601 15710687
Cui X Churchill GA 2003 Statistical tests for differential expression in cDNA microarray experiments Genome Biol 4 210 12702200
Smyth GK 2004 Linear models and empirical Bayes for assessing differential expression in microarray experiments Stat App Gen Mol Biol 3 Article 3.
Jarvinen AK Hautaniemi S Edgren H Auvinen P Saarela J 2004 Are data from different gene expression microarray platforms comparable? Genomics 83 1164 1168 15177569
Kaplan F Kopka J Haskell DW Zhao W Schiller KC 2004 Exploring the temperature-stress metabolome of Arabidopsis
Plant Physiol 136 4159 4168 15557093
Klotke J Kopka J Gatzke N Heyer AG 2004 Impact of soluble sugar concentrations on the acquisition of freezing tolerance in accessions of Arabidopsis thaliana with contrasting cold adaptation—Evidence for a role of raffinose in cold acclimation Plant Cell Environ 27 1395 1404
Zuther E Buchel K Hundertmark M Stitt M Hincha DK 2004 The role of raffinose in the cold acclimation response of Arabidopsis thaliana
FEBS Lett 576 169 173 15474032
Uemura M Joseph RA Steponkus PL 1995 Cold acclimation of Arabidopsis thaliana (effect on plasma membrane lipid composition and freeze-induced lesions) Plant Physiol 109 15 30 12228580
Fukuchi-Mizutani M Tasaka Y Tanaka Y Ashikari T Kusumi T 1998 Characterization of delta 9 acyl-lipid desaturase homologues from Arabidopsis thaliana
Plant Cell Physiol 39 247 253 9559566
Gibson S Arondel V Iba K Somerville C 1994 Cloning of a temperature-regulated gene encoding a chloroplast omega-3 desaturase from Arabidopsis thaliana
Plant Physiol 106 1615 1621 7846164
Gracey AY Fraser EJ Li W Fang Y Taylor RR 2004 Coping with cold: An integrative, multitissue analysis of the transcriptome of a poikilothermic vertebrate Proc Natl Acad Sci U S A 101 16970 16975 15550548
Savitch LV Barker-Astrom J Ivanov AG Hurry V Oquist G 2001 Cold acclimation of Arabidopsis thaliana results in incomplete recovery of photosynthetic capacity, associated with an increased reduction of the chloroplast stroma Planta 214 295 303 11800395
Hurry V Strand A Furbank R Stitt M 2000 The role of inorganic phosphate in the development of freezing tolerance and the acclimatization of photosynthesis to low temperature is revealed by the pho mutants of Arabidopsis thaliana
Plant J 24 383 396 11069711
Winkel-Shirley B 2002 Biosynthesis of flavonoids and effects of stress Curr Opin Plant Biol 5 218 223 11960739
Huner NPA Oquist G Sarhan F 1998 Energy balance and acclimation to light and cold Trends Plant Sci 3 224 230
Puhakainen T Hess MW Makela P Svensson J Heino P 2004 Overexpression of multiple dehydrin genes enhances tolerance to freezing stress in Arabidopsis
Plant Mol Biol 54 743 753 15356392
Steponkus PL Uemura M Joseph RA Gilmour SJ Thomashow MF 1998 Mode of action of the COR15a gene on the freezing tolerance of Arabidopsis thaliana
Proc Natl Acad Sci USA 95 14570 14575 9826741
Koag MC Fenton RD Wilkens S Close TJ 2003 The binding of maize DHN1 to lipid vesicles. Gain of structure and lipid specificity Plant Physiol 131 309 316 12529538
Kazuoka T Oeda K 1994 Purification and characterization of Cor85-oligomeric complex from cold-acclimated spinach Plant Cell Physiol 35 601 611
Houde M Daniel C Lachapelle M Allard F Laliberte S 1995 Immunolocalization of freezing-tolerance-associated proteins in the cytoplasm and nucleoplasm of wheat crown tissues Plant J 8 583 593 7496403
Grelet J Benamar A Teyssier E Avelange-Macherel M-H Grunwald D 2005 Identification in pea seed mitochondria of a late-embryogenesis abundant protein able to protect enzymes from drying Plant Physiol 137 157 167 15618423
Wolkers WF McCready S Brandt WF Lindsey GG Hoekstra FA 2001 Isolation and characterization of a D-7 LEA protein from pollen that stabilizes glasses in vitro Biochim Biophys Acta 1544 196 206 11341929
Xu D Duan X Wang B Hong B Ho T 1996 Expression of a late embryogenesis abundant protein gene, HVA1 , from Barley confers tolerance to water deficit and salt stress in transgenic Rice Plant Physiol 110 249 257 12226181
Dunaeva M Adamska I 2001 Identification of genes expressed in response to light stress in leaves of Arabidopsis thaliana using RNA differential display Eur J Biochem 268 5521 5529 11683875
Ledger S Strayer C Ashton F Kay SA Putterill J 2001 Analysis of the function of two circadian-regulated CONSTANS-LIKE genes Plant J 26 15 22 11359606
Toledo-Ortiz G Huq E Quail PH 2003 The Arabidopsis basic/helix-loop-helix transcription factor family Plant Cell 15 1749 1770 12897250
Friedrichsen DM Nemhauser J Muramitsu T Maloof JN Alonso J 2002 Three redundant brassinosteroid early response genes encode putative bHLH transcription factors required for normal growth Genetics 162 1445 1456 12454087
Zimmermann P Hirsch-Hoffmann M Hennig L Gruissem W 2004 GENEVESTIGATOR. Arabidopsis microarray database and analysis toolbox Plant Physiol 136 2621 2632 15375207
Dharmasiri N Estelle M 2004 Auxin signaling and regulated protein degradation Trends Plant Sci 9 302 308 15165562
Nemhauser JL Mockler TC Chory J 2004 Interdependency of brassinosteroid and auxin signaling in Arabidopsis
PLoS Biology 2 e258 15328536
Goda H Sawa S Asami T Fujioka S Shimada Y 2004 Comprehensive comparison of auxin-regulated and brassinosteroid-regulated genes in Arabidopsis
Plant Physiol 134 1555 1573 15047898
Kim G-T Fujioka S Kozuka T Tax FE Takatsuto S 2005 CYP90C1 and CYP90D1 are involved in different steps in the brassinosteroid biosynthesis pathway in Arabidopsis thaliana
Plant J 41 710 721 15703058
Sasse JM 2003 Physiological actions of brassinosteroids: an update J Plant Growth Regul 22 276 288 14676971
Torjek O Berger D Meyer RC Mussig C Schmid KJ 2003 Establishment of a high-efficiency SNP-based framework marker set for Arabidopsis
Plant J 36 122 140 12974817
Rohde P Hincha DK Heyer AG 2004 Heterosis in the freezing tolerance of crosses between two Arabidopsis thaliana accessions (Columbia-0 and C24) that show differences in non-acclimated and acclimated freezing tolerance Plant J 38 790 799 15144380
Gentleman RC Carey VJ Bates DM Bolstad B Dettling M 2004 Bioconductor: Open software development for computational biology and bioinformatics Genome Biol 5 R80 15461798
Gautier L Cope L Bolstad BM Irizarry RA 2004 Affy—Analysis of Affymetrix GeneChip data at the probe level Bioinformatics 20 307 315 14960456
Wu ZJ Irizarry RA Gentleman R Martinez-Murillo F Spencer F 2004 A model-based background adjustment for oligonucleotide expression arrays J Am Stat Assoc 99 909 917
|
16121258
|
PMC1189076
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Aug 19; 1(2):e26
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010026
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1551596702210.1186/1471-2105-6-155Methodology ArticleIterative approach to model identification of biological networks Gadkar Kapil G [email protected] Rudiyanto [email protected] Francis J [email protected] Department of Chemical Engineering, University of California Santa Barbara, CA, USA2005 20 6 2005 6 155 155 3 2 2005 20 6 2005 Copyright © 2005 Gadkar et al; licensee BioMed Central Ltd.2005Gadkar et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Recent advances in molecular biology techniques provide an opportunity for developing detailed mathematical models of biological processes. An iterative scheme is introduced for model identification using available system knowledge and experimental measurements.
Results
The scheme includes a state regulator algorithm that provides estimates of all system unknowns (concentrations of the system components and the reaction rates of their inter-conversion). The full system information is used for estimation of the model parameters. An optimal experiment design using the parameter identifiability and D-optimality criteria is formulated to provide "rich" experimental data for maximizing the accuracy of the parameter estimates in subsequent iterations. The importance of model identifiability tests for optimal measurement selection is also considered. The iterative scheme is tested on a model for the caspase function in apoptosis where it is demonstrated that model accuracy improves with each iteration. Optimal experiment design was determined to be critical for model identification.
Conclusion
The proposed algorithm has general application to modeling a wide range of cellular processes, which include gene regulation networks, signal transduction and metabolic networks.
==== Body
Background
A systems level understanding of highly complex biological systems requires an integration of experimental techniques and computational research [1]. Current molecular biology techniques can generate high-throughput quantitative data that support in silico research using mathematical models [2]. These models can be used to simulate and study the dynamic interactions among the components of cellular systems as well as the systems' responses to external perturbations and signals. Such tools offer enormous potential for understanding cellular functions at the organism level [3]. Mathematical models also serve as test beds for generating hypotheses and designing experiments to test them [4]. Furthermore, they provide bases for model-based product and process design applications. Useful insights and predictions have been obtained for several biological systems from computational modeling and analysis. A few examples include the metabolic network analysis of Escherichia coli growth on glucose and acetate [5], the MAP kinase signaling pathways [6] and caspase function in apoptosis [7], and bifurcation analysis of cell cycle in Saccharomyces cerevesiae [8].
An iterative process for model development and the testing of hypotheses has been proposed by many researchers in the field and was recently highlighted by Kitano [1]. A qualitative approach of this process is described in [2]. In addition, Rabitz and co-workers [9] have recently developed an iterative method for closed loop parameter identification in biochemical reaction networks. A global inversion algorithm was used to extract the parameter estimates that minimize the differences between model prediction and experimental data. Unfortunately, global search methods typically have high computational requirements, and thus, do not scale very well with the system size. In this work, a quantitative model identification is developed that effciently obtains parameter estimates and facilitates scalability to very large network sizes. A proposed strategy is described in Section 2, with an emphasis on the modeling element. The modeling strategy is decomposed into three main steps: (1) determining the connectivity of the biological network and the interactions of the sub-components, (2) formulating the kinetics of inter-conversion among the subcomponents, and (3) estimating the parameters in the rate equations. To the authors' knowledge, this work represents the first documented example of multiple iterations for model refinement using such a framework in systems biology.
The parameter estimation from experimental data remains the bottleneck in the model development [4]. Banga and coworkers [10] have compared several advanced deterministic and stochastic global optimization methods for parameter identification from available experimental data. It was observed that the traditional gradient-based optimization methods often failed to arrive at the global optimal solutions. Deterministic methods [11-13] can achieve global optimality for certain classes of problems, but there is no guarantee of convergence in finite time [14]. Stochastic strategies [14-16] can locate the parameter region containing the global solution with relatively better effciency, but global optimality is not guaranteed. Furthermore, both methods suffer from the large computational burden required, even for moderately sized problems. Moreover, the validity of model with the estimated parameters over the entire operating space remains to be determined.
Parameter identifiability tests should be performed prior to the estimation process to ensure that the parameter estimation problem is well-posed. Further, the identifiability tests assist in selection of optimal measurements. Several researchers [17-19] have developed methods to determine whether a parameter is "identifiable a priori", i.e., identifiable from a given experiment design using the available measurements. A similar concept known as "practical identifiability" is concerned with the achievable accuracy of the parameter estimates. The confidence interval for the model parameters are determined using the Fisher Information Matrix (FIM) [20,21]. Doyle and coworkers [22] have performed model identifiability studies for a gene regulatory network using gene expression data, in which the identifiability of the parameters was found to be strongly dependent on the driving function.
The final step in the iterative model development process is the design of "optimal experiments" that would provide rich experimental data for improving the parameter estimates. Experiments can also be designed for discrimination among competing model structures that translates to selection between multiple proposed mechanisms of cellular function. Asprey and Macchietto [23] have developed a strategy of optimal experiment design for model structural identifiability. The strategy was used to identify the kinetics of the reactions in the fermentation of Saccharomyces cerevesiae. Kremling and co-workers [24] propose several strategies for model discrimination to identify the correct reaction mechanism of a test metabolic network. Banga and coworkers [25] have formulated the optimal design problem, using a scalar function of the Fisher Information Matrix (FIM) as the performance index, for parameter estimation of nonlinear dynamic systems.
In this work, an iterative procedure for model identification is proposed and applied to the caspase-dependent apoptosis system. An optimal measurement set is determined using the Fisher Information Matrix (FIM). The parameter estimation from partial measurements is decoupled into two parts. First, the available measurements are used to estimate the profiles of all unmeasured concentrations and reaction rates using a State Regulator Problem (SRP) formulation. In the second part the concentration and rate estimates are used to determine the model parameter values. The SRP formulation in this work is an extension of the dynamic Flux Balance Analysis (dFBA) approach developed by Doyle and coworkers [26]. Finally, the model-based experiment design uses parameter identifiability and D-optimality criteria to obtain the optimal experimental procedure that would generate the most informative data for model refinement in the next iteration.
Results
The iterative scheme for model identification is shown in Figure 1. The optimal set of measurements is determined a priori. For an effcient model identification, a significant fraction of the unknown model parameters should be identifiable. In the case of poor identifiability, a higher number of measurements would be motivated. Also, the model complexity could be reduced to decrease the number of parameters; but this does not guarantee identifiability of the reduced number of parameters. Alternatively, a richer protocol (e.g. perturbation sequence [22]) might yield improved identifiability. In this work, selection of the a priori optimal measurement set is restricted to the "preliminary" experiment design that may be suboptimal.
Figure 1 Iterative scheme for model identification.
The model connectivity and reaction mechanisms are developed from existing biological knowledge and are assumed to be known. The network connectivity, along with the partial measurements (optimal), is used in the State Regulator Problem (SRP) to obtain estimates of all system unknowns (unmeasured concentrations and reaction rates). Here it is important to note that the kinetics of the reaction rates are not used in the SRP algorithm. Next, the full estimates of the concentrations and the reaction rates are used for estimating the parameters in the kinetic model. This decouples the model identification into two parts such that the parameters involved in the kinetic equation of each reaction are independently determined as opposed to simultaneously estimating full model parameters from limited measurements. Next, the model invalidation test, which is a critical step in model development and the last "quality control" step before the desired application [27-29], is performed. Invalidity of the model could be determined by comparing model predictions with experimental data that is not used in the SRP algorithm. Further, model invalidity can occur if the model predictions conflict with documented biological knowledge of the system. In case of an invalid model, the model parameters are refined in subsequent iterations using the information obtained from the optimal experiment or by expanding the measurement set. The process of model identification is repeated in an iterative manner until an "acceptable" model is obtained.
System
The mathematical model considered in this paper has the following structure:
= Ax + Br + C (1)
where
r = f(x, p), (2)
x represents the protein/metabolite/gene concentrations, r the reaction rates, and p the model parameters. This is a very general nonlinear state space model in the variable x. The matrices A and C describe degradation and auto-generation respectively, whereas the matrix B represents the stoichiometry of the biological network. The kinetics among the proteins/metabolites/genes interactions show up in the reaction rates r. The aforementioned model is a continuous time invariant affine system from which a discrete version can be derived by standard techniques using a zero-order hold [30]. The resulting discrete model equation is represented as:
where
The discrete version of the model is used in the SRP estimation algorithm.
Theory
Step 1: Determination of measurement set
The optimal measurement set consists of species whose concentration measurements would have maximum benefit for model identification, e.g, parameter identifiability and accuracy. In this work, the measurement set is determined such that the model parameters can be estimated accurately. The assessment of parameter identifiability in a model is crucial prior to parameter estimation from experimental data [31]. Identifiability is closely linked with parametric sensitivity analysis through the Fisher Information Matrix (FIM) [27]. The unidentifiable parameters are determined using the orthogonal procedure proposed by MacAuley and coworkers [19]. Here, a scaled sensitivity coeffcient matrix () shown below is computed:
where {p1, …, pk } are the model parameters, {η1, …, ηm } are the response variables which include all possible measurable quantities, {t1, t2, …, tN} are the sampling times for the measurements, and is the "initial" parameter value that is either the guess values of the parameter or the value obtained from literature. The orthogonal method is a geometric based approach where the number of identifiable parameters correlates with the rank of the orthogonalization of the scaled sensitivity matrix. The parameters corresponding to the columns of orthogonalized sensitivity matrix are deemed unindentifiable if the norms are smaller than a given tolerance. The details of the orthogonal method are not included here for the sake of brevity.
The next step is to obtain a measurement set that maximizes the expected accuracy in the identifiable parameters (practical identifiability). The Fisher Information Matrix (FIM) along with the Cramer-Rao theorem are used to determine a measurement set such that the estimated parameters have minimum variance. A detailed description of the procedure and its theoretical foundation can be found in [32]. Assuming that the measurement errors are additive and Gaussian, the FIM is given by [33]:
FIM = JTWJ (5)
where W is the inverse of the measurement error covariance matrix and J denotes the sensitivity coefficient matrix for the measured response variables:
The quantity {p1, …, pr} denotes the identifiable parameter vector and {, …, } denotes the measured response variable vector.
The Cramer-Rao inequality establishes a lower bound on the variance of the identifiable parameters given by:
σ2(pi) ≥ (FIM-1)ii (7)
The 95% confidence interval (CI) for a parameter is given by:
CI = ± 1.96σ(pi) (8)
In Equation (8) the lower bound of the variance is used. Symmetry of the confidence region about the nominal value is assumed. This results in the following definition of the percentage deviation from the nominal value:
The optimal measurement set is chosen such that the sum of the percentage error (E) for all the identifiable parameters is minimized. In this work, the optimal set is determined by a brute-force search over all combinations of measurement sets subject to restrictions that may be imposed by the system. Doyle and co-workers have developed effcient rational algorithms to determine the optimal measurement set with minimum computational burden [34]. The confidence intervals for non-identifiable parameters are infinitely large and hence are eliminated from the analysis. Identifiability for these parameters can be obtained only by a change in the experimental design or by the selection of an alternative model structure.
Step 2: State Estimation Algorithm
Generally, it is not possible to measure all time-varying components in a metabolic or signaling network. However, there are several techniques from systems engineering to estimate the behavior of unmeasured components given partial measurements of other system constituents. Bastin and Dochain have used an adaptive nonlinear observer for estimation of specific growth rate and biomass concentration [35]. Given accurate models, Extended Kalman Filters (EKF) have had success in several biological applications [36-38]. Artificial Neural Networks (ANN) have also found applications where dynamic models are not available [39-41].
In this work, an extension of Dynamic Flux Balance Analysis (dFBA) [26] is developed to estimate unmeasured concentration and reaction rate trajectories given partial measurement sets. The premise of this approach is straightforward: cellular processes have evolved regulatory structures that optimally use cellular resources. This premise translates into two postulates; (1) network flows are managed to minimize internal accumulation and (2) networks are managed to minimize the number of edges carrying flux at any given time. These two requirements are analogous to a classic problem in automatic control, namely, the State Regulator Problem (SRP). The SRP based estimator uses the measurement set selected from Step 1 to estimate unknown concentration and reaction rate trajectories via a constrained convex programming problem. The SRP estimator constrained by the key measurements captures the optimal cellular behavior of the system.
Estimates of the reaction rates at time step k and protein/gene/metabolite concentrations at time step k + 1 are determined by the SRP and the discrete mass balance equations. The SRP must be solved at each sampling interval to obtain estimates of the unknown rates and concentrations. Consider the model (Equation 3); the discrete mass balance equations over a p step horizon are given by:
Where the matrices k+1, k, x and c are defined as
and the matrix r is given by
The SRP estimator is a Quadratic Program (QP) with two cost terms, the cost of intermediate accumulation and the cost of operating a network reaction. The SRP penalizes intracellular metabolite or protein accumulation, but does not explicitly forbid it. Moreover, because reactions introduce an additional cost, the SRP only utilizes those reactions required to satisfy the mass balances and thermodynamic constraints. Formally, the estimation problem is given by:
subject to:
k + 1 ≥ 0 (12)
αr (k) ≤ k ≤ βr (k) (13)
The SRP problem is subject to non-negativity constraints (Equation 12), flux-directionality constraints (Equation 13) and constraints imposed by the measurement set. Specifically, constraint of Equation 14 forces state estimates belonging to the measurement set to equal the corresponding measured value. The quantities denote measurements that may have been corrupted with noise. specifies the tolerance around the measurement within which the estimate is constrained to lie (incorporated to avoid numerical inconsistencies that may arise due to noisy measurements). The term ΞX defines the measurement set. The matrix Wx defines the cost of intermediate accumulation whereas the matrix WR represents the reaction cost:
For the caspase system considered in this work, the wx and wr are taken to be the order of magnitude of the inverse of the maximum value of the corresponding state or rate. This requires only approximate information regarding ranges of the protein concentrations and identification of the slow versus the fast reactions:
Step 3: Parameter estimation
The estimates of the concentration profiles and the reaction rates allow efficient determination of the parameter values by decoupling the full parameter estimation into multiple sets. Each set consists of parameters associated with one reaction rate. The parameters are obtained by minimizing the difference between the estimates of each reaction rate and that predicted by the kinetics r (x, p), which is a function of the concentrations. In case of the first iteration the minimization follows:
where ri are the individual reaction rates and NR is the total number of reactions. The kinetic parameters associated with a reaction rate equation are determined independently from those with other reactions, i.e., the parameter estimation is decoupled with respect to each reaction.
For subsequent iterations, the Bayesian estimation formulation in [42] is used. In this formulation, in addition to the difference between the estimates of the reaction rates and the model predictions, the deviations of parameter values from those obtained after the previous iteration are minimized. The formulation can be represented as:
In the above equations, is the estimate of the ith reaction rate and is the estimates of the concentrations obtained from the SRP algorithm, ri(, p) is the predicted rate of the ith reaction from the kinetics in Equation 2, p are the parameters associated with the ith reaction rate, p0 are the parameter values obtained from the previous iteration, and Vε and Vp are the variances of the estimates of the reaction rates and the prior parameter estimates. The parameter variances are determined using the Fisher Information Matrix (Equation 7). The variances of the non-identifiable parameters are infinite and penalty for deviations for these parameters are not considered in Equation 19. The variance for the estimates of the reaction rates can be determined from the expected noise in the measurements from which the estimates are obtained.
Step 4: Model invalidation tests
Given the iterative nature of this framework, a termination criterion must be established. Poolla et al. [29] have shown that for certain experimental data, it is not possible to confirm whether the model is really valid; however, one can conclude whether the model is not contradicted by the given data. Model (in)validation tests are usually based on the difference between the simulated and measured output and some statistics about these differences. Typical statistics for the model errors include maximum absolute value, mean value and variance [28]. In this work, model invalidity is tested by determining the model prediction errors using the estimated parameters. This error is calculated as:
To implement this test, experimental data that was not used in the SRP algorithm is required. The statistic used is the maximum and mean value of the errors for the measured states. When the prediction errors are below a certain desired value, the iterations are terminated.
Step 5: Model-based optimal experiment design
The optimal experiment design determines the optimal experiment to be performed for the next iteration such that there is maximum information content in the measurements. This would maximize the accuracy of the estimated parameters. The model-based optimal experiment design uses the Fisher Information Matrix as a measure of the amount of information contained in a given set of measurements about the model parameters [43]. The optimization searches through the space of experimental conditions or some parameterizations of the experimental protocol. For example, an optimal ligand input can be parameterized into a time series profile such that the optimization variables are the levels of ligand at different times (usually equally spaced in time). Naturally, the optimization will be restricted by the limitations in the experimental conditions and apparatus.
There exist several FIM-based optimality measures that quantify the overall informativeness of the measurements [32]. Among these, parameter identifiability and the D-optimality are the most widely used measures. For accurate model identification it is critical that maximum number of the parameters be estimated accurately. Thus, maximizing the number of identifiable parameters is the primary criterion proposed for determining the next experimental design. The orthogonal procedure proposed by McAuley and co-workers [19] is used to determine the number of identifiable parameters. There can be multiple experimental designs with the same maximum number of identifiable parameters. The selection among these is done so as to maximize the informativeness of measurement data. For this purpose, the D-optimality criteria is proposed. The use of D-optimality translates to minimizing the confidence interval of all the identifiable parameter estimates. The optimal experiment design criterion is shown as follows:
where E denote the feasible experimental conditions (defined by constraints in experiments), FIM is given in Equation 5 and r denotes the number of identifiable parameters. Thus, the identifiability is maximized in the sense that the hyper-dimensional confidence interval is minimized. For experimental designs with the maximum number of identifiable parameters, the one with the highest determinant of the Fisher Information Matrix is selected as the optimal design for the next experiment.
A case study
The proposed iterative model identification is applied to the function of caspase-8 and caspase-9 in apoptosis. Caspase enzymes are at the core of the cell's suicide machinery. These enzymes are activated either by an external signal or by stress, and activated enzymes will then dismantle the cells. Varner and co-workers have developed a model for the caspase function in apoptosis [7]. The model describes the key elements of receptor-mediated and stress-induced caspase activations. The model consists of 19 states (enzymes) and 11 reactions with 27 parameters (11 rate constants and 16 saturation constants; see Appendix for additional details of the model). The in silico experiments in this study use the Varner model as the "actual" system. Measurements are assumed to be obtained from this "actual" system corrupted with up to 10% noise. The iteration starts using a "initial" parameter set, generated by perturbing the parameter values of the Varner model (considered as "exact") by 70–100%. The external and stress signals that activate the caspase system are considered as the manipulated variables in Step 5. Model refinement is performed either by determination of an optimal experiment or by optimal refinement of the measurement set. The performance of each of these criteria for model refinement requires to be tested. Therefore, in the first iteration, a "preliminary" suboptimal experiment with a suboptimal measurement set is considered. Moreover, in most cases of model identification, it is expected that preliminary experimental data is available. This data is usually obtained from a suboptimal experiment design and does not include the optimal set of measurements. The second iteration is performed in two ways; one by improving the measurement set and second by improving the experiment design. This tests the performance of both refinement criteria. The sequence of events is shown in Figure 2. It should be noted that it would be best to use optimal experiment design along with the optimal measurement set. However, it may not always be possible to do so due to feasibility issues specific to the particular system. Hence, this approach is not considered in this work. Model identification under less constrained conditions using a similar framework is included in [44].
Figure 2 Sequence of simulations for model identification to test efficiency of optimal experiment design and optimal measurement selection.
After the first iteration, it is observed that there is a significant improvement in the predictions with the estimated parameters for both the protein concentrations and the reaction rates, as shown in Figures 3 and 4. The high errors with the "initial" parameters demonstrate that there is no bias in the results based on the starting guess values for the parameters and that there is indeed an improvement in prediction of both the protein concentrations and the reaction rates. However, the improvement is not suffcient as observed from the invalidation test (see Methods). This warrants a second iteration.
Figure 3 Prediction profiles of the 19 protein concentrations for the test experiment for the caspase system. Solid line: real system; dashed line: prediction with estimated parameters after first iteration (suboptimal experiment with suboptimal measurements); dash-dotted line: prediction with estimated parameters after second iteration (suboptimal experiment with optimal measurements); dotted line: prediction with estimated parameters after second iteration (optimal experiment with suboptimal measurements)
Figure 4 Prediction profiles of the 11 reaction rates for the test experiment for the caspase system. Solid line: real system; dashed line: prediction with estimated parameters after first iteration (suboptimal experiment with suboptimal measurements); dash-dotted line: prediction with estimated parameters after second iteration (suboptimal experiment with optimal measurements); dotted line: prediction with estimated parameters after second iteration (optimal experiment with suboptimal measurements)
In general, the model predictions improve with the second iteration, as shown again in Figures 3 and 4. However, it is observed that the predictions are better for the case with the optimal experiment design, in spite of a suboptimal measurement set. This is due to the fact that model performance depends strongly on the accuracy of the estimated parameters. Using the suboptimal experiment, only 14 of the 27 parameters were identifiable. The optimal measurement set simply improved the confidence in these 14 parameters. On the other hand, the optimal experiment increased parameter identifiability to 18 parameters even with the suboptimal measurement set. These result indicates that performance of model identification is strongly linked with parameter identifiability.
Discussion
In the proposed algorithm, the network topology and the mechanism of interactions in the pathway are assumed to be known. Several approaches have been proposed in literature to determine the network connectivity from experimental data [45-47]. In the case of unknown connectivity, these approaches should be used prior to the proposed model identification. The mismatch between the model and the actual network can appear in two different aspects of the algorithm. First, the SRP step can fail because there exist no feasible state and flux estimates that satisfy the measurement constraints. Such scenario arises mainly due to the network topology mismatch, and has low probability to occur due to the large degree of freedoms in a typical biological system. Alternatively, a well-designed model (in)validation step catches the mismatch between the model and the real system, e.g., an independent measurement set contradicts the model prediction. Also, if multiple models are proposed for a particular biological process, model discrimination methods [24,48] can be used to identify the correct model structure. The effect of incorrect connectivity and/or mechanism would depend on the degree of mismatch and is case dependent.
The fact that cellular processes are carried out in an optimal manner lends tremendous promise to the success of this approach. In case the assumed optimality does not represent the in vivo behavior, the estimates from SRP may be inaccurate. However, the measurements (through the constraints in Equation 14) can attenuate this problem by restricting part of the estimates to match the observations. The parameter estimates may also be inaccurate if the real system deviates considerably from the assumed optimal behavior and this deviation is not captured by the measurements. The model identification framework has applicability to all systems that could be represented in the form shown in Equations 1 and 2. All biological processes are complex, interconnected networks. A feature common to these processes is that they have a fixed connectivity. The proposed algorithm for model development could be applied to metabolic networks, signaling processes and gene networks.
The computation burden for solving the SRP is minimal as it involves only a quadratic programming problem. The parameter estimation is also not computationally intensive due to the decoupling facilitated by SRP. A global optimization algorithm can also be used in the parameter estimation instead of the gradient search method to avoid convergence to local minima. However, the computation burden of parameter estimation will increase. To avoid local minima and high computation cost, the first few iterates (1 or 2) can utilize a global optimization method, while the remaining iterations can implement a gradient search algorithm. Due to the iterative nature of the approach, errors in parameter estimates can be tolerated as the corrections will be made in the next iteration with the optimal experiment. The optimal measurement selection is performed by a brute force search in this work. For very large systems the computation burden for this process grows exponentially. Computationally efficient algorithms for optimal measurement selection [34] can be used. Efficient measurement selection algorithms and the decoupling of parameter estimation for individual reaction rates into separate optimization problems result in good scalability properties of the proposed algorithm for large scale systems. One limitation of the approach is that it is dependent on the weights in Equation 15 for the minimization of the cellular resources. The choice of weights used in this work has provided accurate results but it requires information of the order of magnitude of the concentrations and rates of the system under study [44]. Efficient schemes for determining the weights for metabolic networks has been developed by Varner and co-workers [49].
Conclusion
An iterative methodology for model identification from experimental data is developed in this paper. Identifiability tests are performed for an optimal measurement set selection for a given experimental design. The optimal measurements represent maximum information such that the model identification process is maximally benefitted. The model identification process is decoupled into two parts. In the first, the measurements are used to estimate all the unmeasured quantities of the system. This is achieved using the State Regulator Problem (SRP) formulation which is based on the assumption that the cell is an optimal strategist and uses its resources in an optimal manner. The SRP algorithm developed in this work has shown promising results. The average errors in the estimates for a significant fraction of the unmeasured responses is less than 10%. The accuracy of the estimates obtained by the SRP decreases with decrease in the information content due to suboptimal measurement set. In the second part, the full state and rate estimates are used to determine the model parameters. The decoupling relaxes considerable computation burden compared to estimating all model parameters simultaneously from the limited measurements. In the final step, a model-based experiment design determines the optimal experimental procedure that generates the most informative measurements for the next iteration. A strong dependence is observed between parameter identifiability and model performance. Thus, it is critical that the experiment design and measurement set be chosen such that maximum number of parameters are identifiable.
Tools developed for quantitative analysis of the dynamics of cellular pathways have tremendous potential in improving the predictive capabilities of biological systems especially in cases where experimental data is available but the kinetic parameters of the pathway reactions are unknown. The model developing tools are used for a host of applications and systems analysis.
The measurement selection algorithm presented in this work is freely available as part of a model analysis and development toolkit, BioSens [50].
Methods
Measurement set selection
The measurement selection analysis is performed using the "initial" parameter set for the "preliminary" experimental conditions shown in Table 1. A sampling time of 5 minute is assumed for a total simulation time of 100 minutes. Using the orthogonal procedure [19], the non-identifiable parameters are eliminated (13 of the 27 parameters). Perturbations of the non-identifiable parameters have no noticeable effect upon system dynamics for the given experimental protocol or have a strong correlation with the perturbation of one or more identifiable parameters. The non-identifiable parameters include the rate constants for auto-activation of the procaspases (parameters 5 and 6). The auto-activation is orders of magnitude lower compared to the activation by initiator. The small contribution of the auto-activation cannot be independently captured by the measurements and hence leads to non-identifiability. All other non-identifiable parameters are reaction saturation constants; the dynamics of which are not captured by the measurements at 5 minute sampling time for the given measurement noise. Equation 8 is used to estimate a bound on the confidence interval of the identifiable parameters and the deviation from the nominal value is calculated using Equation 9. A measurement set of 7 protein concentrations is assumed to be available. No measurements of the reaction rates are available. The choice of the measurement set was such that the maximum confidence was obtained for the identifiable parameters. The choice was made by a rigorous brute force search among all possible combinations. The optimal measurement set is shown in Table 2 and the confidence intervals are shown in Table 3. All the identifiable parameters have a confidence window with percentage error less than 30%. Further reduction in the percentage errors would require assuming more measurements in addition to the current 7 measurements, a faster sampling of the available measurements or a new experimental protocol.
Table 1 Experimental procedures used in model identification for the caspase system. Both receptor and stress signals are increased from zero to their maximum value in 30 minutes after which they are held constant (units same as in Varner model [7]).
Maximum receptor signal Maximum stress signal
Preliminary Experiment 0.24 0
Optimal Experiment 0.09 0.045
Test Experiment 0.15 0.030
Table 2 Measurement sets for the SRP estimator for the caspase system.
States measured
Optimal set 2 3 5 7 10 11 12
Suboptimal set 2 3 4 5 7 10 12
Table 3 Confidence intervals for the model parameters of the caspase system. Case 1: suboptimal experiment with optimal measurement set; Case 2: suboptimal experiment with suboptimal measurement set; Case 3: optimal experiment with suboptimal measurement set.
No. Case 1 Case 2 Case 3
CI % E CI % E CI % E
1 1.05 ± 0.25 23.31 1.05 ± 0.26 24.43 0.55 ± 0.04 07.18
2 1.65 ± 0.06 03.69 1.65 ± 0.06 03.50 0.95 ± 0.02 02.20
3 0.52 ± 0.01 02.45 0.52 ± 0.01 02.45 0.29 ± 0.04 15.53
4 1.69 ± 6e-3 00.34 1.69 ± 5e-3 00.32 1.69 ± 0.54 32.23
5 0.62 ± 0.01 01.81 0.62 ± 0.01 01.81 0.62 ± 0.14 22.38
6 NI - NI - NI -
7 NI - NI - NI -
8 0.87 ± 0.11 12.51 0.87 ± 0.11 12.48 1.33 ± 0.18 13.28
9 0.91 ± 0.25 27.87 0.91 ± 0.87 95.69 0.75 ± 0.44 57.85
10 0.15 ± 2e-3 01.49 0.15 ± 2e-3 01.49 0.09 ± 6e-4 00.70
11 0.23 ± 8e-3 03.39 0.23 ± 0.02 06.67 0.23 ± 0.04 17.69
12 2.12 ± 0.56 26.23 2.12 ± 0.59 27.59 17.8 ± 1.60 08.99
13 0.13 ± 2e-3 01.18 0.13 ± 2e-3 01.17 0.10 ± 8e-4 00.81
14 NI - NI - NI -
15 NI - NI - 14.7 ± 4.39 30.55
16 NI - NI - 128 ± 30.8 24.03
17 NI - NI - NI -
18 NI - NI - NI -
19 NI - NI - 127 ± 3.38 2.66
20 NI - NI - NI -
21 (3.21 ± 0.06) × 1e3 01.76 (3.21 ± 3.14) × 1e3 97.51 (3.22 ± 1.16) × 1e3 36.05
22 NI - NI - NI -
23 NI - NI - NI -
24 NI - NI - NI -
25 NI - NI - (3.09 ± 1.05) 33.95
26 0.79 ± 0.12 14.57 0.79 ± 0.12 14.52 1.04 ± 0.22 21.07
27 8.65 ± 0.31 03.60 NI - 8.62 ± 3.62 42.04
Model identification
1st iteration
The "preliminary" experiment (Table 1) with a suboptimal measurement set (Table 2) is used for obtaining the estimates of the unknown concentration and reaction rate trajectories using the SRP algorithm. The sampling time is taken to be 5 minutes with a prediction horizon of 4. A higher prediction horizon showed no appreciable change in the estimates. The initial condition of the protein concentrations is assumed to be equal to the corresponding "actual" system corrupted by up to 25% relative error. Measurements are obtained from the "actual" system with up to 10% noise. The tolerance of the concentration measurements (14) is taken to be 5%. Figures 5 and 6 show estimated versus "actual" profiles for protein and reaction rate trajectories respectively.
Figure 5 Profiles of the 19 protein concentrations for the caspase system. Solid line: actual system; dashed line: estimate by the SRP algorithm with the suboptimal experiment with suboptimal measurement set (first iteration)
Figure 6 Profiles of the 11 reaction rates for the caspase system. Solid line: actual system; dashed line: estimate by the SRP algorithm with the suboptimal experiment with suboptimal measurement set (first iteration)
The estimation error is determined by calculating the difference between the estimated and "actual" value for a rate/state at time k scaled by the maximum value over the entire simulation. Equation 22 shows the estimation error for concentration (the equation for reaction rates is identical). The scaling is done with respect to the maximum value in order to prevent misleading analysis at low concentrations or reaction rates:
The estimation errors (Equation 22) are given in Tables 4 and 5. Overall, it is observed that the estimates are fairly accurate and the system dynamics are captured. The average errors are less than 15% for all the state estimates and for most of the reaction rate estimates. Poor estimates, especially during the initial sampling times are mainly caused by the mismatch in the initial conditions. Further, the noise in the measurements results in fluctuations in some of the estimates. It should be noted that the estimation errors would not be available in real situations because the "actual" profiles are unknown. These are included here as a proof of concept.
Table 4 Estimation error for the protein concentrations in caspase system.
Protein (xi) First iterationa Second iterationb Second iterationc
Max. Avg. Max. Avg. Max. Avg.
1 08.93 01.91 08.93 01.91 08.93 01.91
2 12.49 04.74 10.54 03.55 12.04 04.32
3 03.64 01.38 04.02 01.31 03.27 01.40
4 02.16 00.82 02.38 00.78 01.94 00.83
5 14.63 03.95 12.46 03.35 13.87 03.62
6 11.54 02.47 11.54 02.47 11.54 02.47
7 09.04 02.58 07.48 02.07 08.47 02.27
8 51.10 10.65 12.65 04.40 51.63 11.24
9 05.11 01.46 05.11 01.58 05.09 01.49
10 10.22 04.18 08.45 02.35 09.02 03.65
11 48.30 09.41 07.79 02.74 50.75 10.49
12 18.56 02.76 17.67 02.59 18.19 02.69
13 22.27 14.66 22.10 15.40 13.13 03.45
14 15.00 03.21 15.00 03.21 15.00 03.21
15 12.04 02.58 12.04 02.58 12.04 02.58
16 07.02 01.50 07.02 01.50 07.02 01.50
17 21.00 04.50 21.00 04.50 21.00 04.50
18 08.42 01.80 08.42 01.80 08.42 01.80
19 19.00 04.07 19.00 04.07 19.00 04.07
a suboptimal experiment with suboptimal measurement set
b suboptimal experiment with optimal measurement set
c optimal experiment with suboptimal measurement set
Table 5 Estimation error for the reaction rates in caspase system.
Reaction rate (ri) First iterationa Second iterationb Second iterationc
Max. Avg. Max. Avg. Max. Avg.
1 98.97 15.98 103.3 15.91 67.05 14.14
2 49.09 05.68 48.76 05.68 45.11 06.54
3 30.86 11.60 30.46 11.64 29.49 11.93
4 225.9 14.86 38.71 18.08 221.1 16.25
5 34.32 10.44 33.49 09.79 33.87 10.30
6 61.73 38.69 66.30 41.42 65.93 38.27
7 79.40 37.65 80.00 38.12 81.26 37.20
8 35.03 10.47 36.67 10.75 36.03 11.34
9 112.1 38.78 109.9 37.32 112.5 36.55
10 34.50 17.92 36.56 17.92 29.43 10.44
11 76.59 62.44 72.72 60.33 29.76 10.03
a suboptimal experiment with suboptimal measurement set
b suboptimal experiment with optimal measurement set
c optimal experiment with suboptimal measurement set
The estimates are then used to determine the model parameters by solving the optimization problem in Equation 18. The optimization to determine the parameters is a nonlinear program for the caspase system in which the model equations for the reaction rates are nonlinear with respect to the parameters. The nonlinear optimization is solved using the MATLAB routine fmincon which employs a gradient descent search method. As a starting point for the search, the "initial" parameter values are used. The optimization is solved for each reaction rate separately to obtain all the parameters in the model equations. Figures 3 and 4 compare the prediction profiles of the protein concentration and the reaction rates obtained with the estimated parameters with the "actual" profiles. These profiles are for an experiment condition ("test" conditions in Table 1) that is different from the one used for model identification. As a model invalidation test, the prediction errors are calculated using Equation 20. A threshold of 35% for the maximum error and 10% for the average error can be considered to be stringent. Figure 7 shows the result for the measured protein concentrations. It is observed that after the first iteration, the threshold is violated by the errors in the predictions of the FADD, FAS/FASL-FADD complex, and the cytochrome c. Tables 6 and 7 show the prediction errors for all the protein concentrations and reaction rates using the estimated parameters and also the "initial" parameters set. Again it is important to note that all these would not be available but are included here as a proof of concept.
Table 6 Error in the model predictions for the protein concentrations using the "initial" parameters, the estimated parameters after the first iteration, and the estimated parameters after second iteration for the "test" experiment of the caspase system.
(xi) Initial parameters First iterationa Second iterationb Second iterationc
Max. Avg. Max. Avg. Max. Avg. Max. Avg.
1 11.21 02.26 05.66 01.14 02.54 00.51 12.04 02.43
2* 38.08 15.37 16.42 07.58 15.92 06.44 06.02 04.55
3* 90.59 16.37 72.90 09.57 78.27 09.91 32.07 02.11
4* 53.58 09.69 45.41 05.91 40.71 04.19 15.79 01.31
5* 75.88 27.09 21.47 13.56 16.54 02.81 06.40 03.96
6 01.74 00.35 04.82 00.97 04.76 00.96 05.52 01.11
7* 69.30 31.54 15.13 09.34 08.36 02.43 02.87 01.35
8 40.52 30.01 21.11 05.55 20.80 04.79 14.21 03.94
9 16.20 10.53 01.95 00.83 10.05 02.95 03.01 01.99
10* 56.81 38.13 17.30 06.43 12.93 05.32 15.92 03.01
11* 47.03 34.93 24.23 06.28 18.50 03.95 05.18 03.79
12* 66.80 36.63 07.88 03.04 13.61 04.96 12.22 06.98
13 53.52 39.26 22.14 14.01 22.25 13.75 04.66 02.79
14 11.13 02.24 04.75 00.96 17.06 03.44 15.18 03.06
15 11.69 02.36 07.36 01.48 03.80 00.77 07.90 01.59
16 05.84 01.18 00.72 00.15 06.79 01.37 04.65 00.94
17 10.47 02.11 09.19 01.85 03.77 00.76 14.38 02.90
18 10.61 02.14 07.92 01.60 08.75 01.77 05.66 01.14
19 13.13 02.65 17.94 03.62 09.46 01.91 01.10 00.22
asuboptimal experiment with suboptimal measurement set
bsuboptimal experiment with optimal measurement set
coptimal experiment with suboptimal measurement set
*measured protein concentration for the invalidation test
Table 7 Error in the model predictions for the reaction rates using the "initial" parameters, the estimated parameters after the first iteration, and the estimated parameters after second iteration for the "test" experiment of the caspase system.
(ri) Initial parameters First iterationa Second iterationb Second iterationc
Max. Avg. Max. Avg. Max. Avg. Max. Avg.
1 76.90 08.74 39.88 04.68 39.23 04.08 15.83 03.08
2 90.69 07.75 80.38 05.66 74.15 05.66 48.14 02.59
3 98.85 59.01 48.39 10.98 39.17 03.33 07.83 02.33
4 92.70 39.32 72.30 11.25 63.99 08.22 20.95 05.45
5 77.79 46.55 14.52 04.71 14.38 07.99 12.95 08.80
6 87.73 60.74 43.40 10.46 43.64 09.30 26.27 07.14
7 32.22 20.58 3.92 01.57 12.40 05.10 05.13 03.52
8 74.11 23.75 28.21 13.53 23.36 09.19 11.27 04.83
9 65.88 37.41 33.29 20.39 14.47 09.46 24.21 12.18
10 100.0 53.42 100.00 16.39 100.0 04.53 06.00 04.87
11 36.65 16.09 80.79 53.88 79.09 55.00 13.76 08.04
asuboptimal experiment with suboptimal measurement set
bsuboptimal experiment with optimal measurement set
coptimal experiment with suboptimal measurement set
Figure 7 Maximum and average prediction errors for the measured protein concentrations. Case 1: first iteration (suboptimal experiment with suboptimal measurement); Case 2: second iteration (suboptimal experiment with optimal measurements); Case 3: second iteration (optimal experiment with suboptimal experiment). The measured protein concentrations are (A) FAS/FASL complex; (B) FADD; (C) FAS/FASL-FADD complex; (D) cytochrome c; (E) Apaf-1-cytochrome c complex; (F) executioner procaspase; (G) caspase-8; (H) caspase-9.
2nd iteration
The second iteration is performed in two ways suggested in Figure 2. The first case involves using suboptimal experiment design with the optimal measurement set; whereas, the second case uses the optimal experiment with the suboptimal measurements. The model obtained after the first iteration is used to identify the optimal experiment that generate maximum information. For the selection of the optimal experiment design it is assumed that the measurement set remains unchanged. For the caspase system, the receptor and stress signals are parameterized such that both signals start at time 0 with constant rate injections reaching the maximum level in 30 minutes. The design variables are the final levels of the receptor and stress signals of the caspase activation. The search was constrained over a range of 0–0.4 for the receptor signal and 0–0.05 for the stress signal. A brute force search results in an optimal experiment (Table 1) with maximum information content for 18 identifiable parameters. The confidence interval for the identifiable parameters (Equation 8) and the deviations from the "nominal" parameter values (Equation 9) are shown in Table 3. Here it should be noted that the "nominal" parameters are the values obtained after the first iteration. The optimal levels for the signals suggest an optimal experiment with low receptor concentrations and high stress signal.
In each of the two cases, the model identification procedure is repeated in a similar manner as in the first iteration. The errors in the estimates of the protein concentrations and the reaction rates for both cases are shown in Tables 4 and 5 respectively. It is observed that the optimal measurement set improves the estimates of the states for which the information content increases. For example, the optimal measurement set includes caspase-8 (state 11), a state that is not included in the suboptimal measurements. The measurement of the caspase-8 improves the estimates of both the caspase-8 (state 11) and the procaspase-8 (state 8). The estimates are used to refine the parameter estimates using Equation 19.
Figure 7 shows the maximum and average prediction errors for the measured concentrations for the "test" experiment for the two cases in the second iteration. With the optimal measurements, although the overall errors are reduced, the threshold values are still violated. However, the predictions with parameters obtained from the optimal experiment reduce all the errors below the threshold. This support the termination of the iterative process with an acceptable model. The model prediction for all the concentrations and reaction rates for the "test" experiment are shown in Figures 3 and 4 and the prediction errors in Tables 6 and 7.
Authors' contributions
KG performed the model identification work and drafted the manuscript. KG and RG performed the optimal experiment design and measurement selection work. FJD conceived of the study and participated in its design and co-ordination. All authors have read and approved the final manuscript.
Appendix
The model of the caspase activated apoptosis proposed by Varner and co-workers [7] consists of 19 states (protein concentrations) and 11 reaction rates. Figure 8 gives the schematic of the apoptosis mechanism.
Figure 8 Caspase-dependent apoptosis mechanism. The model includes two triggers for the activation of cell suicide mechanism, extracellular death ligand and stress-related factor [7]. The cell death occurs when executioner caspase is activated by caspase-8 (ligand effector) or caspase-9 (stress-related effector).
The model Equations can be represented as:
where xi denotes the ith protein concentration, rj denotes the jth reaction rate, Ωk denotes the rate of synthesis of the protein k and μ denotes the protein complex degradation rate. The reaction rates are as follows:
where
Tables 8 and 9 represent the nomenclature of the protein complexes and parameter values in the apoptosis model, respectively [7].
Table 8 Nomenclature in caspase-dependent apoptosis model.
No. Protein complex name No. Protein complex name
1 total receptor ligands 11 caspase-8
2 clustered FAS/FASL complex 12 caspase-9
3 FADD 13 executioner caspase
4 FAS/FASL-FADD complex 14 decoy protein
5 cytochrome c 15 decoy protein
6 Apaf-1 16 decoy protein
7 Apaf-1-cytochrome c complex 17 activator protein
8 procaspase-8 18 Bcl-2
9 procaspase-9 19 Bcl-xL
10 executioner procaspase
Table 9 Parameter values in caspase-dependent apoptosis model.
No. Parameter No. Parameter No. Parameter No. Parameter
1 kl 8 k83a 15 KH 22 KK
2 ka 9 k93a 16 KI 23 KL
3 kh 10 αCE 17 KJ 24 KN
4 k8za1 11 ku 18 KC 25 KO
5 k9za1 12 KS 19 KD 26 KP
6 k8za2 13 KA 20 KF 27 KR
7 k9za2 14 KB 21 KG
Acknowledgements
This work is supported by National Science Foundation (BES-0000961), by the Institute for Collaborative Biotechnologies through grant DAAD19-03-D-0004 from the U.S. Army Research Office, and by the DARPA BioComp program.
==== Refs
Kitano H Computational Systems Biology Nature 2002 420 206 210 12432404 10.1038/nature01254
Ideker T Thorsson V Ranish JA Christmas R Buhler J Eng JK Bumgarner R Goodlett DR Aebersold R Hood L Integrated Genomic and Proteomic Analyses of a Systematically Perturbed Network Science 2001 292 929 934 11340206 10.1126/science.292.5518.929
Zhu H Huang S Dhar P The Next Step in Systems Biology: Simulating the Temporospatial Dynamics of Molecular Networks Bioessays 2003 26 68 72 14696042 10.1002/bies.10383
Cho KH Shin SY Kolch W Wolkenhauer O Experimental Design in Systems Biology, Based on Parameter Sensitivity Analysis Using a Monte-Carlo Method: A Case Study for the TNFα-Mediated NF- κB Signal Transduction Pathway Simulation 2003 79 726 739 10.1177/0037549703040943
Edwards JS Ibarra RU Palsson BO In silico Predictions of Escherichia coli Metabolic Capabilities are Consistent with Experimental Data Nat Biotechnol 2001 19 125 130 11175725 10.1038/84379
Schöberl B Jonsson CE Gilles ED Müller G Computational Modeling of the Dynamics of the MAP Kinase Cascade Activated by Surface and Internalized EGF Receptors Nat Biotechnol 2002 20 370 375 11923843 10.1038/nbt0402-370
Fussenegger M Bailey JE Varner J A Mathematical Model of Caspase Function in Apoptosis Nat Biotechnol 2000 18 768 774 10888847 10.1038/81208
Chen KC Csikasz-Nagy A Gyorffy B Val J Novak B Tyson JJ Kinetic Analysis of a Molecular Model of the Budding Yeast Cell Cycle Mol Bio Cell 2000 11 369 391 10637314
Feng XJ Rabitz H Optimal identification of biochemical reaction networks Biophys J 2004 86 1270 1281 14990460
Moles CG Mendes P Banga JR Parameter Estimation in Biochemical Pathways: A Comparison of Global Optimization Methods Genome Res 2003 13 2467 2474 14559783 10.1101/gr.1262503
Esposito WR Floudas CA Global Optimization for the Parameter Estimation of Differential-Algebraic Systems Ind Eng Chem Res 2000 39 1291 1310 10.1021/ie990486w
Grossmann IE Global Optimization in Engineering Design 1996 Dordrecht, Netherlands: Kluwer Academic Publishers
Papamichail I Adjiman CS A Rigorous Global Optimization Algorithm for Problems with Ordinary Differential Equations J Global Optim 2002 24 1 33 10.1023/A:1016259507911
Guss C Boender E Romeijn HE Handbook of Global Optimization Dordrecht, Netherlands: Kluwer Academic Publishers 1995 chap. Stochastic methods 829 869
Ali MM Storey C Törn A Application of Stochastic Global Optimization Algorithms to Practical Problems J Optim Theory Appl 1997 95 545 563 10.1023/A:1022617804737
Törn A Ali MM Viitanen S Stochastic Global Optimzation: Problem Classes and Solution Techniques J Global Optim 1999 14 437 447 10.1023/A:1008395408187
Audoly S Bellu G D'Angio L Saccomani MP Cobelli C Global Identifiability of Nonlinear Models of Biological Systems IEEE Trans Biomed Eng 2001 48 55 65 11235592 10.1109/10.900248
Jacquez JA Perry T Parameter estimation: local identifiability of parameters Amer J Physiol 1990 258 E727 E736 2333964
Yao KZ Shaw BM Kou B McAuley KB Bacon DW Modeling Ethylene/Butene Copolymerization with Multi-Site Catalysts: Parameter Estimability and Experimental Design Polymer Reaction Eng 2003 11 563 588 10.1081/PRE-120024426
Landaw EM DiStefano III JJ Multiexponential, Multicompartmental, and Noncompartmental Modeling. II. Data Analysis and Statistical Considerations Amer J Physiol 1984 246 R665 R677 6720989
Petersen B Gernaey K Vanrolleghem PA Practical Identifiability of Model Parameters by Combined Respirometric-Titrimetric Measurements Water Science Tech 2001 43 347 355
Zak DE Gonye GE Schwaber J Doyle III FJ Importance of Input Perturbations and Stochastic Gene Expression in the Reverse Engineering of Genetic Regulatory Networks: Insights from an Identifiability Analysis of an in Silico Network Genome Res 2003 13 2396 2405 14597654 10.1101/gr.1198103
Asprey SP Macchietto S Statistical Tools for Optimal Dynamic Model Building Comput Chem Eng 2000 24 1261 1267 10.1016/S0098-1354(00)00328-8
Kremling A Fischer S Gadkar K Doyle III FJ Sauter T Bullinger E Allgower F Gilles ED A Benchmark for Methods in Reverse Engineering and Model Discrimination: Problem Formulation and Solutions Genome Res 2004 14 1773 1785 15342560 10.1101/gr.1226004
Banga JR Versyck KJ Impe JFV Computation of Optimal Identification Experiments for Nonlinear Dynamic Process Models: A Stochastic Global Optimization Approach Ind Eng Chem Res 2002 41 2425 2430 10.1021/ie010183d
Mahadevan R Edwards JS Doyle III FJ Dynamic Flux Balance Analysis of Diauxic Growth in E coli Biophys J 2002 83 1331 1340
Ljung L System Identification: Theory for the User 1999 Englewood Cliffs, NJ: Prentice Hall
Ljung L Guo L The Role of Model Validation for Assessing the Size of the Unmodeled Dynamics IEEE Trans Automat Contr 1997 42 1230 1239 10.1109/9.623084
Poolla K Khargonekar P Tikku A Krause J Nagpal K A Time-Domain Approach to Model Validation IEEE Trans Automat Contr 1994 39 951 959 10.1109/9.284871
Brogan WL Modern Control Theory 1991 Upper Saddle River, NJ: Prentice Hall
Vajda S Rabitz H Walter E Lecourtier Y Qualitative and Quantitative Indentifiability Analysis of Non-Linear Chemical Kinetic Models Chem Eng Commun 1989 83 191 219
Emery AF Nenarokomov AV Optimal Experimental Design Meas Sci Technol 1998 9 864 876 10.1088/0957-0233/9/6/003
Beck JV Arnold KJ Parameter Estimation in Engineering and Science 1977 Toronto, Canada: John Wiley & Sons
Gadkar KG Gunawan R Doyle III FJ Heuristic Methods for Measurement Selection for Identification of Biological Systems Int Conf Systems Biology, Heidelberg, Germany 2004
Bastin G Dochain D On-Line Estimation and Adaptive Control of Bioreactors 1990 Amsterdam: Elsevier
Albiol J Robusté J Casas C Poch M Biomass Estimation in Plant Cell Cultures Using Extended Kalman Filter Biotechnol Prog 1993 9 174 178 10.1021/bp00020a009
Gee DA Ramirez WF On-Line State Estimation and Parameter Identification for Batch Fermentation Biotechnol Prog 1996 12 132 140 10.1021/bp950049m
Stephanopoulos G San KY Studies on on-Line Biorector Identification. I. Theory Biotechnol Bioeng 1984 26 1176 1188 10.1002/bit.260261006
Glassey J Ignova M Ward AC Montague GA Morris AJ Bioprocess Supervision: Neural Networks and Knowledge Based Systems J Biotechnol 1997 52 201 205 9084212 10.1016/S0168-1656(96)01645-8
Karim MN Rivera SL Comparision of Feed-Forward and Recurrent Neural Networks for Bioprocess State Estimation Comp Chem Eng 1992 16 S369 S377
Simutis R Lübbert A Exploratory Analysis of Bioprocesses Using Artificial Neural Network-Based Methods Biotechnol Prog 1997 13 479 487 10.1021/bp9700364
Gunawan R Jung MYL Seebauer EG Braatz RD Maximum a Posteriori Estimation of Transient Enhanced Diffusion Energetics AIChE J 2003 49 2114 2123 10.1002/aic.690490819
Cover TM Thomas JA Elements of Information Theory 1991 John Wiley & Sons
Gadkar KG Varner J Doyle III FJ Model Identification of Signal Transduction Networks from Data Using a State Regulator Problem IEE Systems Biology 2005 2
Wagner A How to reconstruct a large genetic network from n gene perturbations in fewer than n2 easy steps Bioinformatics 2001 17 1183 1197 11751227 10.1093/bioinformatics/17.12.1183
Gardner TS Bernardo DD Lorenz D Collins JJ Inferring genetic networks and identifying compound model of action via expression profiling Science 2003 301 102 105 12843395 10.1126/science.1081900
Basso K Margolin AA Stolovitzky G Klein U Dalla-Favera R Califano A Reverse engineering of regulatory networks in human B cells Nat Genet 2005 37 382 390 15778709 10.1038/ng1532
Chen BH Asprey SP On the Design of Optimally Informative Deynamic Experiments for Model Discrimination in Multiresponse Nonlinear Situations Ind Eng Chem Res 2003 42 1379 1390 10.1021/ie0203025
Frey AD Kallio PT Gadkar KG Varner J Dynamic Flux Balance Analysis of VHb Expression in Escherichia coli MG1655 Biotechnol Bioeng 2005 accepted
Taylor S Gunawan R Doyle III FJ BioSens 2.0: Sensitivity Analysis Toolkit 2005
|
15967022
|
PMC1189077
|
CC BY
|
2021-01-04 16:27:27
|
no
|
BMC Bioinformatics. 2005 Jun 20; 6:155
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-155
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1611598517410.1186/1471-2105-6-161SoftwareFACT – a framework for the functional interpretation of high-throughput experiments Kokocinski Felix [email protected] Nicolas [email protected] Gunnar [email protected] Lars [email protected] Grischa [email protected] Peter [email protected] Molecular Genetics, Deutsches Krebsforschungszentrum, 69115 Heidelberg, Germany2 Wellcome Trust Sanger Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge CB10 1HH, UK2005 28 6 2005 6 161 161 14 3 2005 28 6 2005 Copyright © 2005 Kokocinski et al; licensee BioMed Central Ltd.2005Kokocinski et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Interpreting the results of high-throughput experiments, such as those obtained from DNA-microarrays, is an often time-consuming task due to the high number of data-points that need to be analyzed in parallel. It is usually a matter of extensive testing and unknown beforehand, which of the possible approaches for the functional analysis will be the most informative
Results
To address this problem, we have developed the Flexible Annotation and Correlation Tool (FACT). FACT allows for detection of important patterns in large data sets by simplifying the integration of heterogeneous data sources and the subsequent application of different algorithms for statistical evaluation or visualization of the annotated data. The system is constantly extended to include additional annotation data and comparison methods.
Conclusion
FACT serves as a highly flexible framework for the explorative analysis of large genomic and proteomic result sets. The program can be used online; open source code and supplementary information are available at .
==== Body
Background
A variety of algorithms and programs have been introduced to accomplish the processing of raw data as well as the statistical analysis of data from high-throughput experiments. But besides the mathematical complexity that needs to be handled, there is a biological complexity inherent to the data sets, too. Current means to analyze large-scale data sets usually target very specific questions and often fail to provide solutions that can be adapted to different types of data. Nevertheless, common and generalized questions for the interpretation of such data can be established as follows: i) What information is known about the analyzed features (clones, genes, e.g.)? ii) Are there correlations between the experimental outcomes and the additional information (shared pathways, etc.)? iii) Is the outcome comparable with results of other experiments (genomic or gene expression data sets, publications, etc.)?
The program Flexible Annotation and Correlation Tool (FACT) was developed to address these questions by integrating data sources, tools and algorithms in a single open framework. First, FACT allows merging information from various data sources into one comprehensive annotation for an experimental data set. It then provides functional analysis tools to inspect and correlate this heterogeneous information. The functionality of FACT can be extended through the inclusion of new data sources, algorithms and programs by defining additional modules from a prototype. This flexibility is achieved by a strong level of abstraction from the actual data, by the design of the underlying database and by the modular architecture of the software itself. The task to identify relevant biological interconnections reflected by the experimental results (e.g. participation of the analyzed genes in shared pathways) is what we are targeting with the software introduced here.
Implementation
Integration of data sources
The integration of bio-molecular data from diverse sources such as public databases or clinical parameters (annotation) is a key challenge in the process of the analysis of high-throughput experiments. While the interpretation of the outcome of a standard experiment used to depend on the knowledge of one human expert from the field, today's screening tools produce data quantities not manageable by human inspection. After receiving a list of differentially regulated genes from a microarray gene expression experiment comprising several hundreds or thousands of entries, it is not efficient to start the interpretation of these results by manually searching through publications. As a first step, broad biological themes should be identified and followed into a more detailed inspection.
Using network technologies, the availability of data sources is no more the limiting factor, but if accomplished manually, the obstacles for their integration are numerous. Often data are made available in different formats (HTML pages, flat files, direct database access) and very heterogeneous layouts. In addition the nomenclature (e.g. gene names) as well as the relationship of different systems to each other are often inconsistent and require many manual selection and modification steps. At the same time, as much knowledge as possible should be integrated about the data features analyzed, since interesting unknown pathways and interconnections might be hidden behind biological complexity.
As the first aspect, FACT accomplishes the task of integrating heterogeneous information sources by abstraction from the specific data types to one basic concept. (figure 1, lower part). The smallest entities are data features, which are single items of information, either a name/value pair or additional textual description thereof. This could be a list of ids of clones with their relative expression as measured on a cDNA microarray. They are grouped into data sets, combining data features relating to the same experiment or a group of annotation terms for a certain set. In our example the clone/ratio pairs measured in one hybridization would be stored as one data set. The data sets originate from specific data sources, defining distinct types of experiments or annotation sources. One data source would be "cDNA microarray measurements with textual clone ids and numerical results". This architecture follows the idea, that the primary data must be represented at a sufficient level of abstraction to make the data independent of the source technology [1]. The concept applies to experimental as well as to annotation data. Meta-data about the different data sources is stored in dedicated tables of the underlying database, describing the source with date and type of data.
Figure 1 Layered architecture of the FACT framework. The database reflects the abstraction of any experimental or annotation data to DataSets with DataFeatures, originating from a specific DataSource for which DataTypes and Parameters have been defined. The core library (API) supplies all functionality for accessing the database and for the operation of diverse modules, which are adaptors for specific DataSources or functions. The web interface or other applications are using FACT API functions.
Differences in nomenclature and the problem of relating one type of experiment to an other, as the second obstacle for data integration, has been addressed for gene and protein centered research by the development of the GeneOntology system (GO) [2]. This hierarchical framework of a directed acyclic graph of annotations for gene attributes has become a de facto standard, which can be employed in the functional analysis of experiments. Similar projects have been initiated for example to organize the classification of molecular interactions in pathways and molecular complexes (Genome Knowledgebase / Reactome [3]). Using these resources FACT is able to compare experimental results from technically distant applications.
However, most experiments differ in focus and design and usually no standard solution for their interpretation can be applied. The third aspect for an integrating approach therefore is high flexibility concerning the application of diverse analysis methods that have already been developed for the interpretation of results or might be used in future.
Architecture
FACT consists of a MySQL database, a core library (as an Application Programming Interface, API) and various modules written in the language Perl. We also created an interface for the web-based usage of all functions of FACT [4]. The database reflects the transformation of heterogeneous data into a generalized format, storing information in DateSet and DataFeature tables (figure 2). The core software framework supplies the basic functionality for data access in an object-oriented fashion and for adding and operating modules. These modules are adaptors and can be classified into three categories (figure 3):
Figure 2 FACT database schema. The database schema reflects the generalized handling of heterogeneous data. At the definition layer the data sources are defined as experimental, annotational or analysis sources. Also the types of data that they use are specified here. These types are linked to the individual sources which are defined in the data source layer. Parameters that the functions handling the sources can take are stored as well. The actual data – experimental as well as annotational – are saved as data set and data features in the data set layer.
Figure 3 Outline of the flow of information. Modular DataSource-Adapters accomplish data access and data transformation from heterogeneous sources, making FACT a flexible framework.
I. Data loading
Different types of experimental data can be loaded using dedicated parser functions. This can be a simple functions to read tab-delimited data defined e.g. as a gene list with associated expression values. It can also be a more complex solution to handle case descriptions from comparative genomic hybridizations (CGH), a method that is employed to monitor copy numbers changes of all regions of a genome simultaneously on chromosome spreads. The individual modules read the specific file format, perform transformations to the generalized format (i.e. convert them into data features) and use core library functions to store the data.
II. Annotation
Varying data sources can be utilized for the annotation of experimental data sets by different data-access functions (e.g. GO terms for gene names). Modules achieve this for instance through access to an online database or to a local copy of such database. Data of interest are then gathered and stored.
III. Analysis
Different functions can be used to inspect the annotated information and highlight underlying patterns (e.g. overrepresented GO-terms). The modules typically produce textual and graphical output to draw the researcher's attention to the most promising features of his data.
Currently available functions of FACT are described below. Further flexibility is achieved by the concept of experimental and annotational data being reduced to the basic model of one data set with several data features, as described above.
All these modules use the FACT API. It offers a defined interface for the effortless extension to new sources and functions. Prototype modules for each category implementing this API are supplied. For the integration of annotation sources, available data can either be transferred to the local system (data warehousing) or linked to the original source (database federation); the FACT system allows both options to be used by the module functions. Currently remote databases are accessed by the EnsEMBL, BBID and Reactome modules; the CpG and CGAP functions use locally stored information (see below). The update of the local data is accomplished semi-automatically be invoking of the respective update function in the separate modules.
Finally, as there is an active development of software for the annotation and analysis of gene expression data in the language R (Bioconductor project [5]), and the handling of large data matrices is accomplished faster in R, we used the Perl/R interface RSPerl [6] in different modules to encapsulate analysis functions written in R. Other modules employ the functionality from the BioPerl [7] and Ensembl Perl API [8].
Available functionality
A variety of modules for the handling of different data sources (table 1) as well as for the application of basic data analysis and display functions (table 2) were developed. Most of the current functionality is focused on handling human and murine gene annotation information. Additional functionality can be added in a "plug-and-play" fashion, since new modules can be loaded dynamically into FACT.
Table 1 Data Types and Sources accessible by current annotation modules.
Data source, access method Data provider, data location Type of annotation used by FACT
Ensembl, Perl API access to local or remote database European Bioinformatics Institute and Wellcome Trust Sanger Institute (GB) [8], Ensembl ID, Gene Symbol, Gene Name, Chromosomal Location, Homologues Genes, Interpro Domains, RefSeq Accession Number, Affymetrix ID
euGenes, local database University of Indiana (USA) [10], euGene ID, Gene Symbol, Gene Name, GDB ID, OMIM ID, Genomic Localization, GeneOntology Terms, Protein Accession Numbers
Image Consortium, local database Lawrence Livermore National Laboratory (USA) [28], Clone Image ID
Biological Biochemical Image Database, HTTP parser and HTTP request National Institute of Aging, NIH (USA) [11], Pathway Name and Image-link
GeneOntology, local database GeneOntology Consortium [2], ID and Name of GO-Term (Biological Process, Molecular Function, Cellular Localization)
Cancer Genome Anatomy Project, local database National Cancer Institute, NIH (USA) [29], Biocarta name, Biocarta short name, KEGG Pathway Name, KEGG Pathway ID, PFAM ID
LocusLink / EntrezGene, local database NCBI/NIH (USA) [30], / A. LocusLink ID, Gene Symbol, Gene Name, Genomic Localization, GeneOntology Terms, OMIM ID B. Key references (PubMed links)
Mouse Genome Database, local database Jackson Laboratory (USA) [31], MGI ID / Gene Symbol
Internal CloneBase, local database Deutsches Krebsforschungs zentrum, Div. Molecular Genetics (D) General Information on available Clones
CpG, local database University of California Santa Cruz (USA), Calculated relative CpG content of genomic region
STRING, local database EMBL (D) [12], (medium or better confidence) Protein interaction data (computed and imported from other databases)
Affymetrix CEL files Affymetrix Inc. / FACT, Use of Affymetrix probe IDs
Reactome, local database and HTTP request European Bioinformatics Institute (GB) [3], Pathway information
Table 2 Current data analysis and display modules
Method Name Reference Method Description
Simple Count FACT Count and display of occurrences of annotation terms
GO-Term Comparison In part from GO::TermFinder [15] Detection of significantly overrepresented GO terms in Gene List, based upon hypergeometric tail probability
MedLiner Bio::Biblio (M. Senger, EBI) List Publications with co-occurrences of terms
CGH database Deutsches Krebsforschungs-zentrum, Div. Molecular Genetics (D) Compare CGH results to archived data
goCluster goCluster, G. Wrobel, available at Detection of significantly overrepresented GO terms (based upon Fisher's exact test) in Clusters built with k-means algorithm
Hypergeometric Tail In part from GeneMerge [14] Detection of significantly overrepresented terms of any kind, based upon hypergeometric tail probability
CGH – Expression Comparison FACT Detect correlation between genomic and expression data sets, based on two-sided T-Tests
Chromosomal Plot FACT Display values or occurrences in genomic context
To read in experimental results, a simple list (with terms or term-value pairs) or table can be used, or more specialized parser functions can be employed to read and decipher the notations of different types of results. One parser (2_Colums) reads tab-, or semicolon-separated lists and stores the data as terms of the data type that is passed as a parameter (e.g. gene symbol or clone id) and the respective value. Another function (LongList) expects terms only (list of genes that are to be annotated) or Affymetrix probe ids (AffyCelFile). The parser for CGH results translates ISCN notations of cytogenetic alterations [9] into the distinct chromosomal bands that are affected while reading the data file. The bands are stored with the alteration -1 (loss of genomic material), +1 (gain), or +2 (high level amplification).
Data sources that can then be used for annotating these experimental results contain among others EnsEMBL databases [8], with functions providing numerous gene annotations on the human and murine genome (gene name, accession numbers, genomic localization, GO terms and other ids). The annotation data is fetched by using the EnsEMBL API or direct sql queries. Chromosomal locations expressed as cytogenetic bands can be translated in megabasepair positions. This can permit the direct comparison of results from genomic and expression experiments. Most common identifiers (IMAGE IDs, DDBJ/EMBL/GenBank accession numbers, international clone names, MGD (Mouse-Genome Database) IDs or Probe-IDs as used on expression microarrays in the Affymetrix system) are recognized and used by the different annotation modules. Additionally, homologous genes, sequence features, InterPro protein domains, CpG content and PubMed references can be acquired. The euGenes database [10] modules delivers an additional set of broad annotations (Gene Symbol and Name, GDB ID, OMIM ID, Genomic Localization, GO terms, etc.). The BBID module searches for representations of affected pathways in the Biological Biochemical Image Database [11]. We store links to the images which sometimes allow the clarification of interactions better than textual description alone. Additionally data is used from STRING [12] and Reactome [3] to point out protein interactions and involvement in molecular complexes.
Annotation steps can be concatenated, allowing deriving from specific (e.g. Affymetrix Probe-IDs) to more general terms (e.g. gene symbols). This broad annotation that is facilitated by FACT is crucial for the researcher to acquire a complete picture of his data. The user can export the combined lists of annotated data in HTML, XML or text format. If desired the system can send them by email.
Modules to correlate these annotated datasets with each other have been developed for FACT, incorporating existing algorithms or presenting new approaches (table 2). For example, a counting procedure (SimpleCount) reports the number of occurrences of each annotation term. The module is independent of the type of data, one or more data sets can be added up and a threshold for reporting can be defined. The results are displayed graphically in a chart and in a table format, directing the researcher's attention to distinct characteristics (figure 4). A more specific approach to interpret list of genes lies in the explicit usage of GO terms. As originally introduced by Khatri et al. in the program OntoExpress [13] there are different methods and implementation to search for those parts of the GO tree that appear more often in the gene list analyzed than by chance alone. For FACT we used an implementation from the GeneMerge program [14] and the GO::TermFinder perl module [15]. Occurrences of terms in a gene list are normalized against a background, which might contain the annotations of all terms spotted on a chip or the entire genome, by using the hypergeometric tail probability (with Bonferroni correction if desired, example: figure 5). In our implementation, the function can be applied on GO data as well as on any other kind of annotation to detect overrepresented terms. Alternatively the Fisher's exact test can be used for this purpose. We included part of this method taken from the EASE program [16] in a FACT function. We also developed a combined algorithm, which calculates a K-Means similarity matrix based on the experimental values and reports on significant terms with the Fisher's exact test within the identified groups afterwards (goCluster, G. Wrobel, available at ). Another analysis module dedicated to the comparison of genomic and expression data runs pairs of two-sided T-Tests on the groups of over- and under-expressed genes against the groups of enhanced/amplified and deleted genomic regions. This allows detecting a correlation between an altered genomic locus and the corresponding expression pattern. FACT can also represent experimental values or frequency counts in the genomic context which can be helpful for the identification and representation of localization effects in the genome. This is achieved by reading in the genomic locations of data sets and generating bar chart images on top of prepared chromosome ideograms (figure 6). The MedLiner module presents a simple literature mining tool which uses the Bio::Biblio perl functions to find citations that are shared by two or more gene (Figure 5). Example output files are also available at [4].
Figure 4 Application of the FACT system for the functional analysis of microarray data of the development of non-melanoma skin cancer (SimpleCount function). Occurrences of annotation terms are counted and displayed to draw the researchers attention to potentially characteristic features of the data set. In this case the genomic bands at 1q21 seem to play an important role in the experiment.
Figure 5 Application of the FACT system in non-melanoma skin cancer research (GoTerm function). Overrepresented terms from a Gene Ontology annotation are displayed in a chart. The usage of the GO system is the most common approach for the functional interpretation of gene lists.
Figure 6 Application of the FACT system in non-melanoma skin cancer research (ChromPlot function). Visual representation of the genomic distribution of analyzed features highlights the involvement of the genomic band 1q21 (first 5 chromosomes shown). In this case the localization of human homologues genes corresponding to murine clones over- and under expressed in squamous cell carcinoma are displayed.
Application of the program
The Flexible Annotation and Correlation Tool has proven especially helpful with the interpretation of results from genomic and expression microarray experiments, but most functions can be applied to a broad variety of experimental data.
We demonstrate the benefits of FACT for the functional interpretation of the results from a comprehensive analysis of gene expression patterns in the development of non-melanoma skin-cancer conducted within our group [L. Hummerich et al.: Identification of novel tumor-associated genes in the process of squamous cell cancer development; submitted]. Using two different sets of microarrays with 15.000 and 20.000 cDNA fragments respectively, the chemically induced multi-step development of squamous cell carcinoma was monitored. We used the dorsal skin of mice as a well-studied system for the development of epithelial cancer [18] with the carcinogen 7,12-dimethylbenz-[a]-anthracene and the tumor promoter 12-O-tetradecanoylphorbol-13-acetate as inducing agents. Expression values were measured at different time-points of tumor formation. Genes with differentiated expression patterns are expected to play a role in the development of human epidermal tumor development as well.
Preprocessed results where loaded into the FACT system as text files containing the murine clone identifier and expression values. Using mainly information from the Ensembl database, FACT annotation functions acquired corresponding gene names, genomic localizations, functional information from Gene Ontology, homologues human gene names and the genomic localization of those genes (supplement S1, complete set of annotated data). We applied different FACT analysis functions to explore the expression data (for example the TermFinder function to search for enriched functional groups of genes, figure 5). The function SimpleCount was used to search for enriched functional categories of annotations. Using the information of homologues genes, we were able to identify the human chromosomal band 1q21, as a region of accumulated differentially expresses genes in murine skin cancer formation (figure 4). The visual representation of the loci using the ChromPlot function highlights these findings (figure 6). With the MedLiner module we were to confirm that several regulated genes were collectively mentioned in previous publications (figure 7). A number of genes (S100A3, S100A6, S100A8, S100A9) which are part of the S100 family of calcium-binding proteins are involved in the regulation of AP-1 and NFκB-dependent transcription. Using FACT we were able to focus our analysis and to gain understanding of the relevance of the results from the microarray study. With the application, it was possible to find human homologues of the murine tumor-associated genes and to confirm the involvement of the S100 protein family in human epidermal malignancies.
Figure 7 Application of the FACT system in non-melanoma skin cancer research (MedLiner function). FACT's simple automated literature screen function displays publications mentioning groups of genes identified in the study (top 5 hits shown).
Comparison to and inclusion of other tools
There are several large-scale database projects that incorporate an immense spectrum of information about genes and gene product (Ensembl, euGene, LocusLink/EntrezGene, etc.). The Ensembl project for example also allows the user to display his own selected data sources in the context of the full genome annotation through the Distributed Annotation System [19], and it offers the possibility of mining the data of several genomes using the EnsMart software [20]. FACT uses Ensembl, LocusLink/EntrezGene and euGene data and complements them with other annotation resources; it allows the user to apply different analysis functions on the combined data.
Recently, a variety of computational tools have been introduced to aid in the interpretation of results, some of which are of interest concerning FACT. The majority of the programs use GO annotations to gain an interpretation of gene expression data. OntoExpress [13] was introduced in 2002 and offers the options to use hypergeometric, chi-square, binomial and Fischer's exact test to score annotation term derived from gene lists. It also allows the appliance of different methods (False Discovery Rate (FDR), Bonferroni, Holm, Sidak) for the multiple experiment correction [17]. The program can also include KEGG pathway information and chromosomal localization and is now part of the Onto-Tools collection to offer further functionality [21]. EASE [16]/ DAVID [22] offer a broad variety of annotation options in their latest version including all major database identifiers, protein domain and pathway information. The Fisher's exact test is used for the detection of enriched terms. GoMiner [23] and numerous other tools listed at the GO website [24] can be used for the annotation of gene lists with GO terms. GeneMerge [14] uses the hypergeometric tail probability with Bonferroni correction to test GO terms, genomic localizations and KEGG pathway information and is used in parts within the FACT system. FACT uses the available Perl code of GO::TermFinder [15] for the GO annotation and the detection of significantly overrepresented terms using the FDR. It also includes a function combining K-means clustering and Fisher's exact test. GEPAS [25] and GECKO [26] are two recently introduced large software packages that include functional analysis and visualization steps. In contrast to FACT they are focused on the initial statistical evaluation and on the analysis of gene expression microarray data. GFINDer [27] is a system that offers annotations on GO, pathway information, protein domains and genetic disorders. It analyses with count functions and appropriate tests (Hypergeometric, Binomial, Fisher's, Z or Poisson Test).
This list of available tools is far from complete and not all aspects are covered. The FACT system was developed with the focus to include and extend the functionality of tools like these. To our knowledge, the individual programs do not offer the same degree of flexibility and openness to different data sources and analysis methods. New functions can be added to FACT by simply uploading the respective module. The system is designed as an open framework for the explorative analysis using a variety of methods on annotational data. It is not restricted to or focused on Gene Ontology-based interpretation or the analysis of gene expression data alone and should facilitate the development and application of new analysis approaches. The system is constantly being extended to include additional aspects. With the submission as an open-source project we want to encourage other researchers to participate in this development.
Conclusion
To gain a more complete picture of results obtained from high-throughput experiments such as DNA-microarrays, automated procedures are required for annotation and analysis. At the same time it is usually a matter of testing and not known beforehand, which of the possible approaches for the functional analysis will be the most informative or appropriate. The Flexible Annotation and Correlation Tool offers the flexibility to integrate and compare annotation data and different algorithms in one environment by using a unified data basis. Data sets of different nature and format can be incorporated, diverse analytical algorithms can be applied and the user can add his own data integration and analysis functions. As a flexible framework for the explorative meta-analysis of genomic, proteomic or other experiments, FACT can help with the task of analyzing the biological complexity, allowing researchers to bridge gaps between different kinds of experiments and acquiring a more complete interpretation of large-scale experiments.
Availability and requirements
- Project name: Flexible Annotation and Correlation Tool (FACT).
- Project home page:
- Operating system: tested on Linux SUSE 9.1
- Programming language: Perl (5.8.1)
- Other requirements: MySQL database (4.0.15); for specific modules: R (1.8.0 with RSPerl) and Bioconductor (1.4.0); for full installation: Apache web-server (apache2-prefork-2.0.48); additional Perl modules: BioPerl (1.2.1), Ensembl (currently 28). Please refer to website for full listing.
- License: Open Source GNU GPL (see licence document)
- Any restrictions to use by non-academics: written licence needed
Abbreviations
API – application programming interface, CGH – comparative genomic hybridization, FACT – Flexible annotation and correlation tool, GO – Gene Ontology, ISCN – international system for human cytogenetic nomenclature
Authors' contributions
FK designed and implemented the FACT system and the web-interface, ND re-designed and extended it, GW helped with the initial design and supplied R-modules, LH carried out the microarray experiments and supplied additional ideas, GT conducted the integration into other analysis systems, PL supervised the FACT project. All authors read and approved the final manuscript.
Acknowledgements
We are grateful for the contributions of Regina Mueller, Jochen Hess and Peter Angel within the non-melanoma skin cancer project and the helpful comments from Anja Kolb-Kokocinski and Imre Vastrik. The FACT project was supported by grants from the German Ministry for Education and Research (NGFN 01 GR 0101, NGFN 01GR 0417 and NGFN 01GR 0418).
==== Refs
Mangalam H Stewart J Zhou J Schlauch K Waugh M Chen G Farmer AD Colello G Weller JW GeneX: An Open Source gene expression database and integrated tool set IBM Systems J 2001 40
GeneOntology Consortium Creating the gene ontology resource: design and implementation Genome Res 2001 11 1425 33 11483584 10.1101/gr.180801
Joshi-Tope G Gillespie M Vastrik I D'Eustachio P Schmidt E de Bono B Jassal B Gopinath GR Wu GR Matthews L Lewis S Birney E Stein L Reactome: a knowledgebase of biological pathways Nucleic Acids Res 2005 33 D428 32 15608231 10.1093/nar/gki072
FACT website
Gentleman RC Carey VJ Bates DM Bolstad B Dettling M Dudoit S Ellis B Gautier L Ge Y Gentry J Hornik K Hothorn T Huber W Iacus S Irizarry R Leisch F Li C Maechler M Rossini AJ Sawitzki G Smith C Smyth G Tierney L Yang JY Zhang J Bioconductor: open software development for computational biology and bioinformatics Genome Biol 2004 5 R80 15461798 10.1186/gb-2004-5-10-r80
RSPerl
Stajich JE Block D Boulez K Brenner SE Chervitz SA Dagdigian C Fuellen G Gilbert JGR Korf I H Lapp Lehvaslaiho H Matsalla C Mungall CJ Osborne BI Pocock MR Schattner P Senger M Stein LD Stupka E Wilkinson MD Birney E The Bioperl toolkit: Perl modules for the life sciences Genome Res 2002 12 1611 1618 12368254 10.1101/gr.361602
Hubbard T Barker D Birney E Cameron G Chen Y Clark L Cox T Cuff J Curwen V Down T Durbin R Eyras E Gilbert J Hammond M Huminiecki L Kasprzyk A Lehvaslaiho H Lijnzaad P Melsopp C Mongin E Pettett R Pocock M Potter S Rust A Schmidt E Searle S Slater G Smith J Spooner W Stabenau A Stalker J Stupka E Ureta-Vidal A Vastrik I Clamp M The Ensembl genome database project Nucleic Acids Res 2002 30 38 41 11752248 10.1093/nar/30.1.38
Standing Committee on Human Cytogenetic Nomenclature An International System for Human Cytogenetic Nomenclature (1985) Birth Defects Orig Artic Ser 1985 21 1 117 4041569
Gilbert DG euGenes: a eukaryote genome information system Nucleic Acids Res 2002 30 145 148 11752277 10.1093/nar/30.1.145
Becker KG White SL Muller J Engel J BBID: the biological biochemical image database Bioinformatics 2000 16 745 6 11099263 10.1093/bioinformatics/16.8.745
von Mering C Huynen M Jaeggi D Schmidt S Bork P Snel B STRING: a database of predicted functional associations between proteins Nucleic Acids Res 2003 31 258 61 12519996 10.1093/nar/gkg034
Khatri P Draghici S Ostermeier GC Krawetz SA Profiling gene expression using Onto-express Genomics 2002 79 266 70 11829497 10.1006/geno.2002.6698
Castillo-Davis CI Hartl DL GeneMerge – post-genomic analysis, data mining, and hypothesis testing Bioinformatics 2003 19 891 2 12724301 10.1093/bioinformatics/btg114
Boyle EI Weng S Gollub J Jin H Botstein D Cherry JM Sherlock G GO::TermFinder – open source software for accessing Gene Ontology information and finding significantly enriched Gene Ontology terms associated with a list of genes Bioinformatics 2004 20 3710 5 15297299
Hosack DA Dennis G JrSherman BT Lane HC Lempicki RA Identifying biological themes withing lists of genes with EASE Genome Biol 2003 4 R70 14519205 10.1186/gb-2003-4-10-r70
Draghici S Khatri P Martins RP Ostermeier GC Krawetz SA Global functional profiling of gene expression Genomics 2003 81 98 104 12620386 10.1016/S0888-7543(02)00021-6
DiGionvanni J Multistage carcinogenesis in mouse skin Pharmacol Ther 1992 54 63 126 1528955 10.1016/0163-7258(92)90051-Z
Dowell RD Jokerst RM Day A Eddy SR Stein L The distributed annotation system BMC Bioinformatics 2001 2 7 11667947 10.1186/1471-2105-2-7
Kasprzyk A Keefe D Smedley D London D Spooner W Melsopp C Hammond M Rocca-Serra P Cox T Birney E EnsMart: a generic system for fast and flexible access to biological data Genome Res 2004 14 160 9 14707178 10.1101/gr.1645104
Khatri P Bhavsar P Bawa G Draghici S Onto-Tools: an ensemble of web-accessible, ontology-based tools for the functional design and interpretation of high-throughput gene expression experiments Nucleic Acids Res 2004 32 W449 56 15215428 10.1093/nar/gkh086
Dennis G JrSherman BT Hosack DA Yang J Gao W Lane HC Lempicki RA DAVID: Database for Annotation, Visualization, and Integrated Discovery Genome Biol 2003 4 P3 12734009 10.1186/gb-2003-4-5-p3
Zeeberg BR Feng W Wang G Wang MD Fojo AT Sunshine M Narasimhan S Kane DW Reinhold WC Lababidi S Bussey KJ Riss J Barrett JC Weinstein JN GoMiner: A Resource for Biological Interpretation of Genomic and Proteomic Data Genome Biol 2003 4 R28 12702209 10.1186/gb-2003-4-4-r28
Gene Ontology website
Herrero J Vaquerizas JM Al-Shahrour F Conde L Mateos A Diaz-Uriarte JS Dopazo J New challenges in gene expression data analysis and the extended GEPAS Nucleic Acids Res 2004 32 W485 W491 15215434
Theilhaber J Ulyanov A Malanthara A Cole J Xu D Nahf R Heuer M Brockel C Bushnell S GECKO: a complete large-scale gene expression analysis platform BMC Bioinformatics 2004 5 195 15588317 10.1186/1471-2105-5-195
Masseroli M Martucci D Pinciroli F GFINDer: Genome Function INtegrated Discoverer through dynamic annotation, statistical analysis, and mining Nucleic Acids Res 2004 32 W293 300 15215397 10.1093/nar/gkh108
Lennon G Auffray C Polymeropoulos M Soares MB The I.M.A.G.E. Consortium: an integrated molecular analysis of genomes and their expression Genomics 1996 33 151 2 8617505 10.1006/geno.1996.0177
Strausberg RL Buetow KH Emmert-Buck M Klausner R The Cancer Genome Anatomy Project: building an annotated gene index Trends in Genetics 2000 16 103 106 10689348 10.1016/S0168-9525(99)01937-X
Pruitt KD Katz KS Sicotte H Maglott DR Introducing RefSeq and LocusLink: curated human genome resources at the NCBI Trends Genet 2000 16 44 7 10637631 10.1016/S0168-9525(99)01882-X
Blake JA Richardson JE Bult CJ Kadin JA Eppig JT the members of the Mouse Genome Database Group MGD: The Mouse Genome Database Nucleic Acids Res 2003 31 193 195 12519980 10.1093/nar/gkg047
|
15985174
|
PMC1189078
|
CC BY
|
2021-01-04 16:27:25
|
no
|
BMC Bioinformatics. 2005 Jun 28; 6:161
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-161
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1661599240610.1186/1471-2105-6-166Methodology ArticleQuality assessment of microarrays: Visualization of spatial artifacts and quantitation of regional biases Reimers Mark [email protected] John N [email protected] Genomics & Bioinformatics Group, Laboratory of Molecular Pharmacology, Center for Cancer Research, National Cancer Institute, Bethesda Maryland, 20892 USA2005 1 7 2005 6 166 166 23 11 2004 1 7 2005 Copyright © 2005 Reimers and Weinstein; licensee BioMed Central Ltd.2005Reimers and Weinstein; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Quality-control is an important issue in the analysis of gene expression microarrays. One type of problem is regional bias, in which one region of a chip shows artifactually high or low intensities (or ratios in a two-channel array) relative to the majority of the chip. Current practice in quality assessment for microarrays does not address regional biases.
Results
We present methods implemented in R for visualizing regional biases and other spatial artifacts on spotted microarrays and Affymetrix chips. We also propose a statistical index to quantify regional bias and investigate its typical distribution on spotted and Affymetrix arrays.
We demonstrate that notable regional biases occur on both Affymetrix and spotted arrays and that they can make a significant difference in the case of spotted microarray results. Although strong biases are also seen at the level of individual probes on Affymetrix chips, the gene expression measures are less affected, especially when the RMA method is used to summarize intensities for the probe sets. A web application program for visualization and quantitation of regional bias is provided at .
Conclusion
Researchers should visualize and measure the regional biases and should estimate their impact on gene expression measurements obtained. Here, we (i) introduce pictorial visualizations of the spatial biases; (ii) present for Affymetrix chips a useful resolution of the biases into two components, one related to background, the other to intensity scale factor; (iii) introduce a single parameter to reflect the global bias present across an array. We also examine the pattern distribution of such biases and conclude that algorithms based on smoothing are unlikely to compensate adequately for them.
==== Body
Background
Microarrays and other new high-throughput technologies are changing the way molecular biology is practiced. However microarray platforms and protocols are still under development, and the causes of common errors and artifacts are still not completely understood or controlled. Over time and replications, many types of errors seem almost random. Others, however, affect many gene expression measures at once, introducing systematic biases into the data. Most statistical methods are designed to deal with measures corrupted by random noise; methods to deal with systematic biases are not so well developed.
Spotted arrays
Over the extent of a single spotted microarray, factors such as temperature, liquid flow rate or RNA diffusion rate may differ among different regions on the array. Often, washing is less thorough near the edges of a slide, contributing to higher local off-spot background near the edges. High local background usually pushes up the spot measures, although not always predictably. To detect such technical artifacts, it is now standard practice to examine images of slides for pronounced irregularities and high backgrounds. Such examination can identify many types of faults, but even a skilled technician may miss regions of higher than average intensity in Affymetrix arrays or moderate biases in the ratios calculated for spotted arrays, since the vast majority of spots appear dim in images calibrated for the dynamic range of the brightest spots. At the moment there is no way of quantifying regional biases, and a lot is left to the technician's judgment.
Affymetrix arrays
To address such issues, Affymetrix has gone to great lengths to standardize their procedure. However, uniform results are rarely achieved in practice. Often a bubble remains after filling an Affymetrix cassette. This bubble will not travel uniformly over the chip during hybridization mixing and may get stuck or move in an irregular circuit. Scratches and other manufacturing imperfections can make a difference. Scratches are sometimes visible with the aid of software such as dChip [1] or RMA [2]. Although a skilled technician can identify some of the grosser faults by examining the images of hybridized Affymetrix chips, he or she has no current standard for measuring how serious the problems are or for knowing whether other sorts of systematic problems are evading scrutiny.
Current practice in quality assessment for spotted arrays considers individual spot measures, such as area and signal/noise ratio [3]. Current quality metrics for Affymetrix arrays consider 3'/5' ratios for selected genes and spike-in ratios. These quality metrics don't take variation within a slide or chip into account. We show here that such bias, which is currently ignored, can be a significant problem.
Methods
All computation was done within the R programming environment [11], and the Affymetrix analysis used the affy package [12].
Detection of regional bias on spotted arrays
An effective way to present information about regional biases is through plots or maps of the ratios or signals over the chip surface. For two-color arrays it is natural to plot ratios as a function of position. Because all ratios are represented at the same brightness, such a plot makes it much easier to see patterns of regional bias than does inspection of the raw image file. Such a plot is shown in Figure 1A. Similar plots are available through arrayMagic [4]. However we still face the problem that many different ratios (high and low) are juxtaposed on the slide, making it difficult to see subtle but consistent biases.
Figure 1 The advantages of using a reference to highlight regional biases. Figure 1A shows log2 spot ratios at a constant intensity. Red corresponds to a log ratio of greater than 0.5; yellow to a log ratio of 0, and green to a log ratio of less than -0.5. Figure 1B shows the log2 ratios for the same slide relative to the averages of the log2 ratios across all the slides. Figure 1C shows the log ratios for the same slide after background subtraction. The right portion of the bottom row in each print-tip group was spotted with buffer only.
Many two-color microarray experiments focus on a single tissue but use a common reference RNA not specific to the tissue. In such a design, two neighboring probes will often show consistently different red/green ratios across all slides, reflecting the typical abundance of the probes' mRNA targets in the samples relative to the common reference. We would like to compare each slide's probe ratios with a standard ratio profile, reflecting the typical abundance of all mRNA species in the tissue under study relative to the reference. We approximate such a common standard by computing, for each probe, the 20%-trimmed mean of the probe's log ratios, across all slides. In doing so, we are assuming that the biological variation due to sample and the regional biases on each slide will tend to balance out over the whole experimental set.
For each slide we then compute the difference between the log ratio of each spot and the spot's average log ratio over all of the pertinent arrays:
(1) di,j = log2(Ri,j/Gi,j) – mi; mi = trim(log2(Ri,k/Gi,k)),
where d is the difference, i indexes the spot, j indexes the particular slide, k indexes all slides, and trim refers to the 20% trimmed mean of a set; Ri,j and Gi,j are the red and green channel intensities of spot i on chip j. When these differences di,j are represented as colors over the area of the chip, then often the high and low ratio values are clearly concentrated in some sub-regions. An example is shown in Figure 1B. More examples are in the Supplementary Material. Because the probes for most co-regulated sets of genes are distributed widely throughout the chip, we don't expect that a biological process would generate such a pattern. Hence, such regional inhomogeneity of ratios must be a technical artifact.
Affymetrix arrays
Affymetrix raw data are considerably denser (per unit area) than spotted array data, so a deeper investigation is possible. We present several types of plots here showing different aspects of bias. The first (Figure 2, upper left) shows how the raw intensity data look if we present brightness on a logarithmic scale. Ref [5] shows a similar plot. Because of the log transformation, this plot brings out detail in the low range (intensity values typically between 50 and 150). Typically, this low range contains more than half the probes on a chip. Such a plot often shows striations because probes of similar sequence are placed in rows.
Figure 2 Regional biases on an Affymetrix chip. An Affymetrix chip is represented in log2 scale at upper left and the ratio of the same Affy chip to the standard chip at upper right. Each pixel of the original image represents one probe. The color legend is shown at bottom; bright red represents off-scale high, and white represents off-scale low. Blue rectangles in the upper plots indicate non-coding probes. In the lower row are images for background and scale factor for the same chip. The left plot represents the local background – the lowest levels achieved by probes on this chip relative to the lowest levels achieved by probes in the same region in other chips. The lower right plot represents the effective sensitivity or local scale factor – the log ratio of values of typically bright probes on this chip, to their values across the other chips. The scale factor captures all the variation seen in the top right image while the background image shows almost no variation. This clear separation between background artifacts and scale factor artifacts is typical in Affy chips. This chip is within the range of acceptable by our QC protocol (see discussion).
Even more than with spotted array ratios, it is difficult to see subtle spatial patterns on an Affymetrix chip image because neighboring probes show such a wide range of different intensities next to one another. To make biases visible we would ideally like to compare individual slides with a standard that represents a good, uniform hybridization. Ideally, we would like to have many replicate slides of at least one representative sample and to use their average as a standard. In practice we rarely have such replicates. Hence, in the approach to be presented here, we construct a reference for the Affymetrix chips (hereafter called the 'standard' chip) by taking a trimmed mean of each probe across all chips (from the same tissue) in the experimental series. In the calculations here, we use a 20% trim, which seems to work satisfactorily. This standard chip ideally represents the probe intensities for a 'typical' sample in the experimental series – a virtual sample of the same tissue type with expression values intermediate among those of all samples in the experimental series. We then plot the differences between log values on each chip and the standard chip:
(2) di,j = log2(Inti,j) – trim(log2(Inti,k)),
where i indexes the probe, j indexes the chip, and k indexes all chips; Inti,j is the intensity of probe i on chip j. A plot of dij is shown in Figure 2 at upper right. We note that the differences {di,j} reflect discrepancies from an average, and cannot detect regions on a slide which consistently show the same bias.
Using the greater density of probes on an Affymetrix chip we can investigate in more detail how the log differences in equation (2) (i.e., the log ratios of intensities) between the sample chip and standard chip) differ from one region to another. To estimate the log difference in local background of a region, we adopt a heuristic procedure, first selecting those probes with intensities in the lowest one-fifth of probe intensities for the chip as a whole. Then we compute the 20%-trimmed mean of differences between the log2-intensities of the selected probes on the chip and the corresponding probes on the standard:
(3) Pj = trim(log2(Inti,k)
(4) bg = trim(log2(Inti,j) -Pj) | Pj < qP,0.2).
Here, trim(x|S) represents the 20%-trimmed mean of the variable x restricted to the set S, and Inti,j represents the intensity of probe j on chip i, Pj represents the log2-intensity of probe j on the standard, and qP,a represents the a-th quantile of probe intensities on the standard chip.
To compute the log2 scale factor, S, we use the 20%-trimmed mean among the highest 20% of probe intensities in the region:
(5) S = trim(log2(Inti,j) – Pj | Pj > qP,0.8).
We then construct heat maps of the log2 background factor (bg) and log2 scale factor (S) over the chip. When these plots are placed side by side (the bottom left and right plots in Figure 2), we see regions in which the background is raised but the scale factor is unaffected, and vice versa. Further examples are in the Supplementary Material. The code for making such plots for Affymetrix chips is available on our website .
Quantitation of regional bias
It is important to have some scale on which to measure the distortions introduced by spatial effects, and to have some idea how much difference these distortions make to the final estimates of gene expression. The simplest estimate is correlation between each probe intensity or ratio and the average of its four neighbors. For spotted arrays the measure of correlation is
(6) R = < rlm, (rl,m-1+rl,m+1+rl-1,m+rl+1,m)/4 >,
where l indexes rows, m indexes columns, and rl,m is the log2 red-green ratio at a spot indexed by l and m. In the case where all slides use a common reference, then the difference between the log2 ratio of the spot and the average signal from that spot may be used, as described earlier. The notation <x,y> refers to the Pearson correlation between variables x and y over all values of l and m in the array. For an array with no regional bias, R would equal 0; for one with regional bias, R > 0. For Affymetrix arrays we computed R using the difference between the log2 ratio of each probe and the average signal from that probe, and we used only neighbors within rows, because neighbors within a column include the corresponding mismatch probes, which should be highly correlated with the perfect match probe.
Measuring effect of regional biases on estimates
To test how much biases affect the expression estimates, we selected several very clean-looking chips from several different studies and systematically distorted their CEL file data by multiplying regions of various sizes by factors of 1.41 and 2, corresponding to log2 changes of 0.5 and 1.0. The distortion patterns were selected to mimic patterns that we observe in real chips. We then estimated the gene abundances using the MAS5 and RMA algorithms in the affy package of Bioconductor. The results did not depend much on the exact shape of the region distorted, and results were comparable using different chip types (not shown).
Results
We investigated several hundred spotted microarrays and Affymetrix chips, from over a dozen different studies, finding noticeable bias in almost all slides and in most chips. Many of the studies included some slides or chips whose regional biases were severe enough to compromise at least part of the study.
Spotted arrays
Using our methods we find both sharply defined, high-contrast artifacts, and diffuse regional biases. The most common sort of regional bias on spotted arrays is associated with high backgrounds of one color over a region. It is usually supposed that background subtraction removes such biases. We find, however, that the standard method of subtracting the off-spot local background from each channel does not effectively correct regional biases (see Figure 1C) and sometimes introduces them (see Figure 2). The implicit model underlying background subtraction is that the amount of non-specific DNA binding to the substrate around a spot is equal to that within the spot and additive to the target-specific binding within the spot. There are several different mechanisms that cause fluorescent signal outside of the spots – such as direct binding of dye or labeled cDNA to substrate, reflection from substrate, and binding of labeled target to smeared probe – and only some of them will contribute additively to the measured signal on spots. We think that the issue of adjustment for background needs more thought than is usually given it.
Affymetrix chips
We find three major types of spatial artifact on Affymetrix arrays:
1) Obvious, distinct artifacts with sharp boundaries; most of those defects cover less than 5% of chip area.
2) Regional shifts in non-specific signal background: in wide areas of the chip the tenth percentile may be as much as 50% higher than the corresponding quantile over the remainder of the chip
3) Regional shifts in scale factor: in wide areas, the highest values of both PM and MM appear to be up to 50% lower than corresponding values in other areas; the scale factor appears very uneven, and shows a characteristic turbulent appearance.
We suspect that artifacts of type 2) and 3) are also present on spotted arrays, but we do not detect them as readily because the density of features is lower.
By looking in such detail, we find many irregularities in even the best Affymetrix chips. Since probes for each gene are distributed across the chip, however, a modest area (5–10% of the chip area) of affected probes is not a serious problem. A robust statistical method, such as MAS5, dChip, RMA, or the PLIER method, will down-weight those values, as described later. The problems become more serious when large (more than 20%) regions of the chip are higher in intensity than other regions by a factor of 1.5 or more. If we use a linear algorithm without outlier removal, then the values for some probes may be changed more than 1.5-fold, and the few high-intensity probes may dominate estimates for genes by these methods. If one used a linear algorithm on the log-scale then the distortion over the chip should roughly average out for each probe set. However, it is difficult to predict the effect on estimates made by a robust algorithm such as MAS5 or a linear model, such as Li-Wong, RMA or PLIER, because such methods remove outliers, and these may be found preferentially in one region of the chip. Below, we investigate empirically the effects of regional biases on gene expression estimates from robust algorithms.
Typical measured biases
On spotted arrays we find typical correlations R between raw ratios of 0.05 to 0.1 and typical correlations using log ratios relative to the average of 0.1 to 0.2. Some slides show correlations as high as 0.6 in log ratios relative to the average.
On Affymetrix chips we find, as did Workman, only slight correlation using intensities. The correlations are much stronger for ratios of individual probes to their typical values, as instantiated in the 'standard' virtual chip. A good chip will typically show correlations in ratio relative to standard between nearest neighbors of 0.1 to 0.2. We observe the highest correlations along horizontal straight lines in the most recent generation of chips. That is so because probes with similar sequence motifs are often printed on lines (Earl Hubbell, Affymetrix Inc, personal communication), and sequence similarity may predict similar responses to many variations in conditions.
Effect of regional biases on Affymetrix estimates
Table 1 shows the effects of simulated regional bias on gene expression for one particular sample on a Human Focus array. Systematic experiments with other array types yield comparable results.
Table 1 Effect of deliberate regional bias distortions on MAS5 estimates
Region .05 .10 .25 .50
Factor
1.4 RMA <.001 .011 .019 .029
MAS5 .019 .031 .047 .063
R .426 .549 .69 .74
2.0 RMA .001 .003 .039 .35
MAS5 .034 .057 .100 .155
R .647 .764 .864 .89
Entries for RMA and MAS5 in the table are fractions of probe set abundance estimates (i.e., gene expression estimates) changed by more than 0.5 on a log2 scale as a consequence of the bias introduced. The rows labeled R report values of the local correlation, R.
As expected, both MAS5 and RMA are fairly robust to small distortions but, as would be expected, both methods do worse as more distortion is added to the chip images. RMA is notably more robust than MAS5 to the moderate distortions commonly found in Affymetrix chips. However, RMA does worse than MAS5 when the perturbation is most serious. A little thought makes the reason clear: RMA aims to fit the majority of the intensities on each chip well; it down-weights probes that appear too high or too low relative to the majority of others in the probe set, according to the pattern on other chips. When half of the chip is raised in intensity values relative to the other half, then roughly half the probes for each gene lie in each region. RMA fits one half well and discounts the other half.
Discussion
Consequences for data analysis
As described above, chips with significant regional distortions can be expected to yield gene expression estimates that differ significantly from the true values. Other, less distorted chips in a group will show expression values more indicative of the biology. Several studies that have come to the first author from leading core facilities include chips with very large spatial distortions that went undetected by the (rigorous) QC at the facility. Good intuition leads the data analyst to suspect certain outliers and to include others. However, data analysts prefer to have some objective criterion to reject outliers. In our experience, most chips that are outliers relative to their experimental groups show large regional distortions.
Rather than rely on intuition to discard outliers, we can use a systematic chip QC process to put outlier detection on a firm footing. We recommend that users note the R statistic, as defined in equation 6. We find that the R statistic is a useful guide to the degree of distortion in expression measures as summarized in Table 1. Users can decide how much distortion they are willing to live with, and select slides with R statistics accordingly. Our standard practice is to run the bias detection program in batch on a new set of chips, and to discard chips with R values exceeding 0.5. For chips with R values between 0.4 and 0.5, we scrutinize the images provided by our program, and decide whether the flaw is large and concentrated (in which case a robust procedure will limit damage to only a few probe sets), or moderate and pervasive, in which case more probe sets will be notably affected.
The discovery of systematic bias leads a statistician to try to compensate. Our QC visualization method is based on ratios, and ratios to a standard are a natural choice for normalization. Smoothing is an approach to spatial variation favored by many statisticians, and several authors have proposed compensations for microarray spatial biases using smoothed fits to bias [6-8]. However, those methods have not met with unqualified success [8,9].
We are not sanguine about the prospects for normalization by smoothing. Our observation is that the biases represented in the ratio plots show abrupt transitions from one region within a slide to another and also occur in complex filigree patterns. Often regions within the same print-tip group on a spotted microarray slide show apparent regional biases as large as do regions at greater distances on a slide (see Supplementary Figures). Sometimes there is a repeating pattern of biases in all print-tip groups. Smyth [10] has suggested that such repeating patterns derive from different quality 96-well plates used for printing the arrays. He proposes a print-order normalization, but many arrays show non-repeating, non-random patterns of bias, which can't be compensated in that way.
A reasonable question is whether regional biases in Affymetrix chips can be eliminated by comparing PM with MM. In fact, we find that a plot of log (PMij) – log (MMij) values for a chip, relative to the same quantities for the average chip, shows much less regional bias than does a plot of log probe intensities relative to their averages. That observation suggests that, in practice, the MAS5.0 PM correction reduces regional biases in scale factor, whereas the RMA procedure does not. In the same way, the MAS5.0 background correction reduces regional biases in background, whereas the RMA procedure does not. However our results in synthetically distorted chips suggest this advantage of the MAS5.0 procedure is telling only in the presence of strong regional bias (R>0.4).
Conclusion
We have shown that regional biases are common on microarrays, and that in some cases they may be responsible for apparent large differences in gene expression. We have presented methods for visualizing and quantifying the levels of regional bias (and other spatial artifacts). In our judgment the most practical way to use information about regional biases on microarrays is in the quality assessment step, rather than in an attempt to compensate for it. We hope that others will use the tools we have provided at to visualize and quantify these biases on their microarrays.
Supplementary Material is online at
Authors' contributions
MR conceived of the study, wrote the code, analyzed the results and drafted the manuscript. JNW suggested the perturbation study, and provided extensive and detailed comments on the manuscript. All authors read and approved the final manuscript.
Figure 3 An example where background subtraction induces regional bias. The left image shows the raw spot ratios relative to average; the middle image shows ratios of the off-spot local background, the right image shows the ratios after background subtraction. The color legend for all three images is at bottom.
==== Refs
Li C Wong WH Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection Proc Natl Acad Sci U S A 2001 98 31 36 11134512 10.1073/pnas.011404098
Irizarry RA Bolstad BM Collin F Cope LM Hobbs B Speed TP Summaries of Affymetrix GeneChip probe level data Nucleic Acids Res 2003 31 e15 12582260 10.1093/nar/gng015
Wang X Hessner MJ Wu Y Pati N Ghosh S Quantitative quality control in microarray experiments and the application in data filtering, normalization and false positive rate prediction Bioinformatics 2003 19 1341 1347 12874045 10.1093/bioinformatics/btg154
Buness A Huber W Steiner K Sultmann H Poustka A arrayMagic: two-colour cDNA microarray quality control and preprocessing Bioinformatics 2004
Gautier L Cope L Bolstad BM Irizarry RA affy--analysis of Affymetrix GeneChip data at the probe level Bioinformatics 2004 20 307 315 14960456 10.1093/bioinformatics/btg405
Balazsi G Kay KA Barabasi AL Oltvai ZN Spurious spatial periodicity of co-expression in microarray data due to printing design Nucleic Acids Res 2003 31 4425 4433 12888502 10.1093/nar/gkg485
Colantuoni C Henry G Zeger S Pevsner J SNOMAD (Standardization and NOrmalization of MicroArray Data): web-accessible gene expression data analysis Bioinformatics 2002 18 1540 1541 12424128 10.1093/bioinformatics/18.11.1540
Workman C Jensen LJ Jarmer H Berka R Gautier L Nielser HB Saxild HH Nielsen C Brunak S Knudsen S A new non-linear normalization method for reducing variability in DNA microarray experiments Genome Biol 2002 3 research0048 12225587 10.1186/gb-2002-3-9-research0048
Qian J Kluger Y Yu H Gerstein M Identification and correction of spurious spatial correlations in microarray data Biotechniques 2003 35 42 4, 46, 48 12866403
Smyth GK Print-order normalization of cDNA microarrays
R Development Core Team A language and environment for statistical computing R Foundation for Statistical Computing, Vienna 2004
Gautier L Cope L Bolstad BM Irizarry RA Affymetrix GeneChip data at the probe level Bioinformatics 2004 20 307 315 14960456 10.1093/bioinformatics/btg405
|
15992406
|
PMC1189079
|
CC BY
|
2021-01-04 16:27:25
|
no
|
BMC Bioinformatics. 2005 Jul 1; 6:166
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-166
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1721601179810.1186/1471-2105-6-172Research ArticleThe PD-(D/E)XK superfamily revisited: identification of new members among proteins involved in DNA metabolism and functional predictions for domains of (hitherto) unknown function Kosinski Jan [email protected] Marcin [email protected] Janusz M [email protected] Laboratory of Bioinformatics and Protein Engineering, International Institute of Molecular and Cell Biology, Trojdena 4, PL-02-109 Warsaw, Poland2005 12 7 2005 6 172 172 13 4 2005 12 7 2005 Copyright © 2005 Kosinski et al; licensee BioMed Central Ltd.2005Kosinski et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The PD-(D/E)XK nuclease superfamily, initially identified in type II restriction endonucleases and later in many enzymes involved in DNA recombination and repair, is one of the most challenging targets for protein sequence analysis and structure prediction. Typically, the sequence similarity between these proteins is so low, that most of the relationships between known members of the PD-(D/E)XK superfamily were identified only after the corresponding structures were determined experimentally. Thus, it is tempting to speculate that among the uncharacterized protein families, there are potential nucleases that remain to be discovered, but their identification requires more sensitive tools than traditional PSI-BLAST searches.
Results
The low degree of amino acid conservation hampers the possibility of identification of new members of the PD-(D/E)XK superfamily based solely on sequence comparisons to known members. Therefore, we used a recently developed method HHsearch for sensitive detection of remote similarities between protein families represented as profile Hidden Markov Models enhanced by secondary structure. We carried out a comparison of known families of PD-(D/E)XK nucleases to the database comprising the COG and PFAM profiles corresponding to both functionally characterized as well as uncharacterized protein families to detect significant similarities. The initial candidates for new nucleases were subsequently verified by sequence-structure threading, comparative modeling, and identification of potential active site residues.
Conclusion
In this article, we report identification of the PD-(D/E)XK nuclease domain in numerous proteins implicated in interactions with DNA but with unknown structure and mechanism of action (such as putative recombinase RmuC, DNA competence factor CoiA, a DNA-binding protein SfsA, a large human protein predicted to be a DNA repair enzyme, predicted archaeal transcription regulators, and the head completion protein of phage T4) and in proteins for which no function was assigned to date (such as YhcG, various phage proteins, novel candidates for restriction enzymes). Our results contributes to the reduction of "white spaces" on the sequence-structure-function map of the protein universe and will help to jump-start the experimental characterization of new nucleases, of which many may be of importance for the complete understanding of mechanisms that govern the evolution and stability of the genome.
==== Body
Background
The PD-(D/E)XK superfamily of Mg2+-dependent nucleases groups together protein domains initially identified in structurally characterized type II restriction enzymes (REases) (reviews: [1,2]) and later found in diverse enzymes involved in DNA replication, repair, and recombination, including phage λ exonuclease [3], bacterial enzymes exerting ssDNA nicking in the context of methyl-directed and very-short-patch DNA repair: MutH [4] and Vsr [5], Tn7 transposase TnsA [6], archaeal Holliday junction resolvases (AHJRs) Hjc/Hje [7-9], phage T7 endonuclease I [10], the XPF/Rad1/Mus81 family of nucleases that cleaves branched structures generated during DNA repair, replication, and recombination [11], and RecB nuclease [12].
All members of the PD-(D/E)XK superfamily share a common structural core, comprising a mixed β-sheet of 4 strands flanked on both sides by α-helices [1,2,13](Figure 1). It serves as a scaffold for a weakly conserved active site, typically including two or three acidic residues (Asp or Glu) and one Lys residue, which together form the hallmark bipartite catalytic motif (P)D...Xn.(D/E)X K (where X is any amino acid). It was found that some members of the PD-(D/E)XK superfamily developed different variants of the active site, in which the acidic residues or the lysine have been replaced by Asn or Gln (Lys can be also replaced by an acidic side chain) [14-16], or in which their side-chains have "migrated" to another region of the polypeptide in a way that only the spatial orientation of functional groups in the active site is maintained, but their sequence is not conserved [17-21] (Figure 1). This variability makes it difficult to identify the active site in PD-(D/E)XK nucleases solely based on sequence comparisons and usually requires analysis of the three-dimensional structures (e.g. obtained by comparative modeling techniques). It should be mentioned that the mechanistic/catalytic role of the active site sidechains is not fully elucidated even in the well-characterized members with known structure. Moreover, many of PD-(D/E)XK nucleases contain elaborations of the common fold in the form of large insertions and terminal extensions that form additional subdomains, usually involved in oligomerization or DNA-binding. These regions often contain regular elements of secondary structure and sometimes constitute the majority of the protein, making the detection of the true core a challenging task for protein structure prediction methods (reviews: [2,22]).
Figure 1 A) Conserved topology of the common core and the most typical architecture of the active site in PD-(D/E)XK nucleases, B) Topological variation of the common fold found in the nuclease XPF (1j24), C) Extension of the common core and migration of the carboxylate residue (e.g. in NgoMIV, 1fiu), D) Different extension of the common core and migration of the lysine residue (e.g. in Tt1018, 1wjd). Helices are shown as circles, strands are shown as triangles (the orientation – up or down indicates the direction and the parallel/antiparallel character). The common core is shown in green/black, the variable elaborations are shown in red/white. The most common catalytic residues (known or putative) are indicated by letters in yellow boxes (E, D, and K), structurally important, semi-conserved P residue is also indicated.
We have previously used the comparisons of sequence profiles to detect novel members of the PD-(D/E)XK superfamily [16,18,19,23-25], as well as applied the sequence-structure threading to map the sequences of REases onto the crystal structures of known members [21,26-28]. However, both approaches have shortcomings: the profile-profile analysis is unable to find very remote relationships when families include only a few members (as is often the case in the PD-(D/E)XK superfamily) or if the compared proteins contain large unrelated insertions, and the threading analysis works only if an experimentally solved structure is available for one of the potential homologs. Therefore, we decided to use an intermediate approach, in which the compared sequence profiles are combined with the structural information, which can be obtained either from the crystal structure (if it is available for one of the members of the considered families) or predicted by bioinformatics methods, thereby allowing protein fold recognition without the explicit reference to an experimentally solved protein structure. We used the recently developed method HHsearch that allows comparison of sequence alignments together with secondary structures represented as Hidden Markov Models (HMM) [29] to compare the library of PD-(D/E)XK nuclease families obtained previously [23] with the set of domains in the Clusters of Orthologous Groups (COG and KOG)[30] and PFAM [31] databases. Our aim was to identify the previously unknown PD-(D/E)XK domains in families of proteins that are either completely uncharacterized, or whose function is known (or predicted), but the structure and phylogenetic relationships are unclear.
Results
Identification of new candidate PD-(D/E)XK nucleases with profile HHM searches
In order to carry out a systematic search for new PD-(D/E)XK nucleases, we prepared a set of multiple sequence alignments corresponding to previously identified families [18,23,32](see Methods). For each family we generated a profile HMM that included the sequences and predicted secondary structure (see Methods). These profile HMMs were compared with HHsearch [29] to a database of profile HMMs corresponding to multiple sequence alignments obtained from the COG, KOG, and PFAM databases, also with predicted secondary structures (see Methods for details). It is noteworthy, that searches initiated with most of the known PD-(D/E)XK nucleases in the query dataset identified families containing sequences of other known PD-(D/E)XK nucleases, thereby validating our approach and providing useful threshold values for confident identification of true members of PD-(D/E)XK superfamily. For instance, AtF16A14.4 profile detected herpesvirus alkaline exonuclease (pfam01771) with the e-value of 10-3, Pfu-HJR detected Mrr with the e-value of 10-2, YcjD detected Vsr with the e-value of 10-5, etc. It must be noted, that some of the "best hits" to true positives exhibited poor scores, e.g. Mrr detected HJR (pfam01870) with the e-value of 1.2. As expected, our HHM-HHM comparisons revealed also numerous other potential homologs, not included in the query dataset. Based on these results, we generated a preliminary list of candidate new PD-(D/E)XK subfamilies. For further analysis, we retained only such proteins, which have not been reported as members of the PD-(D/E)XK superfamily in earlier analyses carried out by ourselves or other groups (e.g. [32]).
The preliminary candidates for novel PD-(D/E)XK nucleases were initially validated by reciprocal HHsearches against the database comprising both the initial query profile HMMs as well as all the other COG, KOG, and PFAM profile HMMs. This search confirmed that for most of the candidate families, the closest homologs are either among the bona fide PD-(D/E)XK enzymes or among other preliminary candidates. It also revealed additional families, which were similar to the first set of candidates, but not to the original PD-(D/E)XK queries, suggesting that they may be more remote members of the superfamily. In reciprocal searches, a limited number of profile HMMs showed significant relationship to other families, unrelated to PD-(D/E)XK enzymes – such candidates were regarded as false positives and were not further analyzed. Further, each candidate family was analyzed by fold-recognition methods to evaluate its compatibility with the known PD-(D/E)XK structures (or detect cases, where some other, unrelated structure appeared to be a better template). Finally, preliminary comparative models were built for the parts of the sequence aligned to the template structures and the sequence conservation was analyzed in the structural context to detect potential active site residues. Altogether, our analysis revealed previously unknown membership of 14 families and one "orphan" protein (herein predicted to be a new restriction enzyme) in the PD-(D/E)XK superfamily of nucleases (Table 1).
Table 1 New members of the PD-(D/E)XK superfamily Columns I and II contain accession numbers of Pfam and COG database entries corresponding to protein families analyzed in this work. Typically, COGs contain only subsets of seqeunces from the corresponding Pfam entries. Previously described molecular and/or cellular function and the newly predicted function is shown. * indicates only a general functional prediction based on the fact that a given protein family is found to possess a nuclease domain with a seemingly complete set of catalytic residues.
PFAM COG representative closest homolog other domains known function predicted function
02646 1322 E. coli RmuC McrC transmembrane helix, coiled-coil regions, unknown alpha/beta domain limiting inversions at short-inverted repeats cleaves DNA structures arising during recombination of short- inverted repeats
06054 4469 S. pneumoniae CoiA COG 4636 unknown C-terminal domain DNA uptake process and recombination degradation of one DNA strand during DNA uptake
03749 1489 E. coli SfsA YraN OB-fold N-terminal domain sugar fermentation stimulation DNA cleavage*
- 2143 (KOG) H. sapiens KIAA1018 AHJR TPR-like domain, Rad18-like CCHC zinc finger, uknown 300aa domain unknown DNA repair, protein binding
06250 4804 E. coli YhcG YraN ~200aa unknown N-terminal domain unknown DNA cleavage*
06356 - Phage phi ETA orf25 TnsA or T7 EndoI none unknown degradation of the host DNA upon lytic infection, production of recombinogenic fragments
- 5482 A. tumefaciens AGRL2570 AHJR C-terminal domain with wHTH fold unknown DNA binding, DNA cleavage
06190 - Phage PSA gp51 AHJR none unknown recombinase
- 1395 P. abyssi PAB2104 AHJR second inactive AHJRlike domain, HTH_3 domain unknown DNA cleavage or nicking*
- 4127 S. typhimurium STM4490 Mrr putative inactivated nuclease domain unknown REase
4741 T. volcanium TVN1166 YraN N-terminal membrane helix unknown membrane-associated guardian against foreign DNA
06319 5321 A. tumefaciens AGR_C_8 AHJR none unknown DNA cleavage*
05626 3372 P. abyssi PAB1046 AHJR unknown ~250aa N-terminal domain unknown DNA repair or recombination
- - Phage T4 gp4 TnsA or T7 EndoI none important for the phage T4 head assembly process [52] determination if DNA is packed properly into the phage head
- - P. aerophilum PAE1662 AHJR none unknown type II REase
Discussion
COG1322/pfam02646 (RmuC family)
COG1322/pfam02646 is represented by the RmuC protein from Escherichia coli, which has been implicated in limiting inversions at short-inverted repeats [33]. It has been speculated that RmuC exhibits "limited homology" to human Rad50 protein, centrosome protein pericentrin, nuclear mitotic apparatus proteins and the SbcC proteins, and therefore it may be a structural protein that protects DNA against nuclease action or be itself involved in DNA cleavage at the regions of DNA secondary structure [33]. Our secondary structure predictions revealed that RmuC contains an N-terminal transmembrane helix (aa 1–25), coiled-coil structures (aa 25–200 and ~350–420), a globular α/β domain (aa 200–360), and a disordered C-terminus (aa 420–475). The profile HMM analysis confidently identified a relationship of the central globular α/β domain of RmuC and its homologs to the catalytic domain of the McrC nuclease [23,34](P-value 10-4). This prediction was also confirmed by the fold-recognition analysis, which identified the catalytic domain of REase FokI as the best modeling template, albeit with relatively low scores (mGenTHREADER: 0.4666, SAM-T02: 0.23). Analysis of the multiple sequence alignment (Figure 2) reveals that proteins from the RmuC/COG1322/pfam02646 family exhibit a hallmark PD-(D/E)XK motif associated with the characteristic pattern of predicted secondary structures, which strongly suggests that RmuC is a nuclease that may cleave DNA structures arising during the recombination of short-inverted repeats and thereby thwart the inversion of the internal sequence. We suspect that the initial predictions of homology between RmuC and proteins involved in the structural maintenance of chromosome (Rad50 and SbcC) [33] was due to a spurious similarity in the regions of low sequence complexity, e.g. coiled coils.
Figure 2 Multiple sequence alignment of selected representatives of proteins families predicted to belong to the PD-(D/E)XK superfamily. The selection of representative sequences includes members of all families analyzed in this article (see Table 1). Amino acids are colored according to the physico-chemical properties of their side-chains (negatively charged: red, positively charged: blue, polar: magenta, hydrophobic: green). The putative catalytic Asp/Glu and Lys/Arg residues are highlighted. The variable termini and insertions are not shown. The number of omitted residues is indicated in parentheses. Elements of secondary structure (helices and strands predicted individually for each family) are indicated by H and E, respectively.
COG4469/pfam06054 (CoiA family)
COG4469/pfam06054 is represented by the CoiA protein from Streptococcus pneumoniae, which has been implicated in the DNA uptake process and recombination, but without any clues as to the molecular mechanism of its action [35]. The profile HMM analysis identified a relationship (P-value 10-2) between the central region of CoiA (aa 56–209) and COG4636, a family of predicted nucleases abundant in Cyanobacteria [21]. This prediction was also confirmed by the fold-recognition analysis, which aligned this domain to the recently solved structure of the COG4636 member (Structural Genomics target Tt1808 from Thermus thermophilus Hb8, 1wdj in the PDB, annotated as "to be published") (top matches in FFAS, score -11.0, and mGenTHREADER, score 0.369) and the structure of the Hef nuclease [11](1j22 in the PDB, INBGU top match, score 8.38), which both belong to the PD-(D/E)XK superfamily. While the template structures exhibit unorthodox configurations of catalytic residues (migration of the Lys residue in COG4636 members and topological crossover and migration of the Glu residue in Hef and its homologs), members of the CoiA/COG4469/pfam06054 family exhibit a more typical PD-EXQ variant of the active site, which has been observed in REase BglII [15]. Additionally, CoiA and its homologs possess a conserved Arg residue at the same position as the swapped Lys residue in COG4636. It will be interesting to determine if this residue is involved in substrate binding or catalysis. We hypothesize that in competent pneumococcus cells, the CoiA nuclease degrades one strand of the double-stranded DNA, while the other strand is imported inside the cell. Alternatively, CoiA may be involved in the incorporation of the foreign DNA into the chromosome, but this process is more likely to be carried out by the general homologous recombination machinery, such as RecA.
COG1489/pfam03749 (SfsA family)
COG1489/pfam03749 is represented by the SfsA protein from E. coli, which has been implicated in sugar fermentation stimulation by the so far unknown mechanism and was shown to non-specifically bind to DNA [36]. The profile HMM analysis confidently identified a relationship (P-value 10-5) between the C-terminal domain of SfsA (aa 80–234) and the YraN family of previously predicted endonucleases [23]. This finding was confirmed by the fold-recognition analysis, which confidently identified the Vsr nuclease, a member of the PD-(D/E)XK superfamily (1vsr and 1cw0 in the PDB [37]) as the best template for the sequence of the SfsA C-terminal domain (scores: INBGU: 17.65, mGenTHREADER: 0.428, FUGUE: 3.4). Interestingly, while the Vsr nuclease exhibits a highly unorthodox active site, with the (D/E)XK half-motif replaced by FXH, SfsA and its homologs (including the YraN family) exhibit a typical version of the nuclease catalytic motif, ID-EVK. The fold-recognition analysis revealed also that the N-terminal domain of SfsA (aa 1–80) is a member of the OB-fold superfamily of nucleic acid binding proteins [38]. Thus, both domains of SfsA are likely to be involved in DNA binding, while the C-terminal domain is likely to possess an unforeseen nuclease activity. It will be interesting to determine whether the N-terminal or C-terminal domains or both are important for the activity of sugar fermentation stimulation and how does this activity relate to the predicted nuclease function. Or perhaps SfsA could be a hydrolase involved in another type of reaction than its homologs from the PD-(D/E)XK superfamily? Deletion mutagenesis based on our prediction and site-directed mutagenesis of the PD-(D/E)XK active site may help to address these questions.
KOG2143 (KIA001018 family)
KOG2143 groups together 4 large (>1000 aa) eukaryotic proteins, including the KIA001018 protein from H. sapiens (PMID: 10231032), which has been annotated as a DNA binding protein engaged in DNA repair, probably due to a presence of an easily identified Rad18-like CCHC zinc finger in the N-terminus [39]. However, no suggestions concerning the mechanism of its action have been made. PSI-BLAST searches revealed that KOG2143 has homologs in other Eukaryota (plants and fungi) as well as in several bacteria (shorter versions, about 550 aa). The C-terminal domain (aa 935–1016 in KIA001018) shows significant similarity to HJR families (HHsearch P-values ranging from 10-3 to 10-5) and exhibits the orthodox version of the PD-(D/E)XK motif, perfectly conserved in all members of KOG2143 as well as in their prokaryotic relatives. The fold-recognition analysis identified AHJR (1gef) as one of the possible templates, but with low confidence (mGenTHREADER, score 0.406). The analysis of predicted secondary structures and the preliminary tertiary model of KIA001018 suggest that the putative nuclease domain in KOG2143 and related proteins corresponds to the minimal core of the PD-(D/E)XK fold, i.e. only 4 β-strands and 2 α-helices [40] (Figure 1). Combination of reciprocal HHsearches and fold-recognition carried out for the remaining part of the KIA001018 sequence revealed that KOG2143 members possess a long (300–400 aa) α-helical domain, most probably comprising 5–7 tetratricopeptide-like repeats (TPRs), an unknown domain (~300 aa) with no clear relationship to any protein families and no confident prediction from FR, and the afore-mentioned Rad18-like CCHC zinc finger. The unknown domain and the zinc finger are missing from the shorter prokaryotic members. Such composition of domains suggests that KIA001018 is a DNA binding protein with endonuclease activity, possibly engaged in DNA repair, with a potential to bind other proteins via TPR domains. The analysis of genomic neighborhood did not reveal any obvious associations with other conserved proteins, which could shed more light on the possible function of KIA001018 and its homologs.
COG4804/pfam06250 (YhcG family)
COG4804/pfam06250 is represented by the uncharacterized and functionally not annotated protein YhcG from E. coli, whose homologs can be found in Prokaryota – mostly in Bacteria, but also in Archaea (a few homologs are found in Methanosarcinales) and bacteriophages. The profile HMM analysis confidently identified a relationship (P-value 10-4) between the C-terminal domain of YhcG (aa 235–375) and pfam01939 (DUF91)/COG1637, a family of predicted RecB-like nucleases [32]. Both YhcG and DUF91 families exhibit the same orthodox endonuclease motif: E-D-E(I/L/V)K. Nonetheless, our profile HMM analysis indicates that the endonuclease domain of DUF91 is more similar to YraN-like HJRs rather than to the RecB-like nucleases (P-values 10-6 for YraN and 10-4 for DUF91). This is also confirmed by the fold-recognition analysis for YhcG, which indicates the AHJR structures 1hh1 and 1gef/1ipi as possible templates, however with low scores (INBGU: 1gef, score: 7.7; 1hh1, score: 6.08; FUGUE: 1ipi, score: 2.97; 3DPSSM: 1hh1, score: 7.7). COG4804/pfam06250 members possess an additional N-terminal domain (~200 aa), predicted to be mainly α-helical, but lacking any confident matches to known structures, according to fold-recognition servers. The analysis of genomic neighborhood did not reveal any obvious associations of the yhcG family with conserved genes. In Rickettsiae 5 YhcG paralogs could be found in two gene clusters. Only in Methanosarcina acetivorans an YhcG homolog was found associated with a type I restriction-modification gene cluster, also with another predicted PD-(D/E)XK superfamily member from COG3586 (a putative endonuclease family, whose members were previously identified [32])
Pfam06356 (φH_25 family)
Pfam06356 (DUF1064) groups together functionally uncharacterized and not annotated proteins from tailed bacteriophages and bacteria (Proteobacteria and Firmicutes), and is represented by the hypothetical product encoded by ORF25 from S. aureus phage φH. The profile HMM analysis identified this family as a relative (P-value from 10-3 to 10-5) of YcjD-like proteins – putative members of the PD-(D/E)XK superfamily identified previously [23]. The fold-recognition analysis identified TnsA (1f1z) as one of the possible structural templates (mGenTHREADER: score: 0.302, FUGUE: score: 2.56). The structure of phage T7 Endonuclease I, 1fzr, had slightly better FR scores, but exhibited numerous insertions and deletions in the alignment. Nonetheless, proteins from pfam06356 exhibit a more typical version of the catalytic motif, namely AD-DXK (more similar to T7 Endo I than to TnsA). PSI-BLAST searches using pfam06356 members as queries revealed also a low similarity to the PD-(D/E)XK domain in the Res subunit of Type III restriction-modification systems [41] (alignment only in the endonuclease core region with sequence identity 20% and lower) and to COG3372/pfam5626 (DUF790), an uncharacterized family predicted to belong to the PD-(D/E)XK superfamily in this work, see below. The genomic analysis reveals that pfam06356 members are typically found only in phage genomes or in prophages in bacterial genomes, however, without any strongly conserved neighbors. In only one case, S. aureus subsp. aureus COL prophage L54a, the gene coding a putative DNA:m6 A MTase (M.SauCOLORF346P according to the REBASE nomenclature [42]) was found in a putative operon with a pfam06356 member – hypothetical protein SACOL0347, suggesting that together they may form a novel RM system. We speculate that most members of this family may serve a purpose similar to that of phylogenetically unrelated Endonuclease II of phage T4, namely the degradation of the host DNA upon lytic infection as well as production of recombinogenic fragments [43]
COG5482
COG5482 includes a few functionally uncharacterized and not annotated proteins (~230 aa) mainly from Proteobacteria (only two homologs were found in species from other taxa: Clostridium acetobutylicum and Symbiobacterium thermophilum). The profile HMM analysis suggested that they are remotely related to AHJRs (P-value ~10-3). The fold-recognition analysis has also identified the AHJR structures (1gef/1ipi and 1hh1) as the best templates for modeling of COG5482 sequences, with very good agreement of alignments reported by different servers despite moderate scores (mGenTHREADER: 1hh1, score: 0.543; 1gef, score: 0.413; FUGUE: 1hh1, score: 3.23; 1ipi, score: 3.89; 3DPSSM: 1hh1, score: 3.5; SAM: 1gef/1ipi, scores: 0.17 (0.41). The catalytic domain exhibits the consensus motif: E-CD-ELK. Interestingly, the C-terminal 100 aa was predicted to form a separate domain of the winged helix-turn-helix (wHTH) fold, typically involved in DNA binding [44], which may therefore dictate the target specificity of this putative nuclease. The precise role of these proteins remains to be determined experimentally.
Pfam06190 (gp51 family)
Pfam06190 (DUF944) groups together very small (~100), functionally uncharacterized and not annotated proteins, exemplified by the gp51 protein from the Listeria monocytogenes phage PSA. Homologs of this protein can be found mainly in other phages and in a few bacteria from different taxa, frequently within putative prophages. The profile HMM comparison revealed a remote relationship of Pfam06190 members to AHJRs (P-value ~10-4), while sequence analysis revealed an orthodox version of the catalytic motif, (P/S/C)D-EXK. All fold-recognition servers identified PD-(D/E)XK enzymes as the best templates with high scores (e.g. AHJRs (1hh1, 1gef) identified by INBGU, score 25.16, RecU/PrfA recombinase (1y1o, 1rzn) identified by FFAS, score -10.8). Structure prediction suggests that the members of the gp51 family exhibit a minimal form of the central β-sheet, with only four strands, flanked by three α-helices. However, they seem to possess an additional β-hairpin similar to the element, which in RecU forms a dimerization interface (our unpublished analysis of the 1y1o structure). Interestingly, the analysis of genomic neighborhood reveals that gp51 and its homologs are frequently associated with primases or helicases, which suggests they may be involved in DNA replication. We predict that gp51 and its relatives are recombinases involved in resolution of branched intermediates of phage DNA undergoing replication and/or recombination, similarly to T7 endonuclease I.
COG1395 (PAB2104 family)
COG1395 groups together archaeal sequences of functionally uncharacterized proteins represented by the PAB2104 protein from Pyrococcus abyssi, annotated as putative transcription regulators. As might have been expected based on the database annotation, the profile HMM analysis identified with high confidence (P-value ~10-10) the helix-turn-helix motif (similar to that one from HTH_3 (Xre) family like proteins: pfam01381, smart00530, cd00093) in the central part of the COG1395 sequences. However, using HHsearch we found that the N – terminal domain is related to the family of AHJRs (pfam01870) (P-value ~10-6). The fold-recognition analysis confidently confirmed AHJR (1gef) as the best template for modeling of the N-terminal domain. Preliminary modeling of the N-terminal domain (data not shown) suggests that in PAB2104 and its homologs the catalytic D/E residue from the orthodox (D/E)XK half-motif migrated to the α-helix following the β-strand in which that half-motif was transformed into the "K(I/V)L" form. Interestingly, the profile HMM analysis predicted that the C-terminal domain in COG1395 members may be homologous to the PD-(D/E)XK nucleases as well. This prediction was also supported by the fold recognition analysis, which found AHJR 1hh1 (scores: FFAS: -6.66, INBGU: 13.34, mGenTHREADER: 0.447, SPARKS: -1.74, FUGUE: 4.33, 3DPSSM: 0.061) as the potentially best template structure. However, despite the apparent structural conservation and the presence of hydrophobic residues that confer protein stability, the typical active site is not conserved in this domain (Figure 2). This suggests that the C-terminal domain was generated by intragenic duplication (or fusion of the N-terminal module with a PD-(D/E)XK domain from another source) and then underwent degeneration. Variants of the PD-(D/E)XK domains that lost the active site but presumably retained the ability to bind nucleic acids have been already described (review: [2]).
Thus, COG1395 represents a new family of archaeal putative nucleases with a novel domain architecture: PD-(D/E)XK-HTH-(inactivated)PD-(D/E)XK. It is noteworthy that a related architecture, namely wHTH-PD-(D/E)XK, was observed in the Type IIE REase NaeI, in which the PD-(D/E)XK domain and the wHTH domain, despite the unrelated folds, bind two copies of the same sequence, but only the copy bound to the nuclease domain is cleaved [45]. A PD-(D/E)XK-wHTH fusion (albeit with a reversed order of domains) was also observed in the TnsA transposase [6]. On the other hand, a variant of a tandemly duplicated PD-(D/E)XK domain with a degenerated C-terminal repeat was found in the Sau3AI enzyme [46,47], another Type IIE REase, which also binds as a dimer two copies of the same sequence, but cleaves only the one bound to the pair of "active" N-terminal domains. Tandemly duplicated PD-(D/E)XK domains have been also found in Type IIS restriction enzymes that act as monomers, bind only one asymmetrical DNA target, and use different domains for the cleavage of each strand of the substrate (JMB, JK, and Arvydas Lubys, unpublished data). It will be interesting to study the mode of action of COG1395 members, i.e. whether they act as monomers and dimers, how many copies of the target sequence they bind and whether they act similarly to Type IIE enzymes (like NaeI or Sau3AI) or as nickases, which cleave only one strand of the substrate using the C-terminal active domain.
COG4127
COG4127 groups together 4 uncharacterized proteins, of which only one, STM4490 from S. typhimurium, has been annotated as a predicted restriction endonuclease. PSI-BLAST searches revealed that COG4127 members have numerous homologs in the non-redundant database and are related (E-value ~ 10-10) to the Mrr family of REases [18]. The relationship between the "extended" COG4127 family and the Mrr family was confirmed by the profile HMM analysis (P-value ~10-14). The fold-recognition analysis identified AHJRs (FFAS: 1hh1, score: -13.5; 1gef/1ipi, score: -7.58; 3DPSSM: 1hh1, score: 3.4; SAM-T02: 1gef/1ipi, score: 0.54) as the best templates for modeling of the C-terminal part of the protein, but failed to suggest a good structural match for the N-terminal domain of a similar size. It is noteworthy that some members of COG4127 contain only the N-terminal domain, but either lack the C-terminal Mrr-like domain or possess another, apparently unrelated domain. Analysis of sequence conservation and predicted secondary structures suggested that the N-terminal domain also resembles the PD-(D/E)XK nucleases, but without the hallmark active site, e.g. like in the afore-mentioned COG1395. However, we were unable to confirm the relationship of the N-terminal domain of COG4127 to known PD-(D/E)XK nucleases either by fold-recognition or by HHsearches. Thus, the structure and function of the N-terminal domain remains to be determined experimentally; the availability of members of COG4127 that already lack the C-terminal domain may be particularly useful.
Interestingly, analyses of genomic context revealed that a few members of COG4127 are associated with putative Type I RM systems (R/S/M.CcrMORF620P in Caulobacter crescentus, R/S/M.XorKORF3462P and R/S/M.XorKORF3457P in Xanthomonas oryzae and a putative type I RM system in Methylobacillus flagellatus not yet included in REBASE) or with a Type III RM system (R/M.DetORF1112P in Dehalococcoides ethenogenes). Additionally, another gene cluster from D. ethenogenes contains two representatives of COG4127 (one of them contains only the N-terminal domain) and a member of the SfsA family (predicted to be a nuclease in this work). Another COG4127 member from Azotobacter vinelandii was found in the neighborhood of the putative Mod subunit and a putative protein annotated as a "virulence protein" (COG3943).
COG4741
COG4741 consists of functionally uncharacterized proteins mostly from Archaea, of which only one, the hypothetical product of the locus TVN1166 from T. volcanium is annotated as "a predicted secreted endonuclease distantly related to AHJRs". The profile HMM analysis confidently (P-value ~10-5) identified the relationship of this family to the YraN subfamily of PD-(D/E)XK enzymes [23]. The predicted nuclease domain encompasses 120 C-terminal residues and corresponds to the absolutely minimal core, with only 4 β-strands and 2 α-helices that serve as a scaffold for a well-conserved E-(V/I)D-E(V/I)K motif. The N-terminus reveals a strongly hydrophobic stretch of residues predicted to form a transmembrane helix, which could be either used as a leader peptide to guide the translocation of the nuclease domains through the membrane or it could anchor it to the membrane. Programs for the prediction of transmembrane protein topology HMMTOP [48], TMAP [49], and TMPRED [50] predicted that the nuclease domain has a cytosolic orientation. We speculate that COG4741 members could be released to the environment as toxic agents against other cells, like endonuclease colicins (review: [51]), or be used to guard the cell against the uptake of foreign DNA and/or to cleave the encountered nucleic acids to produce (oligo)nucleotides that can be used by the host.
COG5321/pfam06319
COG5321/pfam06319 (DUF1052) comprises functionally uncharacterized and not annotated short (~160 aa) proteins found almost exclusively in Proteobacteria. The profile HMM analysis confidently identified their relationship to AHJRs (P-value ~10-7). The fold-recognition analysis confirmed the AHJR structures (1gef and 1hh1) as the best templates, however with low scores (data not shown). The reason for this low confidence of fold-recognition could be due to the strong divergence of the C-terminal part of the domain and poor consensus of secondary structure prediction. Nonetheless, COG5321/pfam06319 members exhibit the conserved orthodox AD-E(V/I/C)K motif associated with the characteristic pattern of predicted secondary structures in the N-terminal part of the common fold, which is a strong indication that they are genuine members of the PD-(D/E)XK superfamily, active as nucleases. However, the analysis of genomic neighborhood did not provide any specific clues as to their specific function.
COG3372/pfam5626
COG3372/pfam5626 (DUF790) is represented by the uncharacterized product of PAB1046 gene from Pyrococcus abyssi. Members of this family are found almost exclusively in Archaea and Cyanobacteria. The profile HMM analysis identified their relationship (P-value ~10-4) to Pfam06356 (DUF1064) family of putative endonucleases (identified in this work, see above) and a more remote similarity to AHJRs (P-value ~10-2). All fold-recognition servers with the exception of FUGUE identified with high scores the structure of bacteriophage T7 Endonuclease I (1fzr, 1m0d) as the best templates for modeling of the C-terminal domain of COG3372 (INBGU: 1fzr, score: 60.96; 1hh1, score: 29.12; 1gef, 19.81; FFAS: 1m0d, score -8.15; mGenTHREADER: 1fzr, score: 0.529; 1hh1, score: 0.459; SAM: 1m0d, score: 1.6*10-4; 1fzr, score: 5.4*10-4; 1ob8, score: 0.53; SPARKS: 1m0d, score: -3.62; 1ob8, score: -3.28; 1gef, score: -2.8; 1hh1, score: -2.53; FUGUE: 1hh1, score: 4.48; 1fzr, score: 4.26; 1j24, score: 3.01; 3DPSSM: 1fzr, score: 0.53; 1gef, score: 0.6). However, the fold-recognition failed to identify any confident structural matches for the N-terminal domain (size ~250 aa). Analysis of the genomic neighborhood revealed that COG3372 members are usually associated with hypothetical proteins from COG1061 (SSL2: DNA or RNA helicases of superfamily II). This suggests they may be involved in DNA repair or recombination, but does not exclude their potential selfish character.
gp4 family
The gp4 family is represented by a small (150 aa) protein gp4 from the T4 bacteriophage, annotated as a head completion protein important for the final stages of bacteriophage head assembly process and predicted to be important for the DNA-mediated attachment of independently assembled head and fibers [52]. To our knowledge, only a general functional prediction has been made for this protein and sequences homologous to gp4 have not been grouped into a PFAM or COG family. We identified them by PSI-BLAST, using a Pfam06356 (DUF1064) representative as a query, with a very low E-value of 3.2. The HHsearch analysis carried out for the gp4 family confirmed their relationship to Pfam06356 (DUF1064) (P-value 10-3) as well as to COG3372/pfam5626 (P-value 10-4) and gp51 families (P-value 10-3). Exhaustive PSI-BLAST searches (5 iterations) confidently reported structures of bacteriophage T7 endonuclease I (1fzr, 1m0d) as best templates for the gp4 tertiary structure (E-value = 10-12). The consensus of fold-recognition analysis confirmed the prediction pf the PD-(D/E)XK fold, with TnsA structures (1f1z, 1t0f) as well as T7 Endonuclease I (1fzr, 1m0d) reported as the best structural matches. gp4 is positioned in a cluster of genes encoding structural proteins important for the formation of infectious phage particles and to our knowledge has not been reported to interact with the DNA, but to be necessary for the assembly of protein components of the viral particles. The attachment of independently assembled head and fibers is however DNA-dependent (the process does not occur if there is no DNA loaded to the phage head) (review: [52]). Thus, it is possible that gp4 may be involved in determination if the DNA is packed properly or in some yet unknown process involving the DNA cleavage upon the attachment of the tail and fibers to the head, for instance when one end of the packaged DNA descends into the tail.
PAE1662
PAE1662 from Pyrobaculum aerophilum str. IM2 is an uncharacterized, functionally not annotated protein. We found it during exhaustive PSI-BLAST searches using Pfam06356 (DUF1064) representative as a query, with E-value only about 3.4, but also with a hallmark motif E-AD-ELK. We found a characteristic pattern of predicted secondary structure elements associated with the putative nuclease motif, while the fold-recognition servers reported that PAE1662 is similar to several structures of PD-(D/E)XK superfamily members (mGenTHREADER: 1w36C (RecC), score: 0.651; 1hh1, score 0.594; 1ob8, score: 0.538; SAM: 1hh1, score: 3.8; 1t0f, score: 4.2; 1f1z, score 6.3; 3DPSSM: score: 1hh1, score: 1.5). The consensus predictor Pcons selected 1hh1 (AHJR) as a preferred template. Interestingly, PAE1662 is associated with a gene encoding a putative DNA:m5C MTase (annotated as M.PaeIMORF1659P in the REBASE database [42]). In this context it is noteworthy that a reciprocal PSI-BLAST search revealed no homologs of PAE1662. Altogether, the association with a DNA MTase, the lack of homologs detectable by database searches and a remote relationship to the PD-(D/E)XK nucleases detected by fold-recognition suggest that PAE1662 is a novel putative type II REase, which to date has not been annotated as such in REBASE.
Conclusion
We have carried out a sequence/structure profile HMM search using a new method [29] to identify new members of the PD-(D/E)XK superfamily. Our results revealed the presence of this highly diverged nuclease domain in families of proteins implicated in DNA metabolism but with unknown structure (such as a putative recombinase RmuC, DNA competence factor CoiA, a DNA-binding protein SfsA, a large human protein predicted to be a DNA repair enzyme, predicted archaeal transcription regulators, and the head completion protein of phage T4), and in proteins for which no function or structure was assigned to date (such as YhcG, various phage proteins, and novel candidates for restriction enzymes). The initial predictions were validated by protein fold-recognition, leading to preliminary structural models, which were used as platforms for identification of the potential active sites. Thus far, all known members of the PD-(D/E)XK fold were found to be nucleases, mostly acting on the DNA, or at least have been implicated in nucleic acid metabolism. It cannot be excluded that some of the newly reported members may be hydrolases acting on other substrates (e.g. the SfsA protein involved in stimulation of the sugar fermentation), but we speculate that most of them would cleave DNA. The predictions reported in this article will facilitate the search for the possible substrates.
Our predictions contribute to the reduction of "white spaces" on the sequence-structure-function map of the protein universe and will help to jump-start the experimental characterization of the cellular function of these putative nucleases, as well as the molecular mechanisms of their interactions with the DNA. That we identified several members of the PD-(D/E)XK superfamily with very low scores suggests that more strongly diverged members still await discovery. Our analysis has provided a set of new sequence profiles that may be used to search for even more members of this important group of enzymes and will help to select targets for experimental analyses. For instance, it would be interesting to determine high-resolution structures of the presumably "minimal" members of the superfamily (such as the predicted nuclease domain in the C-terminus of the giant protein KIA001018) or proteins with interesting combinations of domains (such as the fusion of predicted active and inactive nuclease domains with the HTH domain in PAB2104). The availability of the complete catalog of nucleases and the knowledge of their mechanisms of action (and interaction) in the cell under different conditions is essential for the complete understanding of mechanisms that govern the evolution and stability of the genome. Our analysis provides a small, but important step towards this aim.
Methods
Protein structure prediction
Secondary structure prediction and tertiary fold-recognition was carried out via the GeneSilico meta-server gateway at [53]. Secondary structure was predicted using PSIPRED [54], PROFsec [55], PROF [56], SABLE [57], JNET [58], JUFO [59], and SAM-T02 [60]. Solvent accessibility for the individual residues was predicted with SABLE [57] and JPRED [58]. The fold-recognition analysis (attempt to match the query sequence to known protein structures) was carried out using FFAS03 [61], SAM-T02 [60], 3DPSSM [62], BIOINBGU [63], FUGUE [64], mGenTHREADER [65], and SPARKS [66]. Fold-recognition alignments reported by these methods were compared, evaluated, and ranked by the Pcons server [67]. In the cases, where the active site could not be unambiguously identified from the analysis of alignments, we carried out homology modeling and tried to identify the potential catalytic residues based on structural considerations. Accordingly, fold-recognition alignments to the structures of selected templates were used as a starting point for homology modeling using the "FRankenstein's Monster" approach [68], comprising cycles of model building, evaluation, realignment in poorly scored regions and merging of best scoring fragments. The modeling protocol was essentially identical to that published in [21].
Construction and comparison of sequence-structure profile HMMs
A set of previously identified members of the PD-(D/E)XK superfamily (Pfu-HJR, GI:5689160; McrC, GI:585466; Mrr, GI:127320; XisH, GI:1613875; YcjD, GI:1175688; A putative nuclease from Mycobacterium, GI:15609145; PAB0571, GI:7450881, Yor305wp, GI:6324881; YraN, GI: 1176818; AtF16A14.4, GI:8778385; GP69, GI:465349) were used as seeds in PSI-BLAST [69] searches of the non-redundant (nr) database. For each sequence, the search was carried out in two versions: "conservative", with the expectation (e) value threshold for the retrieval of related sequences set to 10-6 and the maximum number of iterations set to 6, and "aggressive", with the e-value threshold of 10-2 and the maximum number of iterations set to 12. The final blast (blunt-end master-slave) alignments together with the predicted secondary structure were used to generate a set of query profile HMMs using HHmake from the HHsearch package [29]. The profile HMMs corresponding to all COG, KOG [30], PFAM [70], PDB [71], CDD [72] and SMART [73] entries were downloaded from the home site of HHsearch (the Department of Developmental Biology (MPI), ). Comparison of the profile HMMs (sequence+structure) was carried out using HHsearch [29], with default parameters.
Abbreviations
aa, amino acid(s); bp, base pair(s); nt, nucleotide; e, expectation; REase, restriction endonuclease; ORF, product of an open reading frame,
Authors' contributions
MF carried out the searches for members of known PD-(D/E)XK families and comparisons of profile HMMs and prepared the initial list of potential new nuclease families. JK validated these predictions using detailed sequence analyses and protein fold-recognition, identified and analyzed "secondary candidates" and participated in writing of the manuscript. JMB provided the set of initial queries for database searches, built the homology models for the validated predictions, interpreted all data and drafted the manuscript. All authors have read and accepted the final version of the manuscript.
Supplementary Material
Additional File 1
An archive file in the zip format, which includes Hidden Markov Models of sequence & secondary structure profiles of PD-(D/E)XK nuclease families analyzed in this article (those known used as queries and newly identified ones). HMMs are in the format of the HHsearch program. The same data are available for download from the authors' FTP server:
Click here for file
Acknowledgements
This analysis was funded by the MNiI (grant 3P04A01124 to JMB). JMB was also supported by the EMBO/HHMI Young Investigator Award. JK was supported by the NIH (grant 1R21CA97899-01A from the NCI). MF was supported by the NIH (Fogarty International Center grant R03 TW007163-01).
==== Refs
Aggarwal AK Structure and function of restriction endonucleases Curr Opin Struct Biol 1995 5 11 19 7773740 10.1016/0959-440X(95)80004-K
Bujnicki JM Pingoud A Molecular phylogenetics of restriction endonucleases Restriction Endonucleases Nucleic Acids and Molecular Biology Gross HJ 2004 14 Berlin, Springer-Verlag 63 87
Kovall RA Matthews BW Structural, functional, and evolutionary relationships between lambda-exonuclease and the type II restriction endonucleases Proc Natl Acad Sci U S A 1998 95 7893 7897 9653111 10.1073/pnas.95.14.7893
Ban C Yang W Structural basis for MutH activation in E.coli mismatch repair and relationship of MutH to restriction endonucleases Embo J 1998 17 1526 1534 9482749 10.1093/emboj/17.5.1526
Tsutakawa SE Muto T Kawate T Jingami H Kunishima N Ariyoshi M Kohda D Nakagawa M Morikawa K Crystallographic and functional studies of very short patch repair endonuclease Mol Cell 1999 3 621 628 10360178 10.1016/S1097-2765(00)80355-X
Hickman AB Li Y Mathew SV May EW Craig NL Dyda F Unexpected structural diversity in DNA recombination: the restriction endonuclease connection Mol Cell 2000 5 1025 1034 10911996 10.1016/S1097-2765(00)80267-1
Nishino T Komori K Tsuchiya D Ishino Y Morikawa K Crystal structure of the archaeal holliday junction resolvase Hjc and implications for DNA recognition Structure (Camb) 2001 9 197 204 11286886 10.1016/S0969-2126(01)00576-7
Bond CS Kvaratskhelia M Richard D White MF Hunter WN Structure of Hjc, a Holliday junction resolvase, from Sulfolobus solfataricus Proc Natl Acad Sci U S A 2001 98 5509 5514 11331763 10.1073/pnas.091613398
Middleton CL Parker JL Richard DJ White MF Bond CS Substrate recognition and catalysis by the Holliday junction resolving enzyme Hje Nucleic Acids Res 2004 32 5442 5451 15479781 10.1093/nar/gkh869
Hadden JM Convery MA Declais AC Lilley DM Phillips SE Crystal structure of the Holliday junction resolving enzyme T7 endonuclease I Nat Struct Biol 2001 8 62 67 11135673 10.1038/83067
Nishino T Komori K Ishino Y Morikawa K X-ray and biochemical anatomy of an archaeal XPF/Rad1/Mus81 family nuclease: similarity between its endonuclease domain and restriction enzymes Structure (Camb) 2003 11 445 457 12679022 10.1016/S0969-2126(03)00046-7
Singleton MR Dillingham MS Gaudier M Kowalczykowski SC Wigley DB Crystal structure of RecBCD enzyme reveals a machine for processing DNA breaks Nature 2004 432 187 193 15538360 10.1038/nature02988
Venclovas C Timinskas A Siksnys V Five-stranded beta-sheet sandwiched with two alpha-helices: a structural link between restriction endonucleases EcoRI and EcoRV Proteins 1994 20 279 282 7892176 10.1002/prot.340200308
Newman M Strzelecka T Dorner LF Schildkraut I Aggarwal AK Structure of restriction endonuclease BamHI and its relationship to EcoRI Nature 1994 368 660 664 8145855 10.1038/368660a0
Lukacs CM Kucera R Schildkraut I Aggarwal AK Understanding the immutability of restriction enzymes: crystal structure of BglII and its DNA substrate at 1.5 A resolution Nat Struct Biol 2000 7 134 140 10655616 10.1038/72405
Pingoud V Sudina A Geyer H Bujnicki JM Lurz R Luder G Morgan R Kubareva E Pingoud A Specificity changes in the evolution of Type II restriction endonucleases: a biochemical and bioinformatic analysis of restriction enzymes that recognize unrelated sequences J Biol Chem 2005 280 4289 4298 15563460 10.1074/jbc.M409020200
Skirgaila R Grazulis S Bozic D Huber R Siksnys V Structure-based redesign of the catalytic/metal binding site of Cfr10I restriction endonuclease reveals importance of spatial rather than sequence conservation of active centre residues J Mol Biol 1998 279 473 481 9642051 10.1006/jmbi.1998.1803
Bujnicki JM Rychlewski L Identification of a PD-(D/E)XK-like domain with a novel configuration of the endonuclease active site in the methyl-directed restriction enzyme Mrr and its homologs Gene 2001 267 183 191 11313145 10.1016/S0378-1119(01)00405-X
Pingoud V Kubareva E Stengel G Friedhoff P Bujnicki JM Urbanke C Sudina A Pingoud A Evolutionary relationship between different subgroups of restriction endonucleases J Biol Chem 2002 277 14306 14314 11827971 10.1074/jbc.M111625200
Tamulaitis G Solonin AS Siksnys V Alternative arrangements of catalytic residues at the active sites of restriction enzymes FEBS Lett 2002 518 17 22 11997010 10.1016/S0014-5793(02)02621-2
Feder M Bujnicki JM Identification of a new family of putative PD-(D/E)XK nucleases with unusual phylogenomic distribution and a new type of the active site BMC Genomics 2005 6 21 15720711 10.1186/1471-2164-6-21
Bujnicki JM Crystallographic and bioinformatic studies on restriction endonucleases: inference of evolutionary relationships in the "midnight zone" of homology Curr Protein Pept Sci 2003 4 327 337 14529527 10.2174/1389203033487072
Bujnicki JM Rychlewski L Grouping together highly diverged PD-(D/E)XK nucleases and identification of novel superfamily members using structure-guided alignment of sequence profiles J Mol Microbiol Biotechnol 2001 3 69 72 11200231
Bujnicki JM Rychlewski L The herpesvirus alkaline exonuclease belongs to the restriction endonuclease PD-(D/E)XK superfamily: insight from molecular modeling and phylogenetic analysis Virus Genes 2001 22 219 230 11324759 10.1023/A:1008131810233
Pingoud V Conzelmann C Kinzebach S Sudina A Metelev V Kubareva E Bujnicki JM Lurz R Luder G Xu SY Pingoud A PspGI, a type II restriction endonuclease from the extreme thermophile Pyrococcus sp.: structural and functional studies to investigate an evolutionary relationship with several mesophilic restriction enzymes J Mol Biol 2003 329 913 929 12798682 10.1016/S0022-2836(03)00523-0
Chmiel AA Bujnicki JM Skowronek KJ A homology model of restriction endonuclease SfiI in complex with DNA BMC Struct Biol 2005 5 2 15667656 10.1186/1472-6807-5-2
Chmiel AA Radlinska M Pawlak SD Krowarsch D Bujnicki JM Skowronek KJ A theoretical model of restriction endonuclease NlaIV in complex with DNA, predicted by fold recognition and validated by site-directed mutagenesis and circular dichroism spectroscopy Protein Eng Des Sel 2005 18 181 189 15849215 10.1093/protein/gzi019
Pawlak SD Radlinska M Chmiel AA Bujnicki JM Skowronek KJ Inference of relationships in the 'twilight zone' of homology using a combination of bioinformatics and site-directed mutagenesis: a case study of restriction endonucleases Bsp6I and PvuII Nucleic Acids Res 2005 33 661 671 15684412 10.1093/nar/gki213
Soding J Protein homology detection by HMM-HMM comparison Bioinformatics 2005 21 951 960 15531603 10.1093/bioinformatics/bti125
Tatusov RL Fedorova ND Jackson JD Jacobs AR Kiryutin B Koonin EV Krylov DM Mazumder R Mekhedov SL Nikolskaya AN Rao BS Smirnov S Sverdlov AV Vasudevan S Wolf YI Yin JJ Natale DA The COG database: an updated version includes eukaryotes BMC Bioinformatics 2003 4 41 12969510 10.1186/1471-2105-4-41
Mulder NJ Apweiler R Attwood TK Bairoch A Bateman A Binns D Bradley P Bork P Bucher P Cerutti L Copley R Courcelle E Das U Durbin R Fleischmann W Gough J Haft D Harte N Hulo N Kahn D Kanapin A Krestyaninova M Lonsdale D Lopez R Letunic I Madera M Maslen J McDowall J Mitchell A Nikolskaya AN Orchard S Pagni M Ponting CP Quevillon E Selengut J Sigrist CJ Silventoinen V Studholme DJ Vaughan R Wu CH InterPro, progress and status in 2005 Nucleic Acids Res 2005 33 Database Issue D201 5 15608177
Aravind L Makarova KS Koonin EV SURVEY AND SUMMARY: Holliday junction resolvases and related nucleases: identification of new families, phyletic distribution and evolutionary trajectories Nucleic Acids Res 2000 28 3417 3432 10982859 10.1093/nar/28.18.3417
Slupska MM Chiang JH Luther WM Stewart JL Amii L Conrad A Miller JH Genes involved in the determination of the rate of inversions at short inverted repeats Genes Cells 2000 5 425 437 10886369 10.1046/j.1365-2443.2000.00341.x
Pieper U Pingoud A A mutational analysis of the PD...D/EXK motif suggests that McrC harbors the catalytic center for DNA cleavage by the GTP-dependent restriction enzyme McrBC from Escherichia coli Biochemistry 2002 41 5236 5244 11955073 10.1021/bi0156862
Pestova EV Morrison DA Isolation and characterization of three Streptococcus pneumoniae transformation-specific loci by use of a lacZ reporter insertion vector J Bacteriol 1998 180 2701 2710 9573156
Takeda K Akimoto C Kawamukai M Effects of the Escherichia coli sfsA gene on mal genes expression and a DNA binding activity of SfsA Biosci Biotechnol Biochem 2001 65 213 217 11272834 10.1271/bbb.65.213
Tsutakawa SE Jingami H Morikawa K Recognition of a TG mismatch: the crystal structure of very short patch repair endonuclease in complex with a DNA duplex Cell 1999 99 615 623 10612397 10.1016/S0092-8674(00)81550-0
Murzin AG OB(oligonucleotide/oligosaccharide binding)-fold: common structural and functional solution for non-homologous sequences Embo J 1993 12 861 867 8458342
Nagase T Ishikawa K Suyama M Kikuno R Hirosawa M Miyajima N Tanaka A Kotani H Nomura N Ohara O Prediction of the coding sequences of unidentified human genes. XIII. The complete sequences of 100 new cDNA clones from brain which code for large proteins in vitro DNA Res 1999 6 63 70 10231032
Bujnicki JM Understanding the evolution of restriction-modification systems: clues from sequence and structure comparisons Acta Biochim Pol 2001 48 935 967 11996004
Davies GP Martin I Sturrock SS Cronshaw A Murray NE Dryden DT On the structure and operation of type I DNA restriction enzymes J Mol Biol 1999 290 565 579 10390354 10.1006/jmbi.1999.2908
Roberts RJ Vincze T Posfai J Macelis D REBASE--restriction enzymes and DNA methyltransferases Nucleic Acids Res 2005 33 D230 2 15608184 10.1093/nar/gki029
Carlson K Kosturko LD Endonuclease II of coliphage T4: a recombinase disguised as a restriction endonuclease? Mol Microbiol 1998 27 671 676 9515694 10.1046/j.1365-2958.1998.00728.x
Brennan RG The winged-helix DNA-binding motif: another helix-turn-helix takeoff Cell 1993 74 773 776 8374950 10.1016/0092-8674(93)90456-Z
Huai Q Colandene JD Topal MD Ke H Structure of NaeI-DNA complex reveals dual-mode DNA recognition and complete dimer rearrangement Nat Struct Biol 2001 8 665 669 11473254 10.1038/90366
Bujnicki JM A model of structure and action of Sau3AI restriction endonuclease that comprises two MutH-like endonuclease domains within a single polypeptide Acta Microbiol Pol 2001 50 219 231 11930990
Friedhoff P Lurz R Luder G Pingoud A Sau3AI, a monomeric type II restriction endonuclease that dimerizes on the DNA and thereby induces DNA loops J Biol Chem 2001 276 23581 23588 11316811 10.1074/jbc.M101694200
Kolade OO Carr SB Kuhlmann UC Pommer A Kleanthous C Bouchcinsky CA Hemmings AM Structural aspects of the inhibition of DNase and rRNase colicins by their immunity proteins Biochimie 2002 84 439 446 12423787 10.1016/S0300-9084(02)01451-7
Milpetz F Argos P Persson B TMAP: a new email and WWW service for membrane-protein structural predictions Trends Biochem Sci 1995 20 204 205 7610486 10.1016/S0968-0004(00)89009-X
Hofmann K Stoffel W TMBASE - A database of membrane spanning proteins segments Biol Chem Hoppe Seyler 1993 347 166
James R Penfold CN Moore GR Kleanthous C Killing of E coli cells by E group nuclease colicins Biochimie 2002 84 381 389 12423781 10.1016/S0300-9084(02)01450-5
Leiman PG Kanamaru S Mesyanzhinov VV Arisaka F Rossmann MG Structure and morphogenesis of bacteriophage T4 Cell Mol Life Sci 2003 60 2356 2370 14625682 10.1007/s00018-003-3072-1
Kurowski MA Bujnicki JM GeneSilico protein structure prediction meta-server Nucleic Acids Res 2003 31 3305 3307 12824313 10.1093/nar/gkg557
McGuffin LJ Bryson K Jones DT The PSIPRED protein structure prediction server Bioinformatics 2000 16 404 405 10869041 10.1093/bioinformatics/16.4.404
Rost B Yachdav G Liu J The PredictProtein server Nucleic Acids Res 2004 32 W321 6 15215403
Ouali M King RD Cascaded multiple classifiers for secondary structure prediction Protein Sci 2000 9 1162 1176 10892809
Adamczak R Porollo A Meller J Accurate prediction of solvent accessibility using neural networks-based regression Proteins 2004 56 753 767 15281128 10.1002/prot.20176
Cuff JA Barton GJ Application of multiple sequence alignment profiles to improve protein secondary structure prediction Proteins 2000 40 502 511 10861942 10.1002/1097-0134(20000815)40:3<502::AID-PROT170>3.0.CO;2-Q
Meiler J Baker D Coupled prediction of protein secondary and tertiary structure Proc Natl Acad Sci U S A 2003 100 12105 12110 14528006 10.1073/pnas.1831973100
Karplus K Karchin R Draper J Casper J Mandel-Gutfreund Y Diekhans M Hughey R Combining local-structure, fold-recognition, and new fold methods for protein structure prediction Proteins 2003 53 Suppl 6 491 496 14579338 10.1002/prot.10540
Rychlewski L Jaroszewski L Li W Godzik A Comparison of sequence profiles. Strategies for structural predictions using sequence information Protein Sci 2000 9 232 241 10716175
Kelley LA MacCallum RM Sternberg MJ Enhanced genome annotation using structural profiles in the program 3D-PSSM J Mol Biol 2000 299 499 520 10860755 10.1006/jmbi.2000.3741
Fischer D Hybrid fold recognition: combining sequence derived properties with evolutionary information Pacific Symp Biocomput 2000 119 130
Shi J Blundell TL Mizuguchi K FUGUE: sequence-structure homology recognition using environment-specific substitution tables and structure-dependent gap penalties J Mol Biol 2001 310 243 257 11419950 10.1006/jmbi.2001.4762
Jones DT GenTHREADER: an efficient and reliable protein fold recognition method for genomic sequences J Mol Biol 1999 287 797 815 10191147 10.1006/jmbi.1999.2583
Zhou H Zhou Y Single-body residue-level knowledge-based energy score combined with sequence-profile and secondary structure information for fold recognition Proteins 2004 55 1005 1013 15146497 10.1002/prot.20007
Lundstrom J Rychlewski L Bujnicki JM Elofsson A Pcons: a neural-network-based consensus predictor that improves fold recognition Protein Sci 2001 10 2354 2362 11604541 10.1110/ps.08501
Kosinski J Cymerman IA Feder M Kurowski MA Sasin JM Bujnicki JM A "FRankenstein's monster" approach to comparative modeling: merging the finest fragments of Fold-Recognition models and iterative model refinement aided by 3D structure evaluation Proteins 2003 53 Suppl 6 369 379 14579325 10.1002/prot.10545
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Bateman A Coin L Durbin R Finn RD Hollich V Griffiths-Jones S Khanna A Marshall M Moxon S Sonnhammer EL Studholme DJ Yeats C Eddy SR The Pfam protein families database Nucleic Acids Res 2004 32 D138 41 14681378 10.1093/nar/gkh121
Deshpande N Addess KJ Bluhm WF Merino-Ott JC Townsend-Merino W Zhang Q Knezevich C Xie L Chen L Feng Z Green RK Flippen-Anderson JL Westbrook J Berman HM Bourne PE The RCSB Protein Data Bank: a redesigned query system and relational database based on the mmCIF schema Nucleic Acids Res 2005 33 D233 7 15608185 10.1093/nar/gki057
Marchler-Bauer A Anderson JB DeWeese-Scott C Fedorova ND Geer LY He S Hurwitz DI Jackson JD Jacobs AR Lanczycki CJ Liebert CA Liu C Madej T Marchler GH Mazumder R Nikolskaya AN Panchenko AR Rao BS Shoemaker BA Simonyan V Song JS Thiessen PA Vasudevan S Wang Y Yamashita RA Yin JJ Bryant SH CDD: a curated Entrez database of conserved domain alignments Nucleic Acids Res 2003 31 383 387 12520028 10.1093/nar/gkg087
Letunic I Copley RR Schmidt S Ciccarelli FD Doerks T Schultz J Ponting CP Bork P SMART 4.0: towards genomic data integration Nucleic Acids Res 2004 32 D142 4 14681379 10.1093/nar/gkh088
|
16011798
|
PMC1189080
|
CC BY
|
2021-01-04 16:27:23
|
no
|
BMC Bioinformatics. 2005 Jul 12; 6:172
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-172
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1771601880510.1186/1471-2105-6-177Methodology ArticleDissecting systems-wide data using mixture models: application to identify affected cellular processes Svensson J Peter [email protected] Menezes Renée X [email protected] Ingela [email protected] Micheline [email protected] Harry [email protected] Department of Toxicogenetics, Leiden University Medical Centre, P.O. Box 9503, 2300 RA Leiden, the Netherlands2 Department of Oncology, Radiology and Clinical Immunology, Academic Hospital, 751 85 Uppsala, Sweden3 Department of Medical Statistics, Leiden University Medical Centre, P.O. Box 9604, 2300 RA Leiden, the Netherlands2005 14 7 2005 6 177 177 8 12 2004 14 7 2005 Copyright © 2005 Svensson et al; licensee BioMed Central Ltd.2005Svensson et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Functional analysis of data from genome-scale experiments, such as microarrays, requires an extensive selection of differentially expressed genes. Under many conditions, the proportion of differentially expressed genes is considerable, making the selection criteria a balance between the inclusion of false positives and the exclusion of false negatives.
Results
We developed an analytical method to determine a p-value threshold from a microarray experiment that is dependent on the quality and design of the data set. To this aim, populations of p-values are modeled as mathematical functions in which the parameters to describe these functions are estimated in an unsupervised manner. The strength of the method is exemplified by its application to a published gene expression data set of sporadic and familial breast tumors with BRCA1 or BRCA2 mutations.
Conclusion
We present an objective and unsupervised way to set thresholds adapted to the quality and design of the experiment. The resulting mathematical description of the data sets of genome-scale experiments enables a probabilistic approach in systems biology.
==== Body
Background
Functional analysis of microarray data, e.g. to reveal enrichment of promoter sequences, metabolic pathways or signalling cascades, requires an extensive selection of differentially expressed genes. Arbitrarily chosen thresholds for fold-changes and/or significance of change are commonly used to split the genes in subsets of alternative (differentially expressed) and null (non-differentially expressed) genes, with acceptable proportions of false positives and false negatives. Which proportion is acceptable depends on the research question. For example, to find markers of a treatment or signatures of a mutation, it is usually enough to fix the FDR-control level [1], and select the most significant alternative genes. However, most differentially expressed genes will not be selected. For other purposes, e.g. when comparing differences between treatments or mutations, there is need for more thorough comparisons. For each gene we would like to know the probability of it being alternative or null. Such an approach will let us select genes that are truly – or truly not – differentially expressed, with associated estimates of false positives.
The differential expression of genes is evaluated after a comparison of gene expression levels, yielding a list of p-values. Commonly, a threshold is then fixed for either the number of genes selected, or for their p-value, and those genes corresponding to the k smallest p-values are classified as alternative. However, because of the large number of p-values involved, even if all features are null some small p-values may be observed due to pure chance. Thus, it is important to consider an estimate for the proportion of alternative features while determining the threshold.
The problem of estimating the proportion of alternative and null features has been handled by several authors [2-5]. In many experimental settings the alternative features make up a considerable fraction. This is a consequence of genes being connected in networks; altered expression of one gene can affect expression levels of multiple targets. Begley and co-workers [6] showed that the methylating agent MMS induced changes in the transcription level of 33% of the genes in the S. cerevisiae genome and oxidizing agent t-BuOOH altered 38% of the genes. In a study of familial and sporadic human breast cancers by Hedenfalk and co-workers [7], Storey and Tibshirani [5] estimated 33% of the genes to be differentially expressed between BRCA1 and BRCA2 mutation positive tumors. However, even though over 1,000 genes were considered to be changed, only 160 genes could be selected with a pFDR of 5% as a consequence of the overlap between alternative and null genes. It is intuitive that using only 16% of the differentially expressed genes for functional analysis is suboptimal, and we will show that this is indeed the case. To detect subtle but coordinated changes in gene expressions, it is advantageous to examine the joint behavior of inter-connected sets of genes [8], which implies the need for a generous selection. Significant changes in cellular processes can be detected in cases where the alteration of individual gene expressions is not significant [9].
The threshold determination has so far been left to the researcher. In practice it is either guided by limitations in the number of features to be verified after being classified as alternative, or by the desired false positive proportion in the final list. However, with the increasing rate at which new experiments are being performed, and decreasing cost of verification experiments, there will be a growing need for unsupervised ways of determining thresholds yielding a list of features with acceptable false positive and false negative rates.
We present here an objective and unsupervised way to set thresholds that are adapted to the quality and design of the experiment. Description of the distribution of both null and alternative p-values as mathematical functions will give the researcher the possibility to select genes depending on the probability of being alternative or null.
Result
A case study
Hedenfalk et al [7] determined gene expression patterns in 21 tumors from breast cancer patients. Seven tumors were sporadic with unknown mutations, whereas the remainder were familial cancers, in which one of the known genes associated with breast cancer was mutated: BRCA1 (7 patients) or BRCA2 (8 patients). RNAs from the three different sources were hybridized to arrays containing 6,512 cDNA clones. After discarding low quality spots [5,7], more than 3,000 clones remained. The measurements from these genes were tested against the null hypothesis that there is no differential expression across two conditions [5]. The resulting p-values can be visualized in density histograms (figure 1A,C and supplementary material). The p-values appear to follow the expected distribution as being composed of a population of truly null (non-differentially expressed) genes with p-values uniformly distributed among [0, 1] and a population of truly alternative (differentially expressed) genes with p-values that tend to be close to zero. Similar shapes of distributions can be generated for all four tested comparisons, BRCA1 to sporadic, BRCA2 to sporadic, BRCA1 to BRCA2 and both BRCA1 and BRCA2 to sporadic.
Figure 1 Distributions of p-values from the Hedenfalk data set. (A,C) Density histograms for the more than 3,000 genes comparing (A) BRCA1-mutated and sporadic tumors or (C) BRCA1- and BRCA2-mutated tumors. The dashed lines show the estimated distributions of the alternative (curves) and null (horizontal lines) genes, and solid lines show the sum the estimated distributions. (B,D) Scatterplots where the p-value of each gene is plotted against the percentile rank. A smooth function is fitted (solid line) and the local maxima of the curvature (dashed line) are used to split the null and alternative populations of genes for the comparisons (B) BRCA1-mutated and sporadic tumors or (D) BRCA1- and BRCA2-mutated tumors.
Decomposing distributions and setting thresholds
To further examine the gene expression responses, we describe the alternative and null distributions as mathematical functions. For this purpose, we plot the genes sorted by their p-values, p, against the percentile ranks (rank position/the total number of p-values), y (figure 1B) to reveal F-1, the inverse of the underlying cumulative density function. F-1 is estimated non-parametrically. When comparing BRCA1 mutation-positive tumors against sporadic tumors, is convex upward on the entire interval y = [0, 1]. The curvature (dotted line in figure 1B) has a major peak (at y0 = 0.30 corresponding to p0 = 0.19) indicating that there is one dominating population of alternative genes. The expression for f(p0) gives us a rough estimated to 0.77. Other methods specifically designed to estimate yielded similar results; Storey and Tibshirani's method [5] led to = 0.83 and the method developed by Schweder and Spjotvoll [4] resulted in = 0.82. Under the assumption that the p-values of the alternative genes follow an exponential distribution, we calculate to 8.3 (see methods section). With the parameters and we can describe the null and alternative density functions (dashed lines in figure 1A). To select differentially expressed genes, we use the critical point p0 as a p-value threshold. As a result, the null hypothesis was accepted for 2,083 genes with a p > 0.19. These genes all have at least a twice larger probability of being null than alternative. In general, since we have a mathematical description of f(p) any point can be chosen as threshold. For instance, we can calculate a point, p1, where genes have at least a two times larger probability of being alternative than null. Hundred-and-eighty genes have a p-value below this threshold of p1 = 0.04, and are selected as alternative. The proportion of false positives, as estimated by the ratio of integrals of f0 and f in the interval [0, p1], amounts to 20%. The pFDR estimated by Storey and Tibshirani [5] is 34%. Notably, in this comparison at a pFDR of 5%, being the proposed threshold by Storey and Tibshirani [5], no gene is selected as being differentially expressed.
Comparing expression data from BRCA2 and sporadic tumors shows similar patterns ( = 0.75 and = 11.1) (supplementary figure 1). Here, using p0 as the p-value threshold 2,168 genes are selected as being null. Only 20% of the genes with p <p0(BRCA1) or p <p0(BRCA2) are in common in both comparisons, suggesting that BRCA1 and BRCA2 affect different target genes.
Information on the genes affected in both BRCA1 and BRCA2 mutation-positive tumors can be obtained by comparing the data from the familial tumors together against the sporadic. The curvature of , revealing the genes in common between BRCA1 and BRCA2 mutated tumors, has two local maxima (supplementary figure 1). Also when the data from BRCA1 is compared to that from BRCA2 mutated tumors (figure 1C–D), the curvature plot has several local maxima. These observations indicate that the original distributions might be constituted of p-values from two or more populations of truly alternative genes in addition to the truly null genes. Each of these groups is represented by a close to linear part in the plot of . The critical points of the curvature of and the corresponding p-values let us estimate the parameters to describe the distributions of the groups.
Simulation study
In order to verify how the method works in a case for which the result is known, a simulation study was performed. It was assumed for simplicity that each alternative p-value followed an exponential distribution with rate λ. For values of λ between 1 and 100, and varying proportions π0 of alternative p-values within [0, 1], a set of 10,000 independent p-values was generated. For each of these sets, an estimate was computed for π0 using the method developed by Storey and Tibshirani [5]. Subsequently, an estimate for λ was determined via a non-parametric approach (see methods section). A graph of the estimate for all combinations of λ and π0 values, is shown in figure 2. The relative error in was determined as and the proportion of false positives and false negatives were calculated by and respectively.
Figure 2 Calculation of p0 and estimations of λ from a simulation study. 10,000 p-values were simulated in a mixture of an alternative and a null distribution. As functions of λ ∈ [1, 100] and π0 ∈ [0, 1] are shown: (A) , (B) the relative error in , (C) the proportion of false positives (PFP) and (D) false negatives (PFN) when selecting the genes with p <p0. NB the inverse arrow of π0 in (C).
The results show that gives a reasonable estimate of λ for 0.20 <π0 < 0.95 and λ > 5 (figure 2). The proportion of false positives obtained for these values is 0.03 ± 0.01 (mean ± s.d.), and the proportion of false negatives is 0.04 ± 0.03. When few alternative features are present (π0 = 0.95), λ cannot be determined. Also when the alternative distribution is close to uniform (λ < 5), fails to correctly estimate λ. For π0 = 0.20, λ is over-estimated and the method yields a conservative threshold, meaning that the proportion of false positives is low on the expense of a high proportion of false negatives. However, π0 being ≤ 0.20 is of relatively little practical interest, as typically by design, studies already contain a large fraction of unchanged genes.
We also compared the proportions of false positives and false negatives between different methods for selecting genes, e.g. while controlling the family wise error-rate (FWER) or the FDR (figure 3). The proportion of false calls largely depends on the separation between alternative and null features (the value of λ) for thresholds based directly on the p-values. As to be expected, when correcting for multiple testing, i.e. controlling the FWER or FDR, the proportion of false positives remains constant.
Figure 3 Comparisons between p0 and other selection methods in a simulation study. 10,000 p-values were simulated in a mixture of an alternative and a null distribution. The proportions of false positives (A) and false negatives (B) are depicted as functions of λ ∈ [1, 100], π0 is set at 0.67. Selecting the genes with p <p0 (black) is compared to: selecting the genes with p < 0.001 (brown) or p < 0.1 (red); controlling the Benjamini-Hochberg FDR at α = 0.05 (dark blue) or α = 0.2 (light blue); the Benjamini-Yekutieli FDR at α = 0.2 (dark green) and the FWER at α = 0.2 (light green). The lines represent means of twenty simulations.
Similar calculations can be made for complex populations with more than one population of alternative features.
Functional analysis
To discern alternated cellular processes, we performed an analysis at gene set level [8,9]. To obtain a description of the function of the affected genes (p <p0) in the BRCA study, the genes on the arrays were associated with Gene Ontology (GO) terms [10]. The annotation of the genes in the original data files were updated to the Unigene 170 build (24 April 2004). Of the informative transcripts on the array, 62% could be attributed to gene sets that represent specific GO terms. Subsequently, the hypergeometric probability of the genes in each gene set being randomly distributed around the threshold, p0, was calculated.
where x is the number of affected genes in the gene set, N is the total number of affected genes, n is the total number of genes in the gene set and m is the total number of genes not in the gene set. In cases of several p0, we chose p0 corresponding to the major peak of the curvature plot, or if two major peaks are close to each other, the one associated with the largest p0.
The significant gene sets and the affected genes in the four comparisons are shown in supplementary tables 1–4.
When comparing BRCA2 and sporadic, only three gene sets were significantly (p < 0.01) altered, one consisting of genes involved in DNA repair. For BRCA1, 15 genes sets are significantly (p < 0.01) altered. The genes involve DNA binding, phosphorylation and cell cycle. Among the six enriched gene sets in the comparison between familial (BRCA1 and BRCA2) and sporadic tumors, the most significant enrichment is that of genes associated with the mitochondria. Almost as significant is the lack of electron transport genes among the selected genes.
We then tried to select genes using alternative methods. When controlling the FWER or the Benjamini-Yuketieli FDR, no genes were selected at the level α = 0.05 and when using the Benjamini-Hochberg FDR only the comparison between BRCA1 and BRCA2 yielded a non-zero list of genes. At the control-level α = 0.2, all four comparisons resulted in lists comprised of 2–450 genes, which is much less (familial versus sporadic) or comparable (BRCA1 versus BRCA2) to the number of genes selected as having p <p0. The significant genes sets after gene set analysis are partially overlapping with the results from the p0 threshold (supplementary tables 5–8).
For visualization of the distributions, we used a modified version of 'gene set enrichment analysis' [8], where the genes were sorted on the p-values to form a sequence. Genes belonging to the tested gene set are attributed a value of and the remaining genes are given the value . The cumulative sum of the sequence is calculated and plotted for a few gene sets in figure 4. The more a gene set deviates from a uniform distribution, the greater absolute value of the sum. For these gene sets, the threshold at p0 coincides with the maximum of the cumulative sum.
Figure 4 Enrichment of gene sets in the BRCA-study. Genes R1, ..., RN (N > 3,000) are ordered on their p-values and the cumulative sum is calculated to determine whether the members of a gene set are enriched. Starting with the top-ranking gene, the sum increases when a gene in the gene set is encountered and decreases otherwise. The gene sets shown here are (A) cell cycle (solid line – BRCA1-sporadic; dashed line – BRCA2-sporadic) and (B) DNA repair (solid line – BRCA1-sporadic; dashed line – BRCA2-sporadic). (C) For the comparison BRCA1/2-sporadic, the gene sets consisting of genes functional in the mitochondria (solid line) and electron transport (dashed line) are depicted. The vertical dotted lines correspond to the thresholds, p0, for BRCA1, BRCA2 and BRCA1/2 compared to sporadic.
Discussion
High-throughput techniques such as gene expression profiling by microarrays allows rapid screening of large amounts of data simultaneously. However, the staggering amount of data produced causes new problems, such as how to determine a meaningful p-value threshold between differentially expressed and unchanged genes. Methods such as Benjamini & Hochberg's [1] or Benjamini & Yekutieli's [11] control false positives proportions, but yield no information about false negatives proportions. Our method can produce a full description of the probability density function for differentially expressed genes, which makes it possible to rationally choose a desired ratio between π0 and 1 - π0 for which the p-value threshold can be calculated.
The most important features of our proposed p-value threshold are that it is data dependent and unsupervised. It relies on an independent estimate of the unchanged genes proportion, . The method works best for π0 ∈ (0.20,0.95) and λ > 5, situations in which the number of alternative features is non-negligible and has a minimum separation from the null features.
The value of λ will increase with the separation of the populations of alternative and null genes. The power of the statistical test will also be reflected on λ, as well as the quality of the data. On commercial platforms with multiple synthetic probes, λ tends to be higher than on microarrays spotted with PCR products (data not shown).
As the tolerable amount of false calls depends on the research question, relaxed selection criteria are needed for questions which require a balance between the inclusion of false positives and the exclusion of false negatives. Conventional ways of selecting genes, such as stringent control of the FWER or FDR, select few, if any, false positives whereas the proportion of false negatives is at its maximum even with good separation between alternative and null features (figure 3). Only when relaxing the thresholds are we able to bring the two rates into equilibrium. Ideally, with no overlapping populations between truly alternative and truly null genes and a single population of alternative genes, the threshold at p0 will select of the total genes as differentially expressed. However, in practice there is always some overlap in p-values between truly alternative and truly null genes. Selection of genes with p <p0 will frequently include many genes which can be useful under some circumstances. For instance, although in the examples of the case study the FDRs are high at p0, almost all alternative genes are included, as can be estimated by . In the functional analysis, the plots of the cumulative sums show that setting the threshold at p0 is biologically valid. Up until these points there is still a contribution of genes in the relevant gene sets. A more stringent selection would neglect the evidence of many genes being marginally differentially expressed. By selecting genes through controlling the FDR, similar results can be obtained with relaxed thresholds. The analysis of biological function that we present on the BRCA data set was expected to reveal groups of genes related to the cellular processes that are affected by either mutant BRCA1 or mutant BRCA2. Indeed for certain predefined gene sets, e.g. genes involved in DNA repair, cell cycle progression or nucleic acid interactions, the proportion of genes with p <p0 is higher than can be expected from a random distribution. The observation that the threshold p0 coincides with the maximum of deviation from zero in figure 4 suggests that p0 is at the breakpoint where the enrichment of the gene set culminates. We noticed that there are more gene sets significantly affected by mutated BRCA1 than mutated BRCA2. Next to "protein phosphorylation", specifically gene sets involved in cell cycle regulation are enriched, including "cell cycle", and the somewhat less significantly induced gene sets "mitosis" and "spindle". As for mutated BRCA2, there is no indication that its impairment affects cell cycle progression; the genes in the gene set "mitosis" appear even to be depleted among the affected genes. The involvement of DNA repair is additionally endorsed by the less significantly induced gene sets "response to stress" and "induction of apoptosis by extracellular signals". From the literature we know that the BRCA network deals with lesions that block or interfere with DNA replication. BRCA1 has a role in DNA damage sensing but its precise function is not known (for review see [12]). BRCA2 is directly involved in DNA repair [13]. Interestingly, when BRCA1 and BRCA2 mutation positive tumors are compared together against sporadic tumors, mitochondrial genes are affected but not electron transport genes. Until recently, the main function of mitochondria was thought to be the provision of energy for the cell through the creation of an electrical potential across its membrane. However, mitochondria are also involved in one route of apoptosis and recently there has been evidence for dysfunctional mitochondria being associated with premature ageing [14]. These are functions that could possibly be connected to the BRCA network.
Conclusion
The method presented here will allow researchers to set unsupervised thresholds to select alternative features from mixed populations of p-values. As the discipline of systems biology evolves, there will be a need to compare global measurements of different levels (RNA, proteins, metabolites, etc.). The evidence of features being alternative can be used as weights in a comparison. In this context, a mathematical description of significant features enables a probabilistic approach to identify affected pathways.
Method
Setting thresholds
The p-values are defined as independent continuous random variables P1, P2, ..., Pn taking values in the interval [0, 1]. Let us represent by f and F the probability density and cumulative probability functions of the distribution of a generic p-value P, respectively. We require that f is twice differentiable.
We represent by f0(p) and f1(p) the p-value densities for a null feature and an alternative feature, respectively. For a generic feature, its p-value density can be written as the mixture
f(p) = π0f0(p) + (1 - π0) f1(p), ∀ p ∈ [0, 1], (1)
where π0 represents the proportion of null features out of the total under study. Note that f0(p) takes the value 1 for all p within [0, 1].
A common way of visualizing all p-values is to make a graph of the sorted p-values according to the features (see for example the solid line in figures 1B and 1D). If there are no alternative features, this line should roughly be a straight line with a 45 degrees angle. The presence of alternative features makes the line more convex, as seen in our examples. This line corresponds to F-1, where F is the cumulative probability function as defined above.
We represent by p0 the p-value threshold defining the largest p-value corresponding to features identified as alternative. We shall define p0 as the maximum of the second derivative of F-1. The second derivative correspond to the curvature of the original function. Using (1), we get that p0 satisfies
Note that there need not be a unique value of p0 satisfying (2). The derivation of the equation is given in the appendix.
For an intuitive understanding, consider the extreme case where the alternative features under study all have p-values equal to zero, whereas the null features have a uniform distribution over [0, 1]. Then F has probability mass at p = 0 equal to the proportion of alternative features (1 - π0), and in the interval (0, 1] is described by a straight line between (0, 1 - π0) and (1, 1). It is clear that the best threshold p0 is the one that selects all features with zero p-values. This corresponds to the turning point of F-1 In practice, the alternative features will not all have zero p-values, but the turning point of F-1, where its second derivative is zero, remains the best threshold to yield a compromise between false positives and false negatives proportions.
Application to exponential distributions
To give some insight about how the proposed threshold works, let us consider the case when the alternative p-values have an exponential distribution. This consists of one of the simplest functional forms for a monotonically decreasing density within [0, 1] with probability density function given by
f1(p) = G(λ)e-λp, 0 ≤ p ≤ 1, (3)
where , a constant guaranteeing that . As λ → 0, f1 approaches the uniform distribution. Equation (2) yields
If we replace this into (3), we obtain
and, by replacing the latter into expression (1) for f, we get
Thus, the threshold is set at a point where the proportion of the alternative features is half the proportion of the null features. In other words, the genes with a p-value less than this threshold all have some evidence against the null hypothesis, whereas very few truly differentially expressed genes are excluded. For many purposes this definition gives a too wide selection, and it is more desirable to have a point where the proportion of alternative genes is higher than the proportion of null genes, e.g. at a point p1 where . At p1, the proportion of alternative genes is twice the proportion of null genes. However, with the functional description given above, any point depending on the ratio between the proportions of null and alternative genes can be found in an unsupervised manner.
The dependency of the threshold p0 on λ and π0, is visualized in figure 5. In this figure, it is clear that the suggested threshold p0 varies with π0 and λ in the desired way: as λ increases, the separation between the null and alternative p-values distributions increases and, therefore, p0 decreases. On the other hand, as the proportion of null genes π0 increases, naturally it becomes harder to identify the alternative genes, and again p0 decreases.
Figure 5 The p-value threshold dependencies on π0 and λ. p-values come from a mixture of π0 null genes and a single population of 1 - π0 alternative genes following an exponential distribution visualized in (A) a 3D plot and (B) a contour plot of p0.
For practical reasons it is desirable to have an estimate of the proportion of false positives in the selection. Provided the described functions, we can integrate f0 and f in the interval [0, max(p)]. The ratio between the two will be an estimate of the proportion of false positives.
Extensions to more than one alternative component are straightforward and given in appendix. Application of the more flexible beta distribution yield comparable results (see appendix).
Non-parametric approach
In practice, there is interest in making as few distributional assumptions as possible so as to yield a robust approach. We suggest estimating F non-parametrically. This can be done by plotting the sorted p-values of all genes against their percentile rank (as the scatterplots in figures 1B and 1D) and estimating F-1 by fitting a smooth function – such as a cubic spline – to the data points. The smoothing parameters are chosen by cross-validation.
In non-parametric regression, the equivalent degrees of freedom for noise (EDF) is defined by EDF = tr{I - A(α)}, where A(α) is the hat matrix associated with spline smoothing with smoothing parameter α. The relation between EDF, the generalized cross validation (GCV) score and the residual sum of squares [15,16] can be written as
The optimal EDF, and thus smoothing parameter, to fit the smoothing spline is found by minimizing the residual sum of squares.
To get the curvature of the spline, , it is differentiated numerically twice. The critical points p0, of this second derivative can be identified.
The R code and additional information is available at .
Authors' contributions
JPS designed and implemented the method and drafted most of the manuscript. RXM accounted for statistical expertise and substantial parts of the draft. MGG, IT and HV supervised the study and participated in manuscript writing. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Supplementary tables 1–4. Gene sets enriched in the comparisons among familial (BRCA1 and BRCA2 mutation positive) and sporadic tumors selecting the genes with p <p0.
Click here for file
Additional File 2
Supplementary tables 5–8. Gene sets enriched in the comparisons among familial (BRCA1 and BRCA2 mutation positive) and sporadic tumors, controling the Benjamini-Hochberg FDR at α = 0.2.
Click here for file
Additional File 3
Supplementary figure 1. Histograms and scatterplots of p-values from comparisons between BRCA2 mutation positive and sporadic tumors, and between BRCA1 and BRCA2 mutation positive and sporadic tumors.
Click here for file
Additional File 4
Appendix. Calculation of derivatives and the application to exponential and Beta distributions.
Click here for file
Additional File 5
The pseudo-code and R code. Methods to calculate p0 and estimate f(p) given a vector of p-values.
Click here for file
Acknowledgements
The authors would like to thank Dr. P. Eilers at the Department of Medical Statistics, LUMC, The Netherlands, for helpful comments and critical reading of the manuscript.
==== Refs
Benjamini Y Hochberg Y Controlling the false discovery rate – a practical and powerful approach to multiple testing Journal of the Royal Statistical Society Series B-Methodological 1995 57 289 300
Ferkingstad E Langaas M Lindqvist B Estimating the proportion of true null hypotheses, with application to DNA microarray Data Preprint Statistics 2003 4
Grenander U On the theory of mortality measurement. Part II Skandinavisk Aktuarietidskrift 1956 39 125 153
Schweder T Spjotvoll E Plots of p-values to evaluate many tests simultaneously Biometrika 1982 69 493 502
Storey JD Tibshirani R Statistical significance for genomewide studies Proceedings of the National Academy of Sciences of the United States of America 2003 100 9440 9445 12883005 10.1073/pnas.1530509100
Begley TJ Rosenbach AS Ideker T Samson LD Damage recovery pathways in Saccharomyces cerevisiae revealed by genomic phenotyping and interactome mapping Molecular Cancer Research 2002 1 103 112 12496357
Hedenfalk I Duggan D Chen Y Radmacher M Bittner M Simon R Meltzer P Gusterson B Esteller M Kallioniemi OP Wilfond B Borg A Trent J Gene-expression profiles in hereditary breast cancer N Engl J Med 2001 344 539 548 11207349 10.1056/NEJM200102223440801
Mootha VK Lindgren CM Eriksson KF Subramanian A Sihag S Lehar J Puigserver P Carlsson E Ridderstrale M Laurila E Houstis N Daly MJ Patterson N Mesirov JP Golub TR Tamayo P Spiegelman B Lander ES Hirschhorn JN Altshuler D Groop LC PGC-1 alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes Nature Genetics 2003 34 267 273 12808457 10.1038/ng1180
Segal E Friedman N Kaminski N Regev A Koller D From signatures to models: understanding cancer using microarrays Nature Genetics 2005 37 S38 45 15920529 10.1038/ng1561
Ashburner M Ball CA Blake JA Botstein D Butler H Cherry JM Davis AP Dolinski K Dwight SS Eppig JT Harris MA Hill DP Issel-Tarver L Kasarskis A Lewis S Matese JC Richardson JE Ringwald M Rubin GM Sherlock G Gene Ontology: tool for the unification of biology Nature Genetics 2000 25 25 29 10802651 10.1038/75556
Benjamini Y Yekutieli D The control of the false discovery rate in multiple testing under dependency Annals of Statistics 2001 29 1165 1188 10.1214/aos/1013699998
Venkitaraman A Tracing the network connecting BRCA and Fanconi anaemia proteins Nature Reviews Cancer 2004 4 266 276 15057286 10.1038/nrc1321
Yang HJ Jeffrey PD Miller J Kinnucan E Sun YT Thoma NH Zheng N Chen PL Lee WH Pavletich NP BRCA2 function in DNA binding and recombination from a BRCA2-DSS1-ssDNA structure Science 2002 297 1837 1848 12228710 10.1126/science.297.5588.1837
Trifunovic A Wredenberg A Falkenberg M Spelbrink JN Rovio AT Bruder CE Bohlooly Y Gidlof S Oldfors A Wibom R Tornell J Jacobs HT Larsson NG Premature ageing in mice expressing defective mitochondrial DNA polymerase Nature 2004 429 417 423 15164064 10.1038/nature02517
Green P Silverman B Nonparametric regression and generalized linear models: a roughness penalty approach 1994 London: Chapman and Hall
Hastie T Tibshirani R Friedman J The elements of statistical learning: data mining, inference, and prediction 2001 New York: Springer
|
16018805
|
PMC1189081
|
CC BY
|
2021-01-04 16:27:23
|
no
|
BMC Bioinformatics. 2005 Jul 14; 6:177
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-177
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1951607640110.1186/1471-2105-6-195Research ArticleSample phenotype clusters in high-density oligonucleotide microarray data sets are revealed using Isomap, a nonlinear algorithm Dawson Kevin [email protected] Raymond L [email protected] Wasyl [email protected] Laboratory for High Performance Computing and Informatics and Section of Molecular and Cellular Biology University of California Davis MCB, One Shields Avenue Davis, CA 95616. USA2005 2 8 2005 6 195 195 17 12 2004 2 8 2005 Copyright © 2005 Dawson et al; licensee BioMed Central Ltd.2005Dawson et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Life processes are determined by the organism's genetic profile and multiple environmental variables. However the interaction between these factors is inherently non-linear [1]. Microarray data is one representation of the nonlinear interactions among genes and genes and environmental factors. Still most microarray studies use linear methods for the interpretation of nonlinear data. In this study, we apply Isomap, a nonlinear method of dimensionality reduction, to analyze three independent large Affymetrix high-density oligonucleotide microarray data sets.
Results
Isomap discovered low-dimensional structures embedded in the Affymetrix microarray data sets. These structures correspond to and help to interpret biological phenomena present in the data. This analysis provides examples of temporal, spatial, and functional processes revealed by the Isomap algorithm. In a spinal cord injury data set, Isomap discovers the three main modalities of the experiment – location and severity of the injury and the time elapsed after the injury. In a multiple tissue data set, Isomap discovers a low-dimensional structure that corresponds to anatomical locations of the source tissues. This model is capable of describing low- and high-resolution differences in the same model, such as kidney-vs.-brain and differences between the nuclei of the amygdala, respectively. In a high-throughput drug screening data set, Isomap discovers the monocytic and granulocytic differentiation of myeloid cells and maps several chemical compounds on the two-dimensional model.
Conclusion
Visualization of Isomap models provides useful tools for exploratory analysis of microarray data sets. In most instances, Isomap models explain more of the variance present in the microarray data than PCA or MDS. Finally, Isomap is a promising new algorithm for class discovery and class prediction in high-density oligonucleotide data sets.
==== Body
Background
The gene expression microarray is an assay that measures expression levels of tens of thousands of genes in parallel on a single chip. Microarrays can be performed from a very small amount of a biological sample, thus allowing for an experimental design involving many sample groups, repeats, dense time series, and samples collected at high-granularity from various anatomic locations. Today, the cost of microarrays is the principal factor limiting the number of samples that can be examined in a particular experiment. In spite of the high cost of microarrays, two thirds of those surveyed by GenomeWeb said they performed more than 200 microarrays and 57% spent more than $100,000 on microarrays in 2003 [2]. Sixty eight percent of these chips were oligonucleotide arrays, mostly Affymetrix chips. With the widespread use of microarrays in basic research and their increasing use in medical diagnostics, biomedical researchers can anticipate lower costs for chips that will lead to more studies utilizing hundreds, if not thousands, of samples. This expansion in sample size will provide researchers with higher resolution insights into biological processes as they are reflected in temporal, spatial, and functional patterns in microarray data sets. To reveal these patterns, several types of pattern recognition and clustering techniques have been developed and applied to microarray data.
A common task in the analysis of large microarray data sets is sample classification based on gene expression patterns. This process can be divided into two steps: class prediction and class discovery. During class prediction samples are assigned to predefined sample classes; whereas class discovery is the process of establishing new sample classes. For example, when gene expression arrays are used for cancer classification, class prediction assigns tumor samples into pre-existing groups of malignancies, while class discovery reveals previously unknown cancer subtypes [3]. The newly discovered tumor subtypes may have different clinical patterns, respond differently to certain drugs, and require more or less aggressive surgical and radiological treatment. Class discovery may also reveal previously unknown processes in cancer biology and define more specific indications for certain drugs. Specific drugs may be used to target newly discovered tumor subtypes, thus facilitating pharmacogenomic drug design and development. These goals will soon become achievable with the results from microarray studies using large samples. Class prediction and class discovery using large data sets will require the evaluation, adaptation, and development of robust mathematical, statistical, and computational tools.
Several mathematical algorithms and computational methods have been applied to class prediction and class discovery in large gene expression data sets. The methods most frequently used are based on clustering techniques such as hierarchical clustering (HC) [4]. HC was used for temporal classification in conjunction with Fourier analysis to detect genes that correlate with periodic changes in synchronized S. cerevisiae cells [5]. HC was also applied to cancer classification, for example breast cancer classification [6]. Other clustering techniques applied to microarray data are the unstructured k-means clustering [7], cluster affinity search (CAST) [8], fuzzy c-means clustering [9], and two-way clustering [10] that was used for the analysis of drug-tumor interactions [11]. Self-organizing maps (SOM) is another technique that is particularly well suited for exploratory data analysis. Unlike HC, SOM does not impose a rigid structure to the data [12]. The utility of SOM was demonstrated in leukemia classification using a weighted voting procedure [3]. Weighted voting classification was also used for predicting chemosensitivity of the NCI-60 tumor collection [13] and human breast cancers [6]. Supervised methods, such as Fisher's linear discriminant analysis, artificial neural networks (ANN), support vector machines (SVM), and boosting, have the advantage that the sample classes are usually defined by another "gold standard" method (e.g., histology, clinical outcome, length of survival, etc). In supervised classification, the choice of classifiers is frequently based on other considerations, e.g., genes that play a role in the pathomechanism of a certain disease or are expressed in a particular tissue. These "enrichment" methods can improve the prediction strength of a classification and decrease the sample number necessary for developing a prediction model. However they also introduce bias that may lead to overtraining or lack of discovery of unexpected sample classes. One of the supervised methods, support vector machines (SVM), has the advantage that it does not make assumptions about the distribution of the data [14]. SVM was tested on ovarian cancer, leukemia, and colon tumor data sets [15] and was also demonstrated to be useful for multi-class cancer classification [16-18]. Artificial neural networks (ANN), another machine learning method, was shown to classify small, round blue-cell tumors (SRBCT) [19]. Tree harvesting, a new method of supervised learning was recently applied for gene expression data [20]. In contrast to these complex procedures, much simpler classifiers may also perform equally well on some data sets. For example, nearest shrunken centroids were applied to SRBCT cells and leukemias [21].
Dimensionality reduction methods are useful in predicting the underlying true dimensionality of a microarray data set and reduce the number of variables applied as inputs to any of the classification procedures. Multi-dimensional scaling (MDS), a linear method, was used for classifying alveolar rhabdomyosarcomas [22], cutaneous malignant melanomas [23], and breast cancers [24]. Principal component analysis (PCA), another dimensionality reduction method, was used for visualizing gene expression maps of central nervous system (CNS) development [25] and classifying embryonal CNS tumors [26]. Probabilistic PCA, a method incorporating biological assumptions in linear factor models, was recently applied to two yeast microarray datasets [27]. Another method, singular value decomposition (SVD), was applied to soft tissue tumors [28], S. cerevisiae cell cycle, and serum-induced fibroblast data sets [29,30]. Generalized SVD (GSVD) was developed to extract conserved gene expression patterns comparable between two different organisms [31]. All these linear methods are inherently sensitive to outliers, missing values, and non-normal distribution. A variant of SVD, robust SVD (rSVD), was recently developed to minimize the effect of these corruptions in the data set [32]. Sammon mapping, a nonlinear mapping algorithm [33] was incorporated in the R multiv package as well as gene expression data processing and exploratory analysis software, such as ENGENE [34]. To address similar issues, Isomap [35] a nonlinear technique of dimensionality reduction originally designed for solving classical problems of pattern recognition, such as visual perception and handwriting recognition, was applied recently to discovery biologically relevant structures in cDNA microarrays [36,37].
We have applied Isomap to the analysis of both breast cancer microarray data sets and prostate cancer proteomics spectra [36] and showed that it consistently outperformed PCA in revealing biologically relevant low-dimensional structures in high-dimensional data sets. Nilsson et al. independently demonstrated the utility of Isomap using a lymphoma and a lung adenocarcinoma cDNA microarray data set [37]. We report here the application of Isomap to three independent Affymetrix GeneChip® data sets. We show that Isomap is capable of discovering temporally, spatially, and functionally relevant structures in gene expression data. To avoid any kind of bias that may be introduced with gene selection or feature enrichment methods, we did not apply any data scrubbing, filtering, or feature enrichment techniques. We also show that Isomap can successfully detect biologically relevant structures even in the background noise of tens of thousands of genes present.
Results
Spinal cord injury data set
The Isomap algorithm was first evaluated on a large data set consisting of 170 rat U34A high density oligonucleotide arrays with 8,799 genes on each array [38]. These data were originally collected for a study on spinal cord injury that illustrated the role of cell cycle in trauma-induced neuronal death [38]. Compared to earlier microarray studies on experimental spinal cord damage, this study applied a lower level spinal cord injury, used individual rat samples rather than pooled spinal cord tissues, evaluated several time points, employed larger Affymetrix arrays, and used both sham-injured and naïve controls.
Unlike this original study, we do not make the assumption that only those genes "consistently expressed above background" are important for further consideration. The DiGiovanni et al. study used a stringent inclusion threshold that included only those genes present in at least 40% of all the samples. In addition to this first filter, a second filter was also applied that eliminated genes that did not have a change in expression level of at least two-fold compared to that of the sham controls, which was determined with Welch ANOVA t-test. Unlike the DiGiovanni et al. study, our analysis considers all 8,799 genes, thus eliminating an important source of potential selection bias. We do not use stringent selection criteria at the level of data scrubbing, because these filters may introduce confounding into the data set, which can lead to separation between sample classes based on subjective filtering criteria rather than existing biological phenomena.
One hundred seventy samples were collected from spinal cords of naïve rats, after sham operations, as well as mild, moderate, and severe injuries. In addition to these five severity classes, samples are also classified into one of three locations, such as below, above, and at the position of the spinal cord injury. The third classification category is the time interval from the injury to the sample collection. These time points are: 0 min, 30 min, 4 h, 24 h, 2 days, 3 days, 7 days, 14 days, and 28 days. All the 170 samples were subjected to Isomap analysis in a completely unsupervised fashion without the a priori knowledge of which classes the sample belongs to. Isomap fits a nonlinear manifold on the 170 samples. This manifold is used to express sample-to-sample distances as path distances on the surface of the manifold rather than the direct Euclidean distance without considering the existence of the manifold. Distances are computed for all pairs of the 170 samples and subjected to multi-dimensional scaling. The result of this procedure is presented in Figure 2 in a three dimensional coordinate system where each sphere represents one sample.
Figure 2 Isomap analysis of the spinal cord injury data set. Three-dimensional Isomap models were generated from 170 rat high density oligonucleotide arrays with 8,799 genes on each array as described in Systems and Methods. Samples were classified based on the A: time, B:location, and C:severity of the spinal cord injury. Sphere diameters express the 95 percentile distance from the nearest neighbor within the group. Neighborhood size: k = 3. D: Residual variance after the application of Isomap, MDS, and PCA models. See text for more detail. Animated images are presented in Additional Files 1, 2, 3.
Figure 2 shows the 170 samples in a three-dimensional model that was generated by Isomap. Samples are classified and colored according to one of three major attributes – time, location, and severity of the injury. The sphere diameters express the compactness of the sample classes; the more compact the class, the smaller the diameter of the sphere. If members of a class spread out in a larger space then larger size spheres are displayed. The sphere diameter is determined as the 95 percentile of the distance distribution from the nearest neighbor within the class. The Isomap model successfully clusters similar samples into groups that are at well-defined locations of the three-dimensional model. The least affected sham-operated samples are shown at the very central core location of the model. By contrast, the most affected samples after moderate to severe injury at 24 h (that is, the most active phase of spinal cord injury) are shown at the peripheries of the model.
The location panel (Fig. 2A) shows that samples from regions below and above the injury are at the central core of the model. At the same time, samples from the injury itself are displayed in the peripheral lobes of the model. There is no visible separation between the unaffected samples and those originating from locations other than the injury itself.
The time panel (Fig. 2B) shows a clear-cut time-dependent separation of the samples along the x-axis. The right-side lobe of the model shows samples at 30 min, 4 h, and 24 h after injury. The earlier samples are at the core of the model. Conversely, samples from later time points are found at more distal locations in the right-side lobe of the model. This time-pattern is in agreement with Di Giovanni et al. [38] who found with temporal clustering that the immediate early genes are over-expressed at 30 min after spinal cord injury. This is followed by genes associated with inflammatory and oxidative stress plateauing at 4 h after injury, as well as cell cycle and neuronal apoptosis-regulating genes at 4–24 h. Isomap effectively separates this early active phase of post-injury damage from the later phase of regeneration that takes place at 48 h – 28 days. Samples from this later regenerative phase, at time points of 2, 3, 7, 14, and 28 days, are displayed in the left-side lobe of the Isomap model. During regeneration, the earlier samples are shown in the more peripheral locations and the later samples are closer to the central core of the model. It is noteworthy that some samples are at the peripheral locations of the model even 28 days after the injury, which is explained by incomplete regeneration and permanent damage caused by a more severe spinal cord injury as will be demonstrated later. In the time panel, Isomap shows the most striking separation between the early and later phases of post-injury events. The right-side lobe of the model contains samples from the early 30 min – 24 h phase of the post-injury damage. In these samples, the dominant events are the spinal cord injury, the secondary biochemical changes, and the endogenous autodestructive events. On the other hand, the left-side lobe of the Isomap model contains samples from the later 48 h-28 day phase when the neuroprotective and recovery promoting phenomena overcome the earlier autodestructive events.
The severity panel (Fig. 2C) shows that samples taken after incremental levels of spinal cord injury appear at increasingly distal locations in the Isomap model. The sham-operated samples are located in the central core of the model surrounded by a shell of samples from rats with mild spinal cord injury. The most peripheral samples in both lobes of the model are those representing moderate to severe injury. In the right lobe of the model, which represents the 30 min – 24 h time points, there is no clear separation between moderate injury samples and those with severe injury. Unexpectedly, in the left lobe that represents the later time points, the samples with moderate injury are more distal than those with severe injury. Although this separation between the moderate and severe injury was not anticipated, the separation between mild and moderate to severe injury is very clear. This separation also answers a question we raised on the time panel. In the left lobe of the model, six samples are visible at 28 days after injury. Three of these samples are in a cluster that is closer to the periphery and the other three samples are closer to the core of the model. The severity panel clearly shows that the three distal samples are those that underwent moderate injury and the more central three samples are those with only mild injury. This finding is consistent with our interpretation that the left lobe of the model represents the regeneration process. While three rats were able to partially recover from a mild injury after 28 days, the other three animals with moderate spinal cord injury did not recover by this time.
For comparison, in addition to Isomap analysis we also used hierarchical clustering [4] to cluster the 170 samples based on their gene expression patterns (Fig. 1). After clustering, the orientation of the branches of the clustering tree is undefined. Therefore, we used SOM to fold the leaves in an order that places the more similar samples closer to each other [12]. Although hierarchical clustering and SOM clusters similar samples into well-defined groups, the result is inferior to the Isomap model because the tree structure can display only one-dimensional structures, which is inadequate for displaying complex multi-dimensional phenomena such as spinal cord injury. In this experiment, several modalities are present, such as location and severity of the injury and the elapsed time after injury. Unlike clustering, Isomap analysis discovers these three major modalities and presents the 170 samples in a fashion that the underlying biological processes can be interpreted based on these three modalities.
Figure 1 Hierarchical clustering of samples in the spinal cord injury data set. One hundred seventy rat high density oligonucleotide arrays were clustered with hierarchical clustering as described in Systems and Methods. The samples are annotated on time, location, and severity of the injury.
Rat multiple tissue gene expression data set
Isomap was applied to another data set containing 122 samples from 11 peripheral tissues and 15 brain regions from three common outbred rat strains, such as Wistar, Wistar Kyoto, and Sprague Dawley. The original study was performed to characterize physiological expression levels in these anatomical locations, to find genes that are important in certain brain regions, and to explain the phenotypic variations between rat strains [39]. This data set consists of 122 Affymetrix rat U34A arrays with 8,799 genes on each array.
Using Affymetrix MAS5, Walker et al. applied several filters on the genes, such as t-test p-value less than 0.05, at least two-fold change of expression levels, and minimum expression thresholds. Additionally, they used Rosetta Resolver with two filters such as Resolver ANOVA p-value less than 0.05 and at least two-fold change of expression levels. Our Isomap analysis of this data set used all genes to compute sample-to-sample differences without the use of input filters. Although our approach may increase the noise in the data set, it also eliminates any potential source of selection bias. We also avoided using any tissue-enrichment techniques based on a priori knowledge. Isomap analysis was carried out in a completely unsupervised fashion.
We were surprised by the ability of Isomap to sort several rat tissue samples in a fashion that reflects the topological anatomy of the source tissues (Fig. 2A). It is apparent in this figure that duplicate samples from the same tissue are displayed at almost identical locations on the Isomap map. For example, on the top left side of the figure, duplicates from small intestine, large intestine, and endothelial samples are presented as three dots. These three tissues are displayed close to each other, which is expected since the primary components of all three are epithelial cells. Nearby is displayed the kidney, a large part of which is also derived from epithelial cells. Below these samples on the mid left part of the figure close to each other are the spleen and thymus samples, two organs of the immune system. On their right is the bone marrow which is the primary site of blood cell generation and therefore functionally related to the spleen and thymus. On the bottom left part of the figure, close to each other are the two major muscle tissues, such as the skeletal muscle and the heart muscle. At the bottom of the figure is the cornea that is an ectodermal tissue close to the brain samples that are also ectodermal in origin and farther away from the endodermal and mesodermal tissues mentioned above. The two endocrine glands of the brain, such as the pineal and pituitary glands, are somewhat separate from the other parts of the brain.
Figure 3D is a three-dimensional Isomap model of an enlarged region of panel A. The bottom left corner of this figure is the cornea below the primary cortical neurons (PCN). Right of the PCN is the dorsal striatum (DS) and the nucleus accumbens. Walker et al. were specifically interested in studying drug-seeking behavior; therefore separate samples were collected from the core and shell regions of the nucleus accumbens. In our Isomap model, these samples are displayed very close to each other. Above the nucleus accumbens is displayed the ventral tegmental area (VTA) and the amygdala (A). Near to the top of the figure are the dorsal raphe (DR) and the hypothalamus (H). Left of the hypothalamus is the pituitary, an endocrine gland under the control of the hypothalamus. Even farther left is the pineal gland, another endocrine organ related to the pituitary gland. On the right middle part of panel D, are two samples from the locus ceruleus presented at almost identical locations in the Isomap model. In the top right portion of panel D are cortical locations that are displayed in panel E in a two-dimensional Isomap model.
Figure 3 Isomap analysis of the rat multiple tissue gene expression data set. Three- and two-dimensional Isomap models were generated from a data set containing 122 samples from 11 peripheral tissues, 15 brain regions of three common out-bred rat strains, such as Wistar, Wistar Kyoto, and Sprague Dawley as described in Systems and Methods. Panel A shows an overview of a two-dimensional model using k = 3 nearest neighbors. Panel B shows the residual variance after the application of Isomap, MDS, and PCA models. Panel C shows a 2-dimensional comparison map generated by Principal Component Analysis. This panel demonstrates that PCA, unlike Isomap, is less accurate in mapping the different tissues correctly. Arrows show that (1) the skeletal muscle and kidney samples are overlapping the forebrain samples, (2) the thymus and spleen samples overlap the nucleus accumbens samples, and (3) the small intestine samples are mapped at a distance from the (4) large intestine samples. Panel D shows a portion of panel A with a three-dimensional model with k = 4. Panel E presents another region of the overview map representing the central brain structures. The neighborhood graphs demonstrate the k = 3 nearest neighbors. Panel F presents the anatomic locations of the samples on brain slice images from the Neuroscience Division, Regional Primate Research Center, University of Washington: BrainInfo (2000) . The abbreviations are: A, amygdala; ACN, amygdala central nucleus; CMFC, cortex minus frontal cortex; DR, dorsal raphe; DRGF, Fisher dorsal root ganglia; DS, dorsal striatum; endoth, endothel; FC, frontal cortex; H, hypothalamus; LIntestine, large intestine; LocCer, locus ceruleus; NAccumb, nucleus accumbens; PCN, primary cortical neurons; PFC, prefrontal cortex; Pineal, pineal gland; Pituit, pituitary gland; SIntestine, small intestine; SkelMusc, skeletal muscle; Thym, thymus; VS, ventral striatum; and VTA, ventral tegmental area. See text for more detail.
Figure 3E is a two-dimensional Isomap model of another enlarged region of panel A. This panel shows the neighborhood graphs used for building the Isomap model. On the left side of this panel is represented the prefrontal cortex (PFC) that is connected to the temporal/parietal/occipital cortex (cortex minus frontal cortex, CMFC). At the bottom of this panel are the samples from the frontal cortex (FC) below two hypothalamus (H) samples. Above these are samples from the amygdala (A), the amygdala central nucleus (ACN), which is a part of the amygdala, the ventral striatum (VS), and the dorsal striatum (DS). Not circled are two samples from the Fisher dorsal root ganglia (DRGF). Figure 3E also shows that most of the amygdala samples are close to the ACN. Unexpectedly, two other amygdala samples are in a very different location showed on panel D close to the VTA. Similar to the amygdala, the hypothalamus (H) is also shown at two different locations; two hypothalamus samples are shown in panel E below the amygdala, and two other samples are in panel D close to the dorsal raphe (DR).
Isomap analysis projects the hypothalamic samples onto two different well-defined locations in the Isomap model. This separation of the hypothalamus samples may reflect an unequal contribution of the different hypothalamic regions to the four hypothalamic samples and provides an example of class discovery. The hypothalamus has several anatomic regions, one of which is the periventricular hypothalamus that is functionally related to the pituitary as well as to the autonomic areas in the brain stem and the spinal cord. Another hypothalamic region, the medial hypothalamus has several connections with the medial division of the amygdala. The third hypothalamic region, the lateral preoptic hypothalamus, has a very complex anatomy with many fibers passing through this region. These subclasses may be explained by spatial or temporal differences. The hypothalamus samples on panel D close to the pituitary samples may originate from samples that are rich in periventricular hypothalamus. On the other hand, samples on panel E close to the amygdala samples may be richer in medial hypothalamus. The separation of the hypothalamus samples into two well defined classes may be explained by not only anatomical but temporal differences. Many of the hypothalamic nuclei synthesize several neurotransmitters and hormones, such as corticotropin-releasing hormone (CRH) and other related releasing hormones. The levels of these molecules and the activity of particular hypothalamic neurons may vary by time depending on the circadian rhythm, stress, food intake, emotional state, estrus cycle, and other factors. The two hypothalamic subclasses discovered by Isomap may be different because of these temporal variations.
Most of the amygdala (A) samples are projected on the Isomap map at a location between the cortical regions and the ventral striatum (Fig. 3E). Two samples, on the other hand, are projected at the ventral tegmental area (VTA) and the nucleus accumbens (panel E). These two amygdala samples are clearly different from the rest of the amygdala samples. This separation is consistent with anatomical dissimilarities between different parts of the amygdala. The largest portion of the amygdala, the basolateral nuclear complex, primarily consists of pyramidal and stellate neurons similar to those in the cerebral cortex. In fact, Isomap found most of the amygdala samples in the proximity of the cerebral cortex. Another part of the amygdala, the centromedial group, including the central nucleus (ACN), is connected to the bed nucleus through fibers called the stria terminalis. Although the bed nucleus is anatomically closer to the hypothalamus, it is histologically very similar to the amygdala. This separation demonstrates a potential class discovery by Isomap, where the discovered classes are in agreement with anatomically, histologically, and functionally well defined structures within the hypothalamus and the amygdala. Some of the hypothalamus and amygdala samples potentially contain more or less tissue from different parts of these anatomically heterogeneous brain structures, which leads to different classifications based on the major constituent.
As a comparison, we analyzed the multiple tissue dataset also with Principal Component Analysis (PCA) (Fig. 3C). Unlike Isomap, PCA was unable to separate the skeletal muscle, kidney, and bone marrow samples from each other (arrow 1). With PCA, the thymus and spleen samples overlap the nucleus accumbens samples (arrow 2); and the small intestine (arrow 3) and large intestine (arrow4) samples are mapped at a distance from each other. This comparison demonstrates that Isomap outperforms PCA in mapping the different tissue samples on a 2-dimensional space. This difference in performance of the two methods is also supported by the higher residual variance of the PCA model compared to that of the Isomap model (Fig. 3B).
High throughput drug screening data set
The Isomap algorithm was also applied to a high throughput drug screening data set [40]. The goal of this study was to develop a general approach for identifying gene expression signatures as surrogates for cellular states in high throughput drug screening experiments. Particularly, Stegmaier et al. screened 1,739 chemical compounds to identify those capable of inducing terminal differentiation of acute myeloid leukemia (AML) cells. They exposed undifferentiated HL-60 samples (undiff) to several chemical compounds. The names and abbreviations of the chemical compounds selected for microarray analysis are listed in Table 1. In addition to the HL-60 samples, Stegmaier et al. also included primary acute promyelocytic leukemia (APL), primary patient AML, normal human neutrophil (poly), and normal human monocyte (mono) cells (Table 2). This gene expression data set consists of 86 human genomic U133A Affymetrix arrays with 18,400 transcripts and variants on each array and 30 human 6800 arrays with 7,129 genes on each array.
Table 1 Chemical compounds and their abbreviations used in the high throughput screening dataset
Apo (R)-(-)-apomorphine HCl
ATRA all trans retinoic acid
Caff 8-(3-chlorostyryl) caffeine
Cyc7p5 cyclazosin HCl
DMSO dimethyl sulfoxide
EGFR 4,5-dianilinophthalimide
Erythro erythro-9-(2-hydroxy-3-nonyl)adenine HCl
Keto 16-ketoestradiol
Methyl α-methyl-L-p-tyrosine
Perg pergolide methanesulfonate
Phen 1,10-phenanthroline
PMA phorbol-12-myristate-13-acetate
Scop (-) scopolamine methyl bromide
Sulma sulmazole
VitD calcitriol
5FU 5-fluorouracil
5FUD 5-fluorouridine
Table 2 Cells and their abbreviations included in the high throughput screening dataset
AML primary patient acute myeloid leukemia cells
APL primary acute promyelocytic leukemia cells
undiff HL-60 cell line
mono normal human monocytes
poly normal human neutrophils
Figure 4 shows a two dimensional Isomap model built from 86 human U133A microarrays. Panel A presents a low resolution overview of this model. The 86 samples are distributed in a Y-shape with the untreated patient AML samples and normal neutrophils clustered at one end of the Y and the normal monocytes at the other. The two perpendicular arrows on this panel represent the granulocytic and monocytic differentiation. The former arrow points from the HL-60 cells (undiff) to the normal neutrophils (poly) and the latter one points to the normal monocytes (mono). As expected, HL-60 cells treated with PMA, a well characterized inducer of monocytic differentiation, are clustered close to the normal monocytes. It is noteworthy that the gene expression patterns of the primary patient AML cells are very different from that of the HL-60 cell line.
Figure 4 Isomap analysis of the high throughput screening data set U133A arrays.Two-dimensional Isomap models were generated from 86 human U133A high density oligonucleotide arrays with 22,283 genes on each array as described in Systems and Methods. Panel A shows an overview map of the Isomap model. Arrows point to the directions of the monocytic and granulocytic differentiation. Panels B, C,and D are zoomed in regions of panel A. Panel E shows Isomap analysis of the monocytic and neutrophilic differentiation markers as described in Systems and Methods. Neighborhood size k: = 4. Panel Fshows the residual variance after the application of Isomap, MDS, and PCA models. Abbreviations of compounds and cell types are listed in Table 1 and 2, respectively. See text for more detail.
Figure 4B, an enlarged portion of panel A, focuses on the monocytic end of the Y distribution showing HL-60 cells treated with several chemical compounds. These compounds, Apo, Keto, Cyc7p5, Sulma, Erythro, DMSO, Methyl, Phen, EGFR, and PMA (Table 1) change the gene expression pattern of HL-60 cells to become more or less similar to normal monocytes; therefore these compounds are potential inducers of monocytic differentiation. ATRA is displayed on the left side of this panel and represents an effect that is more similar to granulocytic than monocytic differentiation.
Figure 4C, another enlarged portion of panel A, shows an area of the Y-distribution around the normal granulocytes (poly). In addition to the granulocytes, this area also contains duplicates of APL samples before and after several drug treatments. The effect of drugs on APL, as measured by the change of gene expression levels, is less compared to the effect of the same drugs on the HL-60 cells. The result of Keto- and Phen-treatment on APL cells is different from that of ATRA; while EGFR and Erythro causes only a minor change in the expression pattern of APL cells.
Figure 4D zooms in the area of the Y-distribution that is surrounding the undifferentiated HL-60 cells (undiff) at the third end of the Y. Two compounds, 5-FU, and 5-FUD, cause a change in the gene expression pattern that is different from monocytic and granulocytic differentiation. The Scop- and Caff-treated samples are very similar to each other and represent only a small change from the untreated HL-60 cells. The effect of two other compounds, Apo and Perg on HL-60 cells is unique and cannot be classified as only monocytic or granulocytic. At the same time, VitD is displayed close to the ATRA-treated samples in the left upper corner of panel D, which suggests that calcitriol's effect on HL-60 cells is similar to that of ATRA. This functional similarity is supported by the known similarities of action between these two compounds. Both chemicals bind to nuclear receptors, VitD binds to vitamin D receptors (VDR) and ATRA binds to retinoic acid receptors (RAR). Both of these receptors heterodimerize with retinoid X receptor (RXR) and act as hormone-dependent transcription factors.
Stegmaier et al. computed "monocyte and neutrophil scores" based on four parameters: interleukin 1 receptor antagonist (IL1RN) and secreted phosphoprotein 1 (SPP1) for the monocyte signature genes and autosomal chronic granulomatous disease protein (NCF1) and orosomucoid 1 (ORM1) for the neutrophil signature genes. In a supplementary table, these four values were presented for undifferentiated HL-60 cells and cells treated with one of 16 compounds. In Figure 4E, we present a two dimensional Isomap model built only from these four parameters. Arrows point to the directions of the granulocytic (left) and monocytic (right) differentiation. It is noteworthy that our unbiased model, using the expression levels of all the 18,400 transcripts and variants as input, offers a very similar result to the model using only four signature genes. In both models, the monocytic and granulocytic differentiations are revealed and the various chemical treatments cause similar changes relative to the two main dimensions. In both maps, Caff and Scop are proximal and represent only a small change compared to the undifferentiated HL-60 cells (undiff). Perg, 5-FU, and 5-FUD cause more change and Erythro, Apo, EGFR, Phen, ATRA, and VitD are all mapped in similar relative locations in panel A and E. The only major difference between the two models is in the location of Sulma and Methyl. However, in both cases these two compounds are mapped close to each other. All these similarities underscore Isomap's ability to recognize treatment-related changes in gene expression patterns even when no a priori information is available about which genes are good predictors of a biological process or difference. Isomap can detect changes in gene expression signals in the background noise of tens of thousands of other genes.
Figure 5 presents the two- and three-dimensional Isomap models built from 30 Affymetrix human 6800 arrays. Panel B is shown to demonstrate the individual samples and the separation between sample classes. Although the sample number is relatively low, Isomap detects good separation between the patient AML, normal monocyte (mono), neutrophil (poly), PMA-treated, and untreated undifferentiated HL-60 sample classes. The ATRA-treated HL-60 class partially overlaps with the untreated HL-60 class. This is in agreement with the result of the U133A arrays that showed that ATRA has less effect than PMA on the gene expression levels in HL-60 cells. Any of these changes is less than the difference between the cell types, such as HL-60, AML, neutrophils, and monocytes. In the 6800 arrays, similar to the U133A arrays, the PMA-treated cells are on the monocytic differentiation axis and the AML samples are mapped on the granulocytic differentiation axis. Within the PMA-treated group, samples after 4 h, 12 h, and 24 h treatment are grouped together. As expected, the 4 h group is located at the right side of the PMA cluster closer to the untreated HL-60 cells and the 24 h group is located at the left side of the PMA cluster closer to the normal monocytes.
Figure 5 Isomap analysis of the high throughput screening data set human 6800 arrays. Isomap models were generated from 30 human 6800 high density oligonucleotide arrays with 7,129 genes on each array as described in Systems and Methods. Panel A shows a three-dimensional Isomap model. Panel B shows a two-dimensional Isomap model. Arrows point to the directions of the monocytic and granulocytic differentiation. Panel C shows the residual variance after the application of Isomap, MDS, and PCA models. Neighborhood size: k = 5. See text for more detail. Animated image is presented in Additional file 4.
Discussion
In this study, we use three Affymetrix high density oligonucleotide microarray data sets to demonstrate Isomap's ability to discover relevant structures in an unbiased fashion. In the rat multiple tissue gene expression data set, the structures revealed correspond to the anatomical topology of the source organs and tissues [39]. Similarly, in the high-throughput drug screening data set, the structures match the monocytic and granulocytic differentiation patterns of myeloid leukemia cells treated with various chemical compounds [40]. We also demonstrate with a complex spinal cord injury data set that Isomap reveals the three main modalities (e.g., location, time, and severity of the injury) of the experimental design [38]. This Isomap model (Fig. 2) is superior to hierarchical clustering (Fig. 1) because Isomap can model structures of higher dimensionality while hierarchical clustering is limited to the discovery of only one-dimensional structures. Clustering is limited in two significant ways; it requires well-separated data and linear correlation. The only relationship clustering can detect is a one-on-one relationship when pair-wise linear comparisons are made [41]. Isomap is a nonlinear algorithm capable of analyzing microarray data sets that are nonlinear in nature. The sources of these nonlinearities may originate from outliers, missing values, and non-normal distribution; all common events for measurements in biological systems.
In all of our examples, Isomap is applied to the expression level values of all of the genes in the microarray data set. To avoid any kind of selection bias, no filtering or data scrubbing procedures were applied. Even under these conditions, we demonstrate that Isomap is capable of finding the biologically relevant structures within noisy data sets. The experimental noise may come from the measurement error and the variations of expression levels of tens thousands of genes. When Isomap is used in production mode, the experimental noise may be decreased by eliminating the genes that are not expressed in the studied tissues and those which have very stable and unchanging expression levels in all of the samples. These procedures will further improve the resolution of the Isomap models and facilitate class discovery by Isomap. The classification power of Isomap may be improved when it is used in combination with enrichment procedures that weight the gene expression levels of different genes dependent on other surrogate information coming from the knowledge of chemical pathways in which the gene plays a role, the tissues in which the gene is expressed [42], and the gene expression patterns that are evolutionarily conserved between different organisms [39,43].
In addition to Isomap, all the three datasets were also analyzed with PCA and MDS. Residual variance curves are presented for all the three datasets. In all instances, PCA and Isomap outperformed MDS and in most instances, Isomap worked the best out of the three methods. In the spinal cord injury dataset, PCA and Isomap give comparable results (Fig. 2D). However with 3–4-dimensional models, Isomap performs better than PCA; and at dimensions over 9, PCA worked better. In the multiple tissue dataset, Isomap outperforms PCA in the range of 2–20 dimensions (Fig. 3B). Similarly, in the two drug screening datasets, Isomap outperforms PCA at dimension 3 and over (Fig. 4F) and at every dimension (Fig. 5C). Typically, Isomap has a better performance in explaining the variance of microarray data at lower dimensions (Fig. 6). This advantage of Isomap is eliminated at higher (over 10) dimensions. The presented examples show that Isomap works well for datasets with several samples that present gradual changes from one another. However, when a few very distinct classes of samples exist then the more classical methods, e.g., PCA work better. Isomap needs samples to be present along the geodesic surface of the manifold. In our examples, many severity levels of spinal cord injury were considered, as well as multiple tissues and multiple chemical compounds, which was the source of Isomap's success. Another prerequisite for a successful application of Isomap is a large enough number of samples. Isomap will become really useful when datasets with hundreds or thousands of microarrays need to be analyzed.
Figure 6 Residual variance differences between PCA and Isomap models.Typically, Isomap has a better performance explaining the variance in microarray datasets at lower dimensions. By increasing the number of dimensions being considered, this advantage will be eliminated.
The structures discovered by Isomap help to understand biological phenomena underlying these structures. Visualization tools of Isomap models may be applied as data mining tools for microarray data sets. Similar tools were used for modeling co-regulated genes in C. elegans with VxInsight [44] and generating high-resolution temporal models of CNS development [25]. Moreover, Isomap is a promising tool of unbiased class discovery because it is performed in a completely unsupervised manner. This is a key advantage of the Isomap algorithm considering the limited number of class discovery tools [23,45,46] compared to the abundance of class prediction methods. The Isomap algorithm is capable of revealing structures in microarray data sets, which may provide insight into underlying biological networks [4,5,7,22,25,29,30,42,43,47].
Another application of the Isomap algorithm is predicting class membership. We previously demonstrated the use of Isomap for class prediction in a breast cancer outcome microarray data set [36]. The selection of good classifiers that are robust, insensitive to outliers, medically interpretable, having high generalization power, and based on the smallest possible subset with the maximal discriminatory features [48] is important and should be evaluated for Isomap in the future. Aclass prediction model using the Isomap algorithm does not need to use all the genes in the microarray. Somorjai et al. argue that no matter what feature selection approach is used to the microarray data, generally 50 or more features are needed for classification [48]. In this paper, we show that only four genes are used as input for an Isomap model (Fig. 4E). Isomap is a type of dimensionality reduction method and may be evaluated as input to supervised machine learning techniques, such as ANN and SVM. Besides gene expression microarrays, Isomap can be used on other types of biological data, such as genomic, proteomic, and metabolomic data sets to reveal low dimensional structures related to diet-genome interactions, genotype-disease associations, and drug-gene-disease relationships.
Conclusion
Isomap, a nonlinear dimensionality reduction algorithm, discovers low-dimensional structures embedded in high-dimensional Affymetrix high-density oligonucleotide microarray data sets. These structures correspond to and help to interpret underlying biological phenomena present in these data. Our work provides examples of three experiments with temporal, spatial, and functional processes revealed by the Isomap algorithm. Visualization of Isomap models helps to understand these processes and provides means of data mining from gene expression data sets. The low-dimensional models generated by Isomap help to reveal new sample classes and are potentially useful for class prediction. In summary, Isomap is a promising new algorithm for the unbiased analysis of high-density oligonucleotide microarray data sets.
Methods
Datasets
Spinal cord injury data set: A large data set consisting of 170 Affymetrix rat U34A high density oligonucleotide arrays with 8,799 genes on each array was accessed at the Gene Expression Omnibus (GEO) GSE 464. Rat multiple tissue data set: Another data set consisting of 122 Affymetrix rat U34A arrays with 8,799 genes on each array was accessed at the GEO GSE 952. High throughput screening data set: A gene expression data set based on high throughput screening (GE-HST) was accessed at the GEO GSE 995. This series is a combination of three other series: GSE 976, 982, and 985. The accessed data set contains 86 human genomic U133A Affymetrix arrays with 18,400 transcripts and variants on each array and 30 human 6800 arrays with 7,129 genes on each array.
Probe-level microarray data analysis
All microarray data sets were downloaded to a local computer in Affymetrix CEL file format. Some of these files were text files, others were binary CEL files. Probe-level data analysis was carried out with Bioconductor 1.3.28 libraries [49] in the R 1.8.0 environment on a dual processor PC with RedHat 8.0 operating system installed. Raw data were first normalized with quantile normalization; background correction and gene expression levels were then computed with Robust Multichip Average (RMA) [50].
Hierarchical clustering
Gene expression levels were expressed as RMA values and imported into Gene Cluster 3.0 [4]. Genes were centered to the mean and filtered with the criteria of genes with expression levels ≥ 1.0 in at least 10% of the samples. Since the RMA values are base-two logarithm-converted values, this statement is equivalent to a requirement that the expression levels of the selected genes be ≥ 2-fold or ≤ 1/2 of the mean expression level of that gene in at least 10% of the samples. In case of the spinal cord injury data set, 397 of 8,799 genes passed these filtering criteria. The selected genes and all the samples were clustered using centroid linkage hierarchical clustering based on uncentered correlation similarity metric. Branches of the clustering trees were then folded using SOM and visualized using Java TreeView 1.0.1. The result of this analysis with the spinal cord injury data set is presented in Figure 1.
Isomap analysis of the microarray data sets
Gene expression levels of all genes were expressed as RMA values and the result table was imported into the Matlab environment. Sample-to-sample differences between all pairs of genes were expressed as Euclidean distances of RMA values. Isomap was carried out in Matlab with the algorithm provided by Joshua Tenenbaum et al. [35]. We used the nearest neighbor method with k = 2, 3, or 4 dependent on which value of k generated the minimal residual variance. Other data processing and visualization steps with the Isomap models were carried out with custom written Matlab functions. Source code is presented in Additional Files 5, 6, 7, 8.
Isomap analysis of the monocytic and neutrophilic differentiation markers
Stegmaier et al. computed "monocyte and neutrophil scores" based on four parameters: interleukin 1 receptor antagonist (IL1RN) and secreted phosphoprotein 1 (SPP1) for the monocyte signature genes and autosomal chronic granulomatous disease protein (NCF1) and orosomucoid 1 (ORM1) for the neutrophil signature genes [40]. In a supplementary table of their paper, these four variables were presented for undifferentiated HL-60 cells as well as for cells treated with one of 16 chemical compounds. In this study, we expressed sample-to-sample differences as Euclidean distances of the base-two logarithm-converted mean values of the four marker genes. An Isomap model using the nearest neighbor method with k = 2 was performed with the algorithm provided by Joshua Tenenbaum [35]. All other data processing and visualization steps were carried out with custom written Matlab functions. The result of this analysis is presented in Figure 4E.
Authors' contributions
KD conceived of and designed the study, carried out the data analysis and visualization, developed the Matlab computer code, and drafted the manuscript. RLR and WM contributed to the study design and editing of the manuscript. All authors read and approved the final manuscript.
Supplementary Material
Additional File 5
Matlab code. Gene expression values are stored in a matrix (M) with variables (genes) in the rows and observations (samples) in the columns. A vector (samples) contains the sample names. Another vector (pheno) contains the classification of each sample. The gene names are stored in another vector (genes). Euclidean distances may be computed as D = L2_distance(M, M); If we want to perform Isomap at dimensions 1 through 20 then we store this range in: options.dims = 1:20; The Isomap algorithm can be executed with a neighborhood size of 5 as follows: [Y, R] = Isomap(D,'k',5, options); 1 through 20 dimensional coordinates are stored in Y. coords and the residual variances are stored in R. The results can be wrapped around to create a 3D 'model' structure, using makemodel: model = makemodel(M, Y, options, samples, pheno,'i',3,5);
Click here for file
Additional File 6
Matlab code. The function showmode can be used l to render an image from this model: showmodel(model); Enter showmodel without input parameters for more options.
Click here for file
Additional File 7
Matlab code. Adding the sample names to the figure: annotate(Y, samples,3);
Click here for file
Additional File 8
Matlab code. Similarly, a 3-D animation (mov) can be created using: mov = makemovie(model); Enter makemovie without input parameters for more options.
Click here for file
Additional File 1
Animated Isomap model of Fig. 2A.
Click here for file
Additional File 2
Animated Isomap model of Fig. 2B.
Click here for file
Additional File 3
Animated Isomap model of Fig. 2C.
Click here for file
Additional File 4
Animated Isomap model of Fig. 5A.
Click here for file
Acknowledgements
This work was funded by grant (P60MD00222) from the NIH to the NCMHD Center of Excellence in Nutritional Genomics.
==== Refs
Nicholson JK Holmes E Lindon JC Wilson ID The challenges of modeling mammalian biocomplexity Nat Biotechnol 2004 22 1268 1274 15470467 10.1038/nbt1015
Arrays Come of Age Genome Technology 2004 42 38 39
Golub TR Slonim DK Tamayo P Huard C Gaasenbeek M Mesirov JP Coller H Loh ML Downing JR Caligiuri MA Bloomfield CD Lander ES Molecular classification of cancer: class discovery and class prediction by gene expression monitoring Science 1999 286 531 537 10521349 10.1126/science.286.5439.531
Eisen MB Spellman PT Brown PO Botstein D Cluster analysis and display of genome-wide expression patterns Proc Natl Acad Sci U S A 1998 95 14863 14868 9843981 10.1073/pnas.95.25.14863
Spellman PT Sherlock G Zhang MQ Iyer VR Anders K Eisen MB Brown PO Botstein D Futcher B Comprehensive identification of cell cycle-regulated genes of the yeast Saccharomyces cerevisiae by microarray hybridization Mol Biol Cell 1998 9 3273 3297 9843569
Perou CM Sorlie T Eisen MB van de Rijn M Jeffrey SS Rees CA Pollack JR Ross DT Johnsen H Akslen LA Fluge O Pergamenschikov A Williams C Zhu SX Lonning PE Borresen-Dale AL Brown PO Botstein D Molecular portraits of human breast tumours Nature 2000 406 747 752 10963602 10.1038/35021093
Tavazoie S Hughes JD Campbell MJ Cho RJ Church GM Systematic determination of genetic network architecture Nat Genet 1999 22 281 285 10391217 10.1038/10343
Ben-Dor A Shamir R Yakhini Z Clustering gene expression patterns J Comput Biol 1999 6 281 297 10582567 10.1089/106652799318274
Wang J Bo TH Jonassen I Myklebost O Hovig E Tumor classification and marker gene prediction by feature selection and fuzzy c-means clustering using microarray data BMC Bioinformatics 2003 4 60 14651757 10.1186/1471-2105-4-60
Alon U Barkai N Notterman DA Gish K Ybarra S Mack D Levine AJ Broad patterns of gene expression revealed by clustering analysis of tumor and normal colon tissues probed by oligonucleotide arrays Proc Natl Acad Sci U S A 1999 96 6745 6750 10359783 10.1073/pnas.96.12.6745
Scherf U Ross DT Waltham M Smith LH Lee JK Tanabe L Kohn KW Reinhold WC Myers TG Andrews DT Scudiero DA Eisen MB Sausville EA Pommier Y Botstein D Brown PO Weinstein JN A gene expression database for the molecular pharmacology of cancer Nat Genet 2000 24 236 244 10700175 10.1038/73439
Tamayo P Slonim D Mesirov J Zhu Q Kitareewan S Dmitrovsky E Lander ES Golub TR Interpreting patterns of gene expression with self-organizing maps: methods and application to hematopoietic differentiation Proc Natl Acad Sci U S A 1999 96 2907 2912 10077610 10.1073/pnas.96.6.2907
Staunton JE Slonim DK Coller HA Tamayo P Angelo MJ Park J Scherf U Lee JK Reinhold WO Weinstein JN Mesirov JP Lander ES Golub TR Chemosensitivity prediction by transcriptional profiling Proc Natl Acad Sci U S A 2001 98 10787 10792 11553813 10.1073/pnas.191368598
Brown MP Grundy WN Lin D Cristianini N Sugnet CW Furey TS Ares MJ Haussler D Knowledge-based analysis of microarray gene expression data by using support vector machines Proc Natl Acad Sci U S A 2000 97 262 267 10618406 10.1073/pnas.97.1.262
Furey TS Cristianini N Duffy N Bednarski DW Schummer M Haussler D Support vector machine classification and validation of cancer tissue samples using microarray expression data Bioinformatics 2000 16 906 914 11120680 10.1093/bioinformatics/16.10.906
Ramaswamy S Tamayo P Rifkin R Mukherjee S Yeang CH Angelo M Ladd C Reich M Latulippe E Mesirov JP Poggio T Gerald W Loda M Lander ES Golub TR Multiclass cancer diagnosis using tumor gene expression signatures Proc Natl Acad Sci U S A 2001 98 15149 15154 11742071 10.1073/pnas.211566398
Su AI Welsh JB Sapinoso LM Kern SG Dimitrov P Lapp H Schultz PG Powell SM Moskaluk CA Frierson HFJ Hampton GM Molecular classification of human carcinomas by use of gene expression signatures Cancer Res 2001 61 7388 7393 11606367
Yeang CH Ramaswamy S Tamayo P Mukherjee S Rifkin RM Angelo M Reich M Lander E Mesirov J Golub T Molecular classification of multiple tumor types Bioinformatics 2001 17 Suppl 1 S316 22 11473023
Khan J Wei JS Ringner M Saal LH Ladanyi M Westermann F Berthold F Schwab M Antonescu CR Peterson C Meltzer PS Classification and diagnostic prediction of cancers using gene expression profiling and artificial neural networks Nat Med 2001 7 673 679 11385503 10.1038/89044
Hastie T Tibshirani R Botstein D Brown P Supervised harvesting of expression trees Genome Biol 2001 2 RESEARCH00031 12 10.1186/gb-2001-2-1-research0003
Tibshirani R Hastie T Narasimhan B Chu G Diagnosis of multiple cancer types by shrunken centroids of gene expression Proc Natl Acad Sci U S A 2002 99 6567 6572 12011421 10.1073/pnas.082099299
Khan J Simon R Bittner M Chen Y Leighton SB Pohida T Smith PD Jiang Y Gooden GC Trent JM Meltzer PS Gene expression profiling of alveolar rhabdomyosarcoma with cDNA microarrays Cancer Res 1998 58 5009 5013 9823299
Bittner M Meltzer P Chen Y Jiang Y Seftor E Hendrix M Radmacher M Simon R Yakhini Z Ben-Dor A Sampas N Dougherty E Wang E Marincola F Gooden C Lueders J Glatfelter A Pollock P Carpten J Gillanders E Leja D Dietrich K Beaudry C Berens M Alberts D Sondak V Molecular classification of cutaneous malignant melanoma by gene expression profiling Nature 2000 406 536 540 10952317 10.1038/35020115
Hedenfalk I Duggan D Chen Y Radmacher M Bittner M Simon R Meltzer P Gusterson B Esteller M Kallioniemi OP Wilfond B Borg A Trent J Gene-expression profiles in hereditary breast cancer N Engl J Med 2001 344 539 548 11207349 10.1056/NEJM200102223440801
Wen X Fuhrman S Michaels GS Carr DB Smith S Barker JL Somogyi R Large-scale temporal gene expression mapping of central nervous system development Proc Natl Acad Sci U S A 1998 95 334 339 9419376 10.1073/pnas.95.1.334
Pomeroy SL Tamayo P Gaasenbeek M Sturla LM Angelo M McLaughlin ME Kim JY Goumnerova LC Black PM Lau C Allen JC Zagzag D Olson JM Curran T Wetmore C Biegel JA Poggio T Mukherjee S Rifkin R Califano A Stolovitzky G Louis DN Mesirov JP Lander ES Golub TR Prediction of central nervous system embryonal tumour outcome based on gene expression Nature 2002 415 436 442 11807556 10.1038/415436a
Girolami M Breitling R Biologically valid linear factor models of gene expression Bioinformatics 2004 20 3021 3033 15201181 10.1093/bioinformatics/bth354
Nielsen TO West RB Linn SC Alter O Knowling MA O'Connell JX Zhu S Fero M Sherlock G Pollack JR Brown PO Botstein D van de Rijn M Molecular characterisation of soft tissue tumours: a gene expression study Lancet 2002 359 1301 1307 11965276 10.1016/S0140-6736(02)08270-3
Holter NS Mitra M Maritan A Cieplak M Banavar JR Fedoroff NV Fundamental patterns underlying gene expression profiles: simplicity from complexity Proc Natl Acad Sci U S A 2000 97 8409 8414 10890920 10.1073/pnas.150242097
Holter NS Maritan A Cieplak M Fedoroff NV Banavar JR Dynamic modeling of gene expression data Proc Natl Acad Sci U S A 2001 98 1693 1698 11172013 10.1073/pnas.98.4.1693
Alter O Brown PO Botstein D Generalized singular value decomposition for comparative analysis of genome-scale expression data sets of two different organisms Proc Natl Acad Sci U S A 2003 100 3351 3356 12631705 10.1073/pnas.0530258100
Liu L Hawkins DM Ghosh S Young SS Robust singular value decomposition analysis of microarray data Proc Natl Acad Sci U S A 2003 100 13167 13172 14581611 10.1073/pnas.1733249100
Sammon JWJ A nonlinear mapping for data structure analysis. IEEE Trans Comp 1969 C-18 401 409
Garcia de la Nava J Santaella DF Cuenca Alba J Maria Carazo J Trelles O Pascual-Montano A Engene: the processing and exploratory analysis of gene expression data Bioinformatics 2003 19 657 658 12651727 10.1093/bioinformatics/btg028
Tenenbaum JB de Silva V Langford JC A global geometric framework for nonlinear dimensionality reduction Science 2000 290 2319 2323 11125149 10.1126/science.290.5500.2319
Dawson K Rodriguez RL Malyj WV Gene Expression and Proteomics Data Analysis in Cancer Diagnosis and Prognosis Using Isomap, a Non-Linear Algorithm Breast Cancer Research 2003 5
Nilsson J Fioretos T Hoglund M Fontes M Approximate geodesic distances reveal biologically relevant structures in microarray data Bioinformatics 2004 20 874 880 14752004 10.1093/bioinformatics/btg496
Di Giovanni S Knoblach SM Brandoli C Aden SA Hoffman EP Faden AI Gene profiling in spinal cord injury shows role of cell cycle in neuronal death Ann Neurol 2003 53 454 468 12666113 10.1002/ana.10472
Walker JR Su AI Self DW Hogenesch JB Lapp H Maier R Hoyer D Bilbe G Applications of a rat multiple tissue gene expression data set Genome Res 2004 14 742 749 15060018 10.1101/gr.2161804
Stegmaier K Ross KN Colavito SA O'Malley S Stockwell BR Golub TR Gene expression-based high-throughput screening(GE-HTS) and application to leukemia differentiation Nat Genet 2004 36 257 263 14770183 10.1038/ng1305
Bittner M Meltzer P Trent J Data analysis and integration: of steps and arrows Nat Genet 1999 22 213 215 10391202 10.1038/10265
Beer MA Tavazoie S Predicting gene expression from sequence Cell 2004 117 185 198 15084257 10.1016/S0092-8674(04)00304-6
Stuart JM Segal E Koller D Kim SK A gene-coexpression network for global discovery of conserved genetic modules Science 2003 302 249 255 12934013 10.1126/science.1087447
Kim SK Lund J Kiraly M Duke K Jiang M Stuart JM Eizinger A Wylie BN Davidson GS A gene expression map for Caenorhabditis elegans Science 2001 293 2087 2092 11557892 10.1126/science.1061603
von Heydebreck A Huber W Poustka A Vingron M Identifying splits with clear separation: a new class discovery method for gene expression data Bioinformatics 2001 17 Suppl 1 S107 14 11472999
Alizadeh AA Eisen MB Davis RE Ma C Lossos IS Rosenwald A Boldrick JC Sabet H Tran T Yu X Powell JI Yang L Marti GE Moore T Hudson JJ Lu L Lewis DB Tibshirani R Sherlock G Chan WC Greiner TC Weisenburger DD Armitage JO Warnke R Levy R Wilson W Grever MR Byrd JC Botstein D Brown PO Staudt LM Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling Nature 2000 403 503 511 10676951 10.1038/35000501
Butte AJ Tamayo P Slonim D Golub TR Kohane IS Discovering functional relationships between RNA expression and chemotherapeutic susceptibility using relevance networks Proc Natl Acad Sci U S A 2000 97 12182 12186 11027309 10.1073/pnas.220392197
Somorjai RL Dolenko B Baumgartner R Class prediction and discovery using gene microarray and proteomics mass spectroscopy data: curses, caveats, cautions Bioinformatics 2003 19 1484 1491 12912828 10.1093/bioinformatics/btg182
Dudoit S Gentleman RC Quackenbush J Open source software for the analysis of microarray data Biotechniques 2003 Suppl 45 51 12664684
Irizarry RA Hobbs B Collin F Beazer-Barclay YD Antonellis KJ Scherf U Speed TP Exploration, normalization, and summaries of high density oligonucleotide array probe level data Biostatistics 2003 4 249 264 12925520 10.1093/biostatistics/4.2.249
|
16076401
|
PMC1189082
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Aug 2; 6:195
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-195
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1651598751310.1186/1471-2105-6-165Methodology ArticleIdentifying differential expression in multiple SAGE libraries: an overdispersed log-linear model approach Lu Jun [email protected] John K [email protected] Thomas B [email protected] Department of Biostatistics & Bioinformatics, Duke University, Durham, North Carolina 27708, USA2005 29 6 2005 6 165 165 4 3 2005 29 6 2005 Copyright © 2005 Lu et al; licensee BioMed Central Ltd.2005Lu et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
In testing for differential gene expression involving multiple serial analysis of gene expression (SAGE) libraries, it is critical to account for both between and within library variation. Several methods have been proposed, including the t test, tw test, and an overdispersed logistic regression approach. The merits of these tests, however, have not been fully evaluated. Questions still remain on whether further improvements can be made.
Results
In this article, we introduce an overdispersed log-linear model approach to analyzing SAGE; we evaluate and compare its performance with three other tests: the two-sample t test, tw test and another based on overdispersed logistic linear regression. Analysis of simulated and real datasets show that both the log-linear and logistic overdispersion methods generally perform better than the t and tw tests; the log-linear method is further found to have better performance than the logistic method, showing equal or higher statistical power over a range of parameter values and with different data distributions.
Conclusion
Overdispersed log-linear models provide an attractive and reliable framework for analyzing SAGE experiments involving multiple libraries. For convenience, the implementation of this method is available through a user-friendly web-interface available at .
==== Body
Background
Serial analysis of gene expression (SAGE) is used to measure relative abundances of messenger RNAs (mRNAs) for a large number of genes [1,2]. Briefly, mRNAs are extracted from biological samples and reverse-transcribed to cDNAs. The double-stranded cDNAs are then digested by a 4-cutter restriction enzyme (anchoring enzymes, usually NlaIII). After digestion, another restriction enzyme (tagging enzymes) is used to release the downstream DNA sequences at 3' of most of the anchoring enzyme restriction sites. The released sequences, usually 10–11 base pairs (bp) long, are called SAGE tags. The tags derived from many different species of mRNAs can be concatenated, cloned and sequenced. In a typical SAGE experiment, a large number of tags (often ranging from 30,000 to 100,000) are collected from each sample, with each tag representing, ideally, one gene; the tag count indicates the transcription level of the gene represented by that specific tag. A natural question of interest is whether a given tag is differentially expressed. Over the past few years, SAGE has been extensively used for expression analysis of cancer samples for identifying diagnostic or therapeutic targets [3,4].
Most SAGE studies focus on comparing expression levels between two samples. For such two-library comparisons, several statistical methods have been proposed, such as the simulation method of Zhang et al. [2], the Bayesian approaches [5-7], and the normal approximation based z-test [8] (which is equivalent to the chi-square test [9]). A comparative review by Ruijter et al. [10] has shown that all these methods perform equally well.
The comparison between two SAGE libraries can identify biologically interesting tags (or genes). However, in many cases it is essential to conduct experiments with replicates in order to account for normal background biological variation. For experiments involving multiple SAGE libraries, between-library variation beyond the binomial sampling variation is introduced. Such between-library variation can be due to additional known factors involved in the experimental design, as well as to unknown genetic or environmental variation between observations. Indeed, major differences in gene expression exist among SAGE libraries prepared from the same tissues of different individuals [11]. Statistical methods are needed for analyzing SAGE experiments involving multiple libraries. In the case of two-group comparisons (e.g. comparisons between a normal group and a cancer group), methods such as pooling the libraries in each group and transforming to two-library comparisons (for example, using the chi-square test), or the two-sample t-test on proportions have been proposed and discussed [12-14]. The pooling approach is often problematic since it ignores gene expression variation among libraries within the same treatment group, which leads to biased estimates for the variance. The two-sample t-test on proportions, however, can be problematic as well; proportions estimated from libraries with smaller sizes are known to be more variable than those from larger libraries.
For two-group comparisons, Baggerly et al. introduced a test statistic, tw, based on a hierarchical beta-binomial model to account for both between-library and within-library variation [13]. The tw test statistic is assumed to have an approximate t-distribution and like the t-test, the tw-test is only good for two-group comparisons. For SAGE experiments with a more general design (e.g. involving 2 or more factors), an approach based on overdispersed logistic regression has been described [15]. Overdispersed models aim to allow for the possibility of overdispersion in the tag counts, i.e., cases where the variance in tag counts exceeds what is expected for binomial or Poisson sampling alone. Besides its flexibility in modeling multiple factors and/or continuous covariates, logistic regression compares group proportions on a logit scale (log of odds) rather than a raw scale as in the t and tw tests. Comparing groups in logistic regression (and any generalized linear model) is equivalent to testing the hypotheses of whether the coefficients β = 0. Baggerly et al. [15] showed evidence suggesting that "the logit scale may be more appropriate" than the original proportion scale. One drawback with overdispersed logistic regression, however, is that it can break down for cases where all the tag counts in any of groups are very small. In such cases, the deviance test rather than the t-test (on the hypothesis that the coefficient β is zero) has been proposed [15]. Besides that a systematic evaluation of the deviance test is needed, a potential drawback with the deviance test is that it may require multiple rounds of model fitting if a model contains multiple factors or covariates. Furthermore, questions still remain on exactly when the deviance test should be used in preference to the t-test.
In this report we introduce an overdispersed log-linear model approach to analyzing SAGE which is closely related to overdispersed logistic regression but has a different mean-variance relationship assumption. We compare its performance in identifying differential expression with that of three other methods, including the t-test, tw test and overdispersed logistic regression. Analysis of simulated and real datasets show that both the log-linear and logistic overdispersion methods generally perform better than the t and tw tests. Based on simulated data, the log-linear method is found to have better performance than the logistic method, showing equal or higher statistical power over a range of parameter values and with different data distributions. The overdispersed log-linear method also appears to have better performance on the real SAGE data which we analyze; a number of cases are seen where a tag is identified by the log-linear approach and appears to be clearly differentially expressed, but which would not have been identified as significant using the logistic regression method. Overdispersed log-linear models also offer the same flexibility as logistic regression, allowing for modeling multiple factors and/or covariates. We conclude that the overdispersed log-linear models provide an attractive and reliable framework for analyzing SAGE experiments involving multiple libraries.
Results
Overdispersed log-linear models: a case study
Overdispersed log-linear models (see details in Methods) are very similar to overdispersed logistic models, but there are two major differences. First, overdispersed log-linear models work with logarithms of proportions (the log link) with logarithms of sample sizes mi as offsets. In contrast, overdisersped logistic models use the log of the odds (the logit link). Second, the assumption of an overdispersed log-linear model leads to derived weights used by iteratively reweighted least squares (IRLS) that depend on the means of the tag counts (i.e. the weights depend on both library sizes and tag proportions). The weights in overdispersed logistic regression, in contrast, are a function of library sizes only (see Methods).
Baggerly et al. [15] illustrated that the overdispersed logistic model can break down in cases where all proportions in one group are 0. Here we show that such a breakdown can also occur when the proportions in one group are small. Table 1 lists the p-values obtained from both the deviance and t tests. Note that we are testing the hypothesis that β = 0. Artificially increasing the tag counts in group 1 so that they approach the level seen in group 2 (which are held fixed), the deviance test in logistic regression and both tests (deviance and t) in the log-linear model show the expected trend of an increasing p-value (Table 1, columns 5, 6, and 7). In contrast, the p-values from the t-test in logistic regression actually decrease first and then increase (Table 1, column 4). This discrepancy between results from the t and deviance tests in the logistic model (a discrepancy not seen in the log-linear case) suggests that logistic regression can be problematic when the tag counts of all samples in one group are small.
Table 1 Comparisons of t- and deviance tests in overdispersed logistic regression and log-linear models and a test based on a Bayesian model
Group 1a logistic regression log-linear model Bayesian model
library 1 library 2 t-testc deviance test t-testc deviance test Ed
1b 0 0 0.645 0.115 0.003 0.001 0.01
2 2 2 0.485 0.122 0.002 0.002 0.02
3 5 5 0.383 0.133 0.003 0.005 0.04
4 10 10 0.324 0.149 0.007 0.01 0.05
5 20 20 0.291 0.183 0.02 0.025 0.07
6 50 50 0.324 0.29 0.104 0.117 0.11
7 100 100 0.494 0.508 0.376 0.404 0.12
aTag counts in group 1 are artificially increased towards the levels observed in group 2 (which are held fixed). Tag counts in group 2 are 312, 549, 246, 65, 41, and 52. The library sizes and tag counts in group 2 are taken from Baggerly et al. [15].
b The empirical tag counts 0.506, and 0.494 are used to replace the zero counts in group 1[15].
c The t-test here is testing the hypothesis that β = 0.
d E, the Bayes Error Rate, is listed. [26].
Simulation study
To systematically evaluate the performance of the various tests in the case of two-group comparisons, we performed a simulation study. The tests compared here are the t, tw, logit-t and log-t. For t and tw, the test is whether pA = pB, where pA and pB are the mean proportion in groups A and B respectively. The logit-t and log-t are t tests on the hypothesis of whether β = 0 in the overdispersed logistic regression and log-linear models respectively. We do not attempt to replace the t-test with the deviance test in the overdispersed logistic regression model since this requires making a possibly subjective decision on when to use one test in preference to the other.
We generated tag counts under three different distributions, choosing different tag proportions and amounts of overdispersion (Table 2). Data generated from the beta-binomial and negative binomial distributions meet the assumptions (i.e. have the mean-variance relationship structure) of the overdispersed logistic regression and log-linear models approaches, respectively. The negative binomial distribution is equivalent to the gamma-Poisson hierarchical model and is considered a robust alternative to the Poisson distribution [16,17]. It should be noted that the tw-test is also derived under the assumption that the data is generated from a beta-binomial distribution [13]. The range of overdispersion parameter values was chosen based on model fits from a real dataset (see section below); we used the 25, 50, and 75 percentile values of the estimated overdispersion from these fits. Note that the overdispersion parameter φ in the logistic model is not directly related to the φ in the log-linear model; φ values from the two models should not be compared. Given an overdispersion value φ and a group mean proportion p, the α and β values for the beta-binomial distribution are derived as α = p(1/φ - 1), and β = (1 - p)(1/φ - 1). The size parameter in the negative binomial distribution is easily derived as 1/φ. We used 5 samples (libraries) for each group, and determined the sizes of each of 10 libraries by randomly sampling from a uniform distribution over the interval between 30,000 and 90,000. This yielded library sizes of 66148, 67094, 53338, 80124, 64984, 70452, 74052, 60086, 52966 and 45377; these values were not changed over the course of the simulations. Results (not shown) from a separate run using a different set of library sizes were found to be in agreement with those shown here. A total of 5,000 sets of tag counts were generated for each combination of parameter values. The sensitivity and specificity of each of the tests were then evaluated and compared through receiver operating characteristic (ROC) curves [18].
Table 2 A list of parameter values used in the simulations
Distribution binomial (i.e. no overdispersion); beta-binomial; negative-binomial
overdispersion parameter (φ) 8e-06, 2e-05, 4.3e-05 for beta-binomial; 0.17, 0.42, 0.95 for negative binomial
number of samples in groups A and B 5 in each group
mean proportion in group A (pA) 1, 5, 10, 20, 50, and 100 out of 50,000
ratio of mean proportions (pB / pA) 1, 2 and 4
Note: the library sizes are 66148, 67094, 53338, 80124, 64984, 70452, 74052, 60086, 52966 and 45377, each of which was determined by a draw from a uniform distribution over the interval from 30,000 to 90,000.
The ROC curves (one for each of the four tests) shown in Figure 1 were obtained using data generated from the beta-binomial distribution (with overdispersion values φ shown on the top of the figure). Given the same false positive rate (x-axis), the overdispersion models (logistic and log-linear) clearly show improved statistical power (y-axis) compared to the two-sample t and tw tests. In contrast, when the four tests are applied to data generated from the negative binomial distribution, the overdispersed log-linear model clearly outperforms the other three tests (Figure 2). Again, the two-sample t and tw tests do not perform well in general. The figures generated using other parameter values are available [see Additional files 1 and 2]. These results suggest that for SAGE data, statistics methods based on raw proportions (as in the t and tw tests) show less power than the logistic or log-linear model approaches. The overdispersed log-linear model not only shows the best performance in cases where the data are generated in a manner consistent with its assumptions (i.e. from the negative binomial distribution), but also has competitive performance when the data come from a different distribution (here the beta-binomial). This suggests that the overdispersed log-linear model approach is more robust.
Figure 1 Comparisons based on simulated data from the beta-binomial distribution. This figure shows the receiver operating characteristic curves (ROC) of the four tests applied to datasets generated from the beta-binomial distribution with various magnitudes of overdispersion (φ) (shown on the top of each graph). For a specific φ, 10,000 observations (tags) are simulated; 5,000 are generated under the assumption that pA = pB and the remaining from pB = 2 pA, where pA and pB are the mean proportions of the two groups and pA = 0.0002 (i.e. 10 out of 50,000). For figures generated under other conditions, see Additional file 1.
Figure 2 Comparisons based on simulated data from the negative binomial distribution. The ROC curves of the four tests are based on datasets generated from the negative binomial distribution with various magnitudes of overdispersion (φ). The data are simulated by the same strategy as used in Figure 1, except that pB = 4pA. Note that the overdispersion parameter here is not directly comparable with that in Figure 1 (the parameter φ for the negative binomial is not directly related to that for the beta-binomial). For figures generated under other conditions, see Additional file 2.
A pancreatic cancer dataset
We further compared the four tests (t-test, tw-test, logit-t, and log-t) using an experimental SAGE data set obtained from the publicly available SAGE Genie website [19]. To identify genes differentially expressed between the pancreatic cancer cells and normal ductal epithelium, Ryu et al. [12] compared the gene expression levels of five pancreatic cancer cell lines and two samples of normal pancreatic ductal epithelial cells. The library sizes and numbers of unique tags for the SAGE libraries are shown in Table 3. Note that the numbers in the table are slightly different from those described in the original paper due to the different SAGE tag processing procedures [20]. In this analysis, we ignore tags with total counts less than 3.
Table 3 Library information on 5 cancer and 2 normal pancreas SAGE libraries
Cancer cell lines Normal cells
Library ASPC PL45 CAPAN1 CAPAN2 Panc-1 HX H126
Library size 31,224 29,557 37,674 23,042 24,749 31,985 32,223
Unique tags 10,622 11,121 14,815 10,157 10,293 12,392 12,360
We first compare the four tests by examining the overlap between the top ranking genes (top 50 and 100) identified by each test (Table 4). For the t and tw tests, the genes are ranked by the absolute value of the t (or tw) statistics instead of by p-values (see Discussion section for details). As shown in Table 4, the results from the logit-t and log-t tests show the highest agreement (~80%); moderate agreement is observed between tw and logit-t or log-t tests (~60%) and the least agreement is seen between the t and the other three tests (~40%). The top ranking genes identified by the t-test are often those with extremely small within-group variances (data not shown). Overall, results from the t-test differ the most from the results of the other tests, while the most similar results are seen between the logit-t and log-t tests. This generally agrees with the trend seen in the simulations.
Table 4 Pair-wise comparisons of the four tests
t-test tw-test logit-t
tw-test 39(12)a -
logit-t 42(17) 66(29) -
log-t 36(16) 63(25) 82(43)
a number of genes shared among the list of top 100 and top 50 (in parenthesis) genes identified by the two tests; we note that for the t and tw tests, the genes were ranked by the absolute t or tw statistic rather than by p-values.
Of the top 100 genes (ranked by p-value) obtained from the logit-t and log-t tests, 82 genes are in common leaving 18 genes from each test that are not within the top 100 identified by the other test. To further examine the discrepancy between the logit-t and log-t tests, we plotted p-values obtained from both tests for these 36 remaining tags (Fig 3). It can be seen that, while tags identified by the logit-t test are also given relatively small p-values by the log-t test (all less than 0.05), those identified by the log-t test show a wide range of p-values according to the logit-t test. Table 5 lists tags which are ranked among the top 100 by the log-t test but which have p-values greater than 0.05 by the logit-t test; 4 of these were also identified by Ryu et al. [12]. Our analysis indicates that the log-t test is relatively robust in that it not only gives reasonably small p-values to genes identified as significant by the logit-t test, but also identifies genes which would never have been considered significant by the logit-t test.
Figure 3 Comparing p-values from the logit-t test and those from the log-t test. Of the top 100 tags (ranked according to p-values) identified by the logit-t test and by the log-t test, 82 are common to both leaving 18 tags from each test that are not within the top 100 identified by the other. The p-values from both tests for these 36 remaining tags are plotted here. The circles represent the 18 in the top 100 by the logit-t test and the triangles those from the log-t test. While all the tags identified by the logit-t test also have reasonably low p-values according to the log-t test, the tags identified by the log-t test show a much wider range of p-values according to the logit-t test.
Table 5 A set of genes identified as significantly differentially expressed (p < 0.05 and also among the list of top 100 genes) according to the log-t test but not by the logit-t test (p > 0.05)
Normal Cancer
Tag P (log-t) P (logit-t) HX H126 ASPC PL45 CAPAN1 CAPAN2 Panc-1
AGCAGATCAG* 0.003 0.088 16 9 272 152 138 135 384
TTGGTGAAGG 0.003 0.069 6 0 90 267 194 187 238
CCCATCGTCC 0.003 0.309 13 34 2047 1333 364 456 408
CCTCCAGCTA 0.006 0.465 3 16 452 1766 292 265 364
ACTTTTTCAA 0.008 0.096 25 43 413 379 226 200 65
CAAACCATCC* 0.01 0.463 9 9 439 1235 154 143 133
TGCCCTCAGG 0.011 0.219 16 6 80 196 276 339 4
GCTGTTGCGC* 0.011 0.151 3 3 35 30 82 126 133
GACATCAAGT* 0.013 0.554 0 0 183 548 85 126 20
TTCACTGTGA 0.014 0.149 0 3 128 105 77 91 16
TTGGGGTTTC 0.015 0.142 69 37 701 507 173 195 230
TGCCCTCAAA 0.016 0.246 3 6 32 112 135 178 0
GGGGAAATCG 0.017 0.066 100 71 339 423 119 291 226
Note: Tag counts have been converted to number of tags per 100,000 for the comparison purpose. This scaling is not used in any statistical tests. Tags with (*) are those also identified by Ryu et al. [12].
Ryu et al. [12] identified 49 up- and 37 down-regulated genes in cancer with the two-sample t-test and a set of rule-based methods. We compared their results with those from the log-t test (choosing the same number of top genes). Of the total of 86 genes, only 18 are in common (with 9 in each down- and up-regulated gene group). The most significant gene that is up-regulated in cancer on our list (but not in the original paper) is tag, "CTTCCAGCTA", which represents the annexin A2 gene. This gene has been reported to be up-regulated in human pancreatic carcinoma cells and primary pancreatic cancers [21]. Another example is tag 'TTGGTGAAGG', which corresponds to the gene encoding thymosin, beta 4. This gene also has been shown to be "expressed at high levels both in tumor cell lines and in primary cultures of normal pancreas, but not in normal tissue" [22]. A list of the top 20 genes up-regulated and the top 20 genes down-regulated in cancer based on the log-t test are listed in Table 6.
Table 6 A list of top 40 genes differentially expressed between pancreatic cancer and normal ductal epithelium
Tag Description P HX H126 ASPC PL45 CAPAN1 CAPAN2 Panc-1
Up-regulated in pancreatic cancer
CTTCCAGCTA annexin A2 0.0011 19 25 128 217 143 148 170
AAAAAAAAAA - 0.0018 6 3 128 210 180 165 133
AGCAGATCAG S100 calcium binding protein A10 (annexin II ligand, calpactin I, light polypeptide (p11)) 0.0027 16 9 272 152 138 135 384
TTGGTGAAGG thymosin, beta 4, X-linked 0.003 6 0 90 267 194 187 238
CCCATCGTCC motichondria gene 0.0032 13 34 2047 1333 364 456 408
CCTCCAGCTA keratin 8 0.0059 3 16 452 1766 292 265 364
GGAAAAAAAA ATP synthase, H+ transporting, mitochondrial F1 complex, epsilon subunit 0.0063 3 6 64 61 74 74 57
CCCCAGTTGC calpain, small subunit 1 0.0066 22 22 64 88 77 61 113
AACTAAAAAA ribosomal protein S27a 0.0078 19 16 45 85 80 61 61
TTCAATAAAA RPLP1, Ribosomal protein, large, P1 0.0079 9 25 147 179 135 104 40
GCAAAAAAAA chromosome 21 open reading frame 97 0.0079 6 3 58 68 40 65 65
ACTTTTTCAA motichondria gene 0.0081 25 43 413 379 226 200 65
CAAACCATCC KRT18, Keratin 18 0.0095 9 9 439 1235 154 143 133
GTGTGGGGGG Junction plakoglobin 0.0096 6 3 29 64 50 56 61
TGCCCTCAGG LCN2, Lipocalin 2 (oncogene 24p3) 0.0106 16 6 80 196 276 339 4
GCTGTTGCGC - 0.0108 3 3 35 30 82 126 133
AAGAAGATAG ribosomal protein L23a 0.0116 16 9 77 108 85 65 24
GAAAAAAAAA SMAD, mothers against DPP homolog 3 (Drosophila) 0.0118 6 0 74 47 40 56 44
ACCTGTATCC IFITM3, interferon induced transmembrane protein 3 (1-8U) 0.0123 13 3 26 81 64 82 53
CAACTTAGTT myosin regulatory light chain MRLC2 0.0128 6 6 51 61 53 48 16
Down-regulated in pancreatic cancer
GACGACACGA ribosomal protein S28 0.0001 428 388 109 122 90 117 154
GGACCACTGA ribosomal protein L3 0.0002 310 270 102 105 101 104 61
GATCTCTTGG S100 calcium binding protein A2 0.0002 188 174 3 10 8 4 0
AGCAGGAGCA S100 calcium binding protein A16 0.0005 144 152 26 41 45 26 16
AGCTGTCCCC capping protein (actin filament) muscle Z-line, beta 0.0005 219 254 13 3 3 4 0
GACTGCGCGT tumor necrosis factor receptor superfamily, member 12A 0.0007 103 93 10 10 24 22 16
GTGGTGTGTG congenital dyserythropoietic anemia, type I 0.0011 59 87 10 10 8 13 8
TAGGCATTCA - 0.0012 119 115 0 0 0 0 0
TGAGTGGTCA microtubule-associated protein 1 light chain 3 beta 0.0017 66 53 0 7 5 13 8
GGCGGCTGCA excision repair cross- complementing rodent repair deficiency, group 1 0.0017 66 53 6 7 3 4 0
AAGTTTGCCT glutaredoxin (thioltransferase) 0.0022 66 62 0 3 3 0 4
AGCTCTCCCT Ribosomal protein L17 0.0023 335 357 77 145 82 143 125
CCGAAGTCGA transcriptional regulating factor 1 0.0024 53 56 0 7 5 0 0
GCTGCTGCGC - 0.0024 228 320 0 0 0 0 4
TTGGGAGCAG isoleucine-tRNA synthetase 0.0031 72 43 10 10 19 4 8
TAAGGAGCTG Ribosomal protein S26 0.0031 344 329 138 85 96 43 101
AACAGAAGCA hypothetical protein FLJ25692 0.0031 75 59 13 24 24 9 16
CCTCCACCTA peroxiredoxin 2 0.0031 56 43 16 10 3 9 4
TGTGAGTCAC - 0.0038 31 62 0 0 0 0 0
TCAGGGATCT - 0.0038 41 53 0 0 0 0 0
Note: tag counts have been converted to tags per 100,000 for comparison purposes. The p values listed are from the log-t test.
Discussion
In this report we introduced a log-linear model with overdispersion for testing differential gene expression in SAGE. This model is closely related to the overdispersed logistic model proposed by Baggerly et al. [15] but with a different mean-variance relationship assumption. The differences between two models can be seen clearly in the weight (used by IRLS) associated with each observation: assuming library sizes are reasonably close, the overdispersed log-linear model tends to assign higher weights to observations in the group with the smaller mean proportion; in contrast, approximately equal weights are assigned to all the observations in the overdispersed logistic model. Although for real SAGE data the true mean-variance relationship is unknown, it has been observed that "for the higher counts data, the between-library variability is the dominant part of the variation" [13]; this suggests that the magnitude of the overdispersion in the group with higher counts is probably larger than that in the group with low counts so that the assumptions of the overdispersed log-linear model is probably more appropriate for SAGE data.
We also compared the model fits of the overdispersed logistic and log-linear models. Due to the introduction of the overdispersion parameter, the deviance statistic is no longer a valid basis for model fit comparison. An alternative is to use the standardized Pearson residuals, which have an asymptotic standard normal distribution [23]. Williams [24] proposed the approach of plotting the standardized Pearson residuals against the predicted proportions; a problem with a model fit is indicated by a significant decrease in the variance of the standardized residuals as estimated proportions approach zero. Figure 4 shows the residual plots from the logistic and log-linear model fits for two tags (the tag counts are listed in Table 5). In the overdipersed logistic regression case (left panels of Figure 4), the variance of the standardized Pearson's residuals is seen to be much smaller in the normal group than in the cancer group. Such a difference is not evident in the overdispersed log-linear model fits (right panels of Figure 4). Although the sample size is very small in this example (only 2 in the normal group), the residual plots give further indication that log-linear models provide a better fit to SAGE data than logistic models.
Figure 4 Plot of standardized residuals against estimated proportions. Standardized Pearson's residuals (y-axis) plotted vs. the proportion estimates (x-axis) for the two groups. The standardized Pearson's residuals are asymptotically distributed as a standard normal. The model fits of two tags (among the list of genes in Table 5) are shown here; the left is from the fit using the overdispersed logistic model and the right from the overdispersed log-linear model. A lower variance of residuals in the group (normal) with lower mean proportion is an indication of poor model fit.
From the simulation study we have shown that, besides their limitation to two-group comparison, both the t- and tw-tests, in general, are not as powerful as tests which allow for the possibility of overdispersion. We mention one specific problem that can arise with the t- and tw-tests if the number of samples in the data set is small. Note that the rank orders from the t-test and the tw test in Table 4 are based on test statistics instead of p-values. The rank orders based on p-values can be different from those based on test statistics if the residual degrees of freedom differ among tests. Both the t-test and the tw-test use the Satterthwaite approximation [25] for the number of degrees of freedom since the variances are assumed to be different in the two groups. An example of how this can be problematic is given by tag "AGCTGTCCCC", which has tag counts 70, 82 in the two normal samples, and 4, 1, 1, 1, 0 in the five cancer cell line samples. The differential expression is highly significant based on the logit-t (p-value 0.0003) and log-t (p-value 0.0005) tests. In contrast, if the tw-test with the Satterthwaite approximation to the degrees of freedom is used, this tag is barely significant at the 5% level (p-value 0.050). The reason is that, while the magnitude of the tw statistic for this tag is actually extremely high (|tw| = 12.01), the calculated degrees of freedom is only about 1 (which leads to low significance). The small value for the degrees of freedom arises here because the estimated variance in the cancer group is very small; the approximated degrees of freedom is then about equal to the sample size of the normal group minus 1 (here, 2-1 = 1). Cases like this occur frequently in this data set since the number of libraries (samples) in one group is very small. It is not uncommon to have small sample numbers with SAGE data.
The four methods compared in this study follow the frequentist approach of hypothesis testing, and can be broadly considered as examples of linear models. For two-group comparisons, Vencio et al. [26] introduced a Bayesian approach to rank tags by the Bayes Error Rate. We compared their approach with the methods based on linear models by looking at differences in gene rankings determined using the pancreatic dataset. Considering the top 100 genes identified by the different tests, the two overdispersed models show the best agreement with the Bayesian method (~70% in common); 63 genes (of the top 100) are identified by all three tests. We also evaluated the Bayesian method using the artificial data in Table 1; as the tag counts in group 1 are increased, the evidence for differential expression decreases (i.e. the Bayes Error Rate goes up), which follows the expected trend. Furthermore, if we recognize tags with p < 0.05 or E<0.1 as being significantly differentially expressed [26], the results from the Bayesian approach are more consistent with those from the log-linear model than from the logistic models (see Table 1). Since the evidence measures used are conceptually very different, to perform a direct comparison between "P-value"-based methods and the Bayesian approach is not straightforward. Our results, however, suggest that the Bayesian approach of Vencio is a competitive Bayesian alternative for analyzing SAGE data in the case of two-group comparisons.
The current study has not considered the issue of multiple testing problems which is still under active research [27,28]. We note that one possible area for further improvement is to use information across genes (tags) with similar magnitude of dispersion to obtain potentially more robust and accurate overdispersion (and therefore, error) estimates. In all the methods compared here, everything is done one tag at a time, i.e., estimates of the amount of overdispersion are done for each tag individually and these can vary widely (see Figure 5). For expression data with continuous values, strategies on information sharing have been proposed [29-31] and these strategies may be adapted for discrete data such as in SAGE.
Figure 5 The distribution of overdispersion estimates (). The estimates are from the overdispersed log-linear model fit to the pancreas data. Tags with the overdispersion estimate 0 are not shown in the figure.
Methods
Data
Suppose that there are a total of n SAGE libraries in an experiment. Let mi be the size (total tag counts) of library i (i = 1..n) and ri be the tag counts for a specific tag in that library.
Also, let xi be the associated vector of explanatory variables and β the vector of coefficients. The comparison of two groups of SAGE libraries is a special case where there is only one explanatory variable associated with each observation (i.e. one factor with 2 levels).
The two-sample t-test
The t-test proposed by Welch [25] was used to test whether the mean of the proportions in one group equals the mean of the other. The proportions are assumed to have unequal variances in the two groups and the degrees of freedom is calculated based on the Satterthwaite approximation as in the tw-test (see below).
The tw-test
Baggerly et al. [13] introduced a beta-binomial sampling model to account for the extra-binomial variation for a simple design in which the comparison is between two groups of SAGE libraries. This is a special case of a linear model that contains one explanatory variable. Briefly, unobserved random variables Pi are introduced to account for the between-library variation. For a given group, Pi is assumed to have a beta distribution (α, β) with mean and variance E(Pi) = α/(α+β), and Var(Pi) = αβ / [(α+β)2 (α+β+1)]. Notice that this is a special case of the form Var(Pi) = φ pi(1 - pi) as in the overdispersed logistic model, where φ = 1/(α+β+1). Next, the group proportion is estimated by a weighted linear combination of individual proportions within the group , where = ri/mi and wi are weights associated with each individual proportion. The unbiased variance estimator of is given as
To avoid having an estimated variance that is less than the binomial sampling variance, a lower-bound for the variance is also provided. All the parameters (i.e. α, β and wi) are obtained through an iterative procedure. The same estimation procedure is applied to data from the other group. For testing whether the proportion in one group (say group A) equals the proportion in the other group (group B), a t-like statistic tw is constructed, where
The tw statistic is assumed to have a t-distribution with the degrees of freedom (df) calculated from the Satterthwaite approximation:
where nA and nB are the number of SAGE libraries in the group A and B respectively. This test is called the tw-test here. The implementation of both the t- and tw-test can be found in [13].
Overdispersed logistic regression approach
Baggerly et al. [15] provided a thorough description on this approach and details can be found in [24]. Briefly, unobserved continuous random variables Pi are introduced to account for the between-library variation, where the mean and variance of Pi have the following forms: E(Pi) = pi ; Var(Pi) = φ pi(1 - pi). Here φ is a nonnegative scale parameter. Conditional on Pi= pi, the ri have a binomial distribution (mi, pi). The unconditional mean and variance of ri can be shown to be E(ri) = mi pi and Var(ri) = mi pi(1 - pi) [1+(mi-1) φ]. Notice that if φ is 0 (i.e. there is no between-library variation or overdispersion), the variance of ri is the usual binomial variance mi pi(1 - pi). The estimation of coefficients β is carried out by the iteratively reweighted least-squares (IRLS) procedure, where the weights wi are 1/ [1+(mi - 1) φ]. Note that the weights wi are equal if the library sizes mi are equal.
The parameter φ is estimated by equating the goodness of fit Pearson's chi-square statistic X2 to its approximate expected value, which is
where vi = mi pi(1 - pi), and di is the variance of the linear predictor . An iterative procedure is introduced to estimate φ and β, where the estimates of φ (and accordingly, the weights wi) and β are updated at each step. Given the estimated coefficients, the testing hypothesis is whether one (or more if there are more than two groups) of the coefficients (β) is 0. For this, the t-test rather than the z-test is recommended due to the introduction of the overdispersion parameter into the model [15,32].
The hypothesis test based on overdispersed logistic regression is called the logit-t test here. The implementation including source code can be found in [15]. We consider overdispersion models (logistic or log-linear) only if the Pearson's chi-square statistic from the usual logistic regression (or log-linear) fit (i.e. without overdispersion) is greater than or equal to its expected value, n-p.
Overdispersed log-linear models
This model is closely related to the overdispersed logistic regression model. One way to derive it is based on the gamma-Poisson hierarchical model assumption [16]. Assume that an unobserved random variables θi is distributed according to
θi ~ Gamma(μi, 1/φ),
where μi = mi pi, φ >0, E(θi) = μi and Var(θi) = . Given pi, the response variable ri is assumed to be distributed as
ri | pi ~ Poisson(μi).
The unconditional mean and variance of ri can be shown to be E(ri) = μi = mi pi and Var(ri) = μi (1+μiφ). Notice that as φ decreases to 0, the variance of ri approaches the usual Poisson variance μi (i.e. mi pi). The same mean-variance relationship can also be derived if we assume ri has a negative- binomial distribution [16]. The mean μi of the response variable ri and the covariates xi are connected through the log link function,
log μi = log(mi pi) = xiβ.
As in the overdispersed logistic regression model, the estimates of the coefficients β are obtained by the iteratively reweighted least-squares procedure, where the weights wi are 1/(1+μi φ) [33]. Note that, in contrast to the overdispersed logistic regression model where the weights only depend on library sizes mi, the weights in the log-linear model depend on μi (i.e. both mi and pi).
The hypothesis test based on overdispersed log-linear models is called the log-t test here. The R [34] source code and a web-interface for implementing this approach are available [35].
Authors' contributions
JL developed the method. JL and JKT carried out the simulation and data analysis. JKT and JL set up the web interface for implementing this approach. TBK supervised the study, and assisted with the methodology. All authors contributed to the writing, read and approved the final manuscript.
Supplementary Material
Additional File 1
This gzipped tar file contains figures showing the receiver operating characteristic curves (ROC) for the four tests applied to datasets generated from the beta-binomial distribution with various magnitudes of overdispersion(φ) and mean proportions. For example, the file 2_8e-06_0.0002.png shows the ROC curves when pB = 2pA, φ = 8e-06 and pA = 0.0002.
Click here for file
Additional File 2
Similar to the file above, this file contains figures of ROC curves but with data generated from the negative binomial distribution.
Click here for file
Acknowledgements
The authors would like to thank anonymous reviewers for several constructive comments. We thank Gregory Riggins for introducing us to SAGE. We gratefully acknowledge the financial support of the NIH through the Duke University Center for Translational Research (5 P30 AI051445-03) and through the Southeast Regional Center of Excellence in Biodefense and Emerging Infections (U54 AI057157-02); and of the NSF through a grant to our collaborator David Bird (NCSU; DBI 0077503) as well as the Duke Center for Bioinformatics and Computational Biology for support through a postdoctoral fellowship to JL.
==== Refs
Velculescu VE Zhang L Vogelstein B Kinzler KW Serial analysis of gene expression.[comment] Science 1995 270 484 487 7570003
Zhang L Zhou W Velculescu VE Kern SE Hruban RH Hamilton SR Vogelstein B Kinzler KW Gene expression profiles in normal and cancer cells Science 1997 276 1268 1272 9157888 10.1126/science.276.5316.1268
Riggins GJ Strausberg RL Genome and genetic resources from the Cancer Genome Anatomy Project Human Molecular Genetics 2001 10 663 667 11257097 10.1093/hmg/10.7.663
Porter D Lahti-Domenici J Keshaviah A Bae YK Argani P Marks J Richardson A Cooper A Strausberg R Riggins GJ Schnitt S Gabrielson E Gelman R Polyak K Molecular markers in ductal carcinoma in situ of the breast Molecular Cancer Research: MCR 2003 1 362 375 12651909
Audic S Claverie JM The significance of digital gene expression profiles Genome Research 1997 7 986 995 9331369
Chen H Centola M Altschul SF Metzger H Characterization of gene expression in resting and activated mast cells Journal of Experimental Medicine 1998 188 1657 1668 [erratum appears in J Exp Med 1998 Dec 21;188(12):2387]. 9802978 10.1084/jem.188.9.1657
Lal A Lash AE Altschul SF Velculescu V Zhang L McLendon RE Marra MA Prange C Morin PJ Polyak K Papadopoulos N Vogelstein B Kinzler KW Strausberg RL Riggins GJ A public database for gene expression in human cancers Cancer Research 1999 59 5403 5407 10554005
Kal AJ van Zonneveld AJ Benes V van den Berg M Koerkamp MG Albermann K Strack N Ruijter JM Richter A Dujon B Ansorge W Tabak HF Dynamics of gene expression revealed by comparison of serial analysis of gene expression transcript profiles from yeast grown on two different carbon sources Molecular Biology of the Cell 1999 10 1859 1872 10359602
Man MZ Wang X Wang Y POWER_SAGE: comparing statistical tests for SAGE experiments Bioinformatics 2000 16 953 959 11159306 10.1093/bioinformatics/16.11.953
Ruijter JM Van Kampen AH Baas F Statistical evaluation of SAGE libraries: consequences for experimental design Physiological Genomics 2002 11 37 44 12407185
Blackshaw S Kuo WP Park PJ Tsujikawa M Gunnersen JM Scott HS Boon WM Tan SS Cepko CL MicroSAGE is highly representative and reproducible but reveals major differences in gene expression among samples obtained from similar tissues Genome Biology 2003 4 R17 12620102 10.1186/gb-2003-4-3-r17
Ryu B Jones J Blades NJ Parmigiani G Hollingsworth MA Hruban RH Kern SE Relationships and differentially expressed genes among pancreatic cancers examined by large-scale serial analysis of gene expression Cancer Research 2002 62 819 826 11830538
Baggerly KA Deng L Morris JS Aldaz CM Differential expression in SAGE: accounting for normal between-library variation Bioinformatics 2003 19 1477 1483 12912827 10.1093/bioinformatics/btg173
Walter-Yohrling J Cao X Callahan M Weber W Morgenbesser S Madden SL Wang C Teicher BA Identification of genes expressed in malignant cells that promote invasion Cancer Res 2003 63 8939 8947 14695211
Baggerly KA Deng L Morris JS Aldaz CM Overdispersed logistic regression for SAGE: modelling multiple groups and covariates BMC Bioinformatics 2004 5 144 15469612 10.1186/1471-2105-5-144
Casella G Berger RL Statistical Inferences 2002 2 Pacific Grove, CA: DuXBURY
Gelman A Carlin J Stern H Rubin D Bayesian Data Analysis 2004 Second Boca Raton, Florida: CHAPMAN & HALL/CRC
Shapiro DE The interpretation of diagnostic tests Stat Methods Med Res 1999 8 113 134 10501649 10.1191/096228099666928387
SAGE Genie:
Boon K Osorio EC Greenhut SF Schaefer CF Shoemaker J Polyak K Morin PJ Buetow KH Strausberg RL De Souza SJ Riggins GJ An anatomy of normal and malignant gene expression.[comment] Proceedings of the National Academy of Sciences of the United States of America 2002 99 11287 11292 12119410 10.1073/pnas.152324199
Vishwanatha JK Chiang Y Kumble KD Hollingsworth MA Pour PM Enhanced expression of annexin II in human pancreatic carcinoma cells and primary pancreatic cancers Carcinogenesis 1993 14 2575 2579 8269629
Paciucci R Berrozpe G Tora M Navarro E Garcia de Herreros A Real FX Isolation of tissue-type plasminogen activator, cathepsin H, and non-specific cross-reacting antigen from SK-PC-1 pancreas cancer cells using subtractive hybridization FEBS Lett 1996 385 72 76 8641471 10.1016/0014-5793(96)00352-3
Agresti A Categorical Data Analysis 2002 2 Hoboken, New Jersey: A John Wiley & Sons, Inc., Publication
Williams DA Extra-binomial variation in logistic linear models Applied Statistics 1982 31 144 148
Welch BL The generalization of 'students' problem when several different population variances are involved Biometrika 1947 34 28 35
Vencio RZ Brentani H Patrao DF Pereira CA Bayesian model accounting for within-class biological variability in Serial Analysis of Gene Expression (SAGE) BMC Bioinformatics 2004 5 119 15339345 10.1186/1471-2105-5-119
Manly KF Nettleton D Hwang JT Genomics, prior probability, and statistical tests of multiple hypotheses Genome Res 2004 14 997 1001 15173107 10.1101/gr.2156804
Choe SE Boutros M Michelson AM Church GM Halfon MS Preferred analysis methods for Affymetrix GeneChips revealed by a wholly defined control dataset Genome Biol 2005 6 R16 15693945 10.1186/gb-2005-6-2-r16
Kepler TB Crosby L Morgan KT Normalization and analysis of DNA microarray data by self-consistency and local regression Genome Biol 2002 3 RESEARCH0037 12184811 10.1186/gb-2002-3-7-research0037
Wright GW Simon RM A random variance model for detecting of differential gene expression in small microarray experiments Bioinformatics 2003 19 2448 2455 14668230 10.1093/bioinformatics/btg345
Cui X Hwang JT Qiu J Blades NJ Churchill GA Improved statistical tests for differential gene expression by shrinking variance components estimates Biostatistics 2005 6 59 75 15618528 10.1093/biostatistics/kxh018
LIoyd CJ Statistical Analysis of Categorical Data 1999 New York, NY: JOhn Wiley & Sons
Breslow NE Extra-Poisson variation in log-linear models Applied Statistics 1984 33 38 44
R Development Core Team R: A language and environment for statistical computing 2003 R Foundation for Statistical Computing, Vienna, Austria
The website of overdispersed log-linear models for SAGE:
|
15987513
|
PMC1189357
|
CC BY
|
2021-01-04 16:27:23
|
no
|
BMC Bioinformatics. 2005 Jun 29; 6:165
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-165
|
oa_comm
|
==== Front
PLoS GenetPLoS GenetpgenplgeplosgenPLoS Genetics1553-73901553-7404Public Library of Science San Francisco, USA 1612125910.1371/journal.pgen.001002405-PLGE-RA-0093R2plge-01-02-11Research ArticleBioinformatics - Computational BiologyGenetics/GenomicsGenetics/Gene FunctionGenetics/Functional GenomicsSaccharomycesGenome-Wide Requirements for Resistance to Functionally Distinct DNA-Damaging Agents Genome-Wide Response to DNA DamageLee William 1St.Onge Robert P Proctor Michael 2Flaherty Patrick 34Jordan Michael I 5Arkin Adam P 46Davis Ronald W 12Nislow Corey 2Giaever Guri 2*1 Department of Genetics, Stanford University School of Medicine, Stanford, California, United States of America
2 Department of Biochemistry, Stanford University School of Medicine, Stanford Genome Technology Center, Palo Alto, California, United States of America
3 Department of Electrical Engineering and Computer Science, University of California, Berkeley, California, United States of America
4 Physical Biosciences Division, Lawrence Berkeley National Laboratory, Berkeley, California, United States of America
5 Division of Computer Science, Department of Statistics, University of California, Berkeley, California, United States of America
6 Howard Hughes Medical Institute, Department of Bioengineering, University of California, Berkeley, California, United States of America
Snyder Michael EditorYale University, United States of America*To whom correspondence should be addressed. E-mail: [email protected] 2005 19 8 2005 1 2 e2425 4 2005 1 7 2005 Copyright: © 2005 Lee et al.2005This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are properly credited.The mechanistic and therapeutic differences in the cellular response to DNA-damaging compounds are not completely understood, despite intense study. To expand our knowledge of DNA damage, we assayed the effects of 12 closely related DNA-damaging agents on the complete pool of ~4,700 barcoded homozygous deletion strains of Saccharomyces cerevisiae. In our protocol, deletion strains are pooled together and grown competitively in the presence of compound. Relative strain sensitivity is determined by hybridization of PCR-amplified barcodes to an oligonucleotide array carrying the barcode complements. These screens identified genes in well-characterized DNA-damage-response pathways as well as genes whose role in the DNA-damage response had not been previously established. High-throughput individual growth analysis was used to independently confirm microarray results. Each compound produced a unique genome-wide profile. Analysis of these data allowed us to determine the relative importance of DNA-repair modules for resistance to each of the 12 profiled compounds. Clustering the data for 12 distinct compounds uncovered both known and novel functional interactions that comprise the DNA-damage response and allowed us to define the genetic determinants required for repair of interstrand cross-links. Further genetic analysis allowed determination of epistasis for one of these functional groups.
Synopsis
Cells have evolved sophisticated ways to respond to DNA damage. This is critical because unrepaired damage can kill cells or cause them to become cancerous. The response to DNA damage has been studied for more than 50 years, and has been found to be extremely complex. The traditional way of understanding this complexity is to divide the process into its component parts with the goal of eventually reconstituting the entire process. In this study, the authors extend classical approaches using genomics—an approach that involves studying all genes in an organism simultaneously. The authors tested 12 distinct compounds (many used in cancer chemotherapy) that damage DNA and uncovered new genes involved in DNA repair. They then grouped the compounds to define how they attack cells. Using this approach, the study found that many similar DNA-damaging agents act in comparable ways to damage DNA, but surprisingly, similar compounds can also act on cells by very different mechanisms. Specifically grouping the findings together and verifying the significant results lends a high degree of confidence in the data. The development of such a reproducible experimental design is important for inspiring future experiments.
Citation:Lee W, St. Onge RP, Proctor M, Flaherty P, Jordan MI, et al. (2005) Genome-wide requirements for resistance to functionally distinct DNA-damaging agents. PLoS Genet 1(2): e24.
==== Body
Introduction
Eukaryotic cells possess multiple mechanisms to cope with structural damage to their DNA. For example, nucleotide excision repair (NER) excises oligonucleotides containing a covalently modified base and resynthesizes the deleted fragment using the undamaged DNA strand as template [1]. Cell-cycle checkpoints respond to damaged DNA by initiating a series of phosphorylation and dephosphorylation events that result in a transient arrest of the cell cycle, providing time for lesions to be repaired [2,3]. During DNA replication, multiple pathways ensure the stability and restarting of stalled replication forks at sites of DNA damage [4,5]. The sum total of these activities and similar functional repair “modules” are commonly referred to as the DNA-damage response (DDR).
The study of gene products and pathways that comprise the DDR has particular relevance to both the etiology and treatment of cancer in man. A causative role for the DDR in carcinogenesis is supported by several observations, including the high degree of genomic instability observed in tumor cells [6,7] and the number of DDR genes that, when mutated, lead to cancer or to one of several inherited disorders characterized by cancer predisposition [8]. Therapeutically, many cytotoxic agents used to treat cancer act by directly targeting DNA. Therefore, pathways that actively repair DNA lesions are likely to contribute to the significant problem of clinical drug resistance [9].
The high degree of conservation between the genes and pathways involved in maintaining genetic integrity in yeast and all metazoans supports the use of model organisms to better understand these pathways in man. Indeed, classical forward genetic screens in Saccharomyces cerevisiae have led to the identification of several genes important for responding to DNA damage, providing the parts list from which the current framework of the DDR has emerged [10,11]. The results of these screens have been augmented by the recent disruption and barcoding of each predicted yeast open reading frame which has proven to be a powerful tool in identifying additional components of the DDR [12–18].
In this study, we used fitness profiling [19–22] to interrogate a pooled collection of ~4,700 S. cerevisiae homozygous deletion strains for sensitivity to 12 agents known to compromise the structural integrity of DNA (Table 1). Fitness analysis of individual deletion strains identified in these global experiments confirmed our microarray data and revealed that the genetic requirements for resistance to DNA-damaging agents may exceed previous estimates. We discovered that those strains sensitive to these compounds carried deletions primarily in genes known to be involved in DNA metabolism, but we also uncovered genes not previously known to be related to the DDR. While resistance to a given compound typically required multiple DDR modules, we found that the relative importance of these modules was varied, even when comparing functionally related compounds. The significance of our results are 4-fold: (1) we developed a robust exportable assay to identify and confirm DDR components; (2) filtering and clustering the data allowed classification of both the mechanism of drug action and gene function; (3) we used epistasis analysis to identify novel functional relationships between DDR components; and (4) we were able to clearly discriminate the genome-wide response to agents that damage DNA by forming interstrand cross-links (ICLs) from those that do not.
Table 1 Summary of Compounds, with References Indicated
Results
Fitness Profiling of the Yeast Deletion Collection
The yeast deletion collection is a powerful tool for identifying genes important for fitness on a genome-wide scale under a diverse set of environmental conditions [13,16,20,22–24]. This resource has been particularly valuable in the study of cellular mechanisms that respond to DNA-damaging agents [12–14,16,18,25–27]. Each of these studies has provided new insights into the DDR. The underlying protocols in these well-executed studies are, however, so disparate that they preclude any direct comparisons beyond general conclusions. For example, some studies were performed on solid media, while others used high doses of compound followed by recovery in liquid media. Furthermore, the data analysis varies from study to study. To provide a consistent and comprehensive dataset of the DDR, we (1) profiled 12 unique DNA-damaging compounds (six of which had not previously been profiled) (Table 1) using a validated protocol [22]; (2) confirmed a subset of our microarray fitness data by individual strain analysis; and (3) where possible, correlated our results with previously published studies. Specifically, we sought to detect mechanistic differences between compounds that form ICLs (cisplatin, oxaliplatin, carboplatin, mechlorethamine, mitomycin C, and psoralen) and those that do not (angelicin, 4-nitroquinoline-1-oxide [4-NQO], 2-dimethylaminoethyl chloride [2-DMAEC], methyl methanesulfonate [MMS], streptozotocin, and camptothecin).
In our experiments, ~4,700 homozygous diploid deletion mutants were grown in pooled cultures in the presence of compound. Cells were then collected, genomic DNA purified, and the unique molecular barcodes present in each strain amplified by PCR and hybridized to an oligonucleotide array carrying the barcode complements. The relative fitness of each strain was then determined by comparing the signal intensity for each strain on the microarray to the corresponding intensities obtained from a series of no-drug control arrays (see Materials and Methods; Dataset S1; Protocol S1).
Validating the Approach by Individual Strain Confirmation
Little experimental evidence directly addresses how well fitness defects or sensitivities measured by microarray analysis correlate with actual growth rates of individually cultured strains. To directly address this issue, we cultured the 233 deletion strains most sensitive to mechlorethamine individually (as determined from three replicate microarray experiments, see Materials and Methods). The individual growth rates of these strains were measured, both in the presence and absence of mechlorethamine, by taking optical-density readings of liquid cultures every 15 min for 30 h (Dataset S2; Protocol S2). Figure 1A shows representative growth curves for 32 of these cultures (16 in dimethyl sulfoxide [DMSO, diluent control] and 16 in mechlorethamine). We defined the sensitivity to mechlorethamine of each strain by calculating the difference between the average doubling time (AvgG) in DMSO and in mechlorethamine (see Materials and Methods). These values were then normalized to wild-type and plotted against their corresponding fitness-defect scores as measured from the microarray (Figure 1B). We observed a highly significant correlation (R
2 = 0.5086, p = 8.5e−38; data not shown). When we removed strains exhibiting fitness defects in the absence of drug from the analysis, this correlation increased (R
2 = 0.7462; p = 5.4e−57). This is consistent with slow-growing strains yielding artificially low fitness-defect scores in microarray-based fitness analysis of pooled cultures (see Overall Experimental Design). Of 233 individual strains analyzed, 206 exhibited significant mechlorethamine-dependent fitness defects compared to that of a wild-type strain (see Materials and Methods).
Figure 1 Comparison of Individual Strain Analysis to Microarray Results
(A) Representative growth curves of 16 individual strains grown in the presence of solvent alone (DMSO, red curve) and 62.5 μM mechlorethamine (black curve). Growth was monitored by measuring the optical density (OD600) of cultures every 15 min for 30 h. The fitness of each strain was defined by the difference between the average doubling times in mechlorethamine and in DMSO (see Materials and Methods).
(B) Correlation between growth rates of individual strains and microarray-based fitness estimates. The ratio of growth rates of the 186 individual homozygous deletion strains (the top 233 ranked mechlorethamine-sensitive strains as determined by three replicate microarray experiments minus 47 strains which exhibited a slow-growth phenotype when individually cultured in the absence of mechlorethamine) over an average wild-type growth rate are plotted on the x-axis against the average fitness-defect scores from three pool experiments on the y-axis. The correlation (R
2 = 0.7462) is highly significant (p = 5.4e−57).
To further test how well our significance calling in our microarray experiments correlated with individual growth rates, we compared a calculated false-discovery rate (FDR) [28,29] with a measured FDR for the top 81 most sensitive strains according to the combined results for pooled growth in mechlorethamine (Table 2). The FDR calculation allows us to establish a fitness-defect score threshold for a given acceptable rate of false discovery. Specifically, we calculated the cut-off for several FDRs and examined the individual growth curves of strains that were designated as significant at these cut-offs. For example, the array results show that a 5% FDR cut-off results in 37 sensitive strains from the pooled experiments. The individual growth curves of these 37 strains reveal that 35 of them are indeed sensitive to mechlorethamine, giving us a 5.4% FDR (two out of 37). For this drug, our measured FDR is lower than the calculated FDR. It should be noted that although the above underscores the low FDR in our experiments, it can not address the extent of the false-negative rate.
Table 2 Calculated FDR Versus Experimentally Measured FDR
Overall Experimental Design
As we refined our assay for the DNA-damaging agents, it became apparent that the length of time for which cells are exposed to compound has a dramatic effect on the sensitivity of deletion strains lacking those genes most required for growth. Specifically, in our standard assay design, frozen aliquots of the deletion collection are recovered in rich media for ten generations until they reach logarithmic growth phase [22]. Compound is then added and the culture is robotically grown for precisely 20 generations. This prolonged chronic exposure allows for detection of those gene products that are required for resistance even for those gene products that have small (<5%) growth defects. We used this standard assay to profile four compounds: mechlorethamine, cisplatin, streptozotocin, and camptothecin. The results of these experiments made it clear that any deletion strain exhibiting a slow-growth phenotype in the absence of compound becomes depleted from the pool even before the compound is applied. As a consequence, this experimental design precludes any meaningful measure of the fitness of this subset of strains. To circumvent this problem, we altered our experimental protocol in the following way: frozen pool aliquots were thawed and immediately exposed to compound for the equivalent of five generations of wild-type growth.
Control experiments showed no non-specific sensitivity of the wild-type strain as a result of cells recovering from a frozen state using this protocol (data not shown). This revised experimental design was used to screen the 12 compounds (see Table 1) because (1) this low number of generations minimizes the false-negative rate resulting from slow-growing strains; and (2) these experiments should identify gene products immediately required for resistance to compound. The advantages of this design are demonstrated by examining the rad51Δ and rad52Δ homozygous deletion strains, both of which are defective for homologous recombination-mediated repair [30]. Both are acutely sensitive to mechlorethamine but also exhibit a slow-growth phenotype in the absence of mechlorethamine (see Figure 1A). In the standard 20-generation assay, however, the fitness-defect scores of these deletion strains in mechlorethamine ranked 1,676 and 2,051, respectively, among the other ~4,700 strains in the pool (where the strain most sensitive to mechlorethamine has a rank of 1). When the revised experimental design was used, the average rank of these strains increased to 21 and 23, respectively. Therefore, in the former case, neither strain was sensitive, but in the latter case, both were extremely sensitive.
Despite our observation that the shorter-generation assay detected many DDR genes, comparing the five-generation assay with the standard 20-generation assay yielded additional insight. For example, there is an increase in strains that are deleted for genes classified as “other” or “unknown” that exhibit fitness defects only after extended exposure (i.e. 20 generations) to compound (Figure 2A). Particularly interesting examples are strains defective in DNA-damage checkpoints. Table 3 summarizes the average rank of five checkpoint-defective strains (rad9Δ, rad24Δ, rad17Δ, ddc1Δ, and mec3Δ) following five or 20 generations of pool growth in 500 μM cisplatin. We followed up on this result by monitoring the growth rate of these individual strains during the first five and second five generations of growth in cisplatin (Figure 2B). Strikingly, both the rad9Δ and rad24Δ strains exhibited accelerated growth relative to wild-type during the first five generations, consistent with a lack of checkpoint-mediated cell-cycle arrest [31,32]. Despite this accelerated growth, these strains exhibited reduced viability even at the five-generation point (Figure 2C). Consistent with this observation during the next five generations (6–10), these mutants did exhibit a reduced rate of growth, presumably due to the accumulation of DNA and chromosomal damage associated with uncontrolled progression through the cell cycle. Notably, the growth rate of deletion strains rad17Δ, ddc1Δ, and mec3Δ, whose protein products form a complex loaded onto DNA at sites of damage in a RAD24-dependent manner [33], were indistinguishable from that of a rad24Δ strain (Figure 2C; data not shown). These results may explain the discrepancy between conflicting reports that address the requirement of checkpoint genes for resistance to cisplatin [18,34,35].
Figure 2 The Effect of Exposure Duration on the Genetic Requirements for Resistance to DNA-Damaging Agents
(A) Pie charts showing relative percentage of sensitive genes categorized into either “DNA metabolism”, “unknown”, or “other”. All of the Gene Ontology [63] slim annotations (ftp://ftp.yeastgenome.org/yeast/data_download/literature_curation/go_slim_mapping.tab. Accessed February 17, 2005) are combined into “other” except for those classified in the unknown or DNA-metabolism category. The relative distributions of the mechlorethamine experiments are shown as a function of both time and gene rank.
(B) BY4743 (wild-type) and the DNA-damage-checkpoint mutants, rad9Δ and rad24Δ, were grown in the presence of 500 μM cisplatin over the course of ten population doublings. Yeast cultures were maintained in an exponential phase of growth by robotic dilution of cultures after five doublings into fresh media containing cisplatin. The growth of each deletion strain (black curve) is compared to that of BY4743 (red curve) between 0–5 (left) and 6–10 (right) population doublings. The rad9Δ and rad24Δ deletion strains exhibit accelerated grow rates in the first five generations, but by ten generations begin to demonstrate slow growth as compared to wild-type.
(C) Viability assay of strains treated with cisplatin or DMSO for five generations. Indicated strains were removed from cultures after they reached an OD600 of 2.0, and were diluted as shown. Strains were chosen based on the criteria that they did not show a growth defect at five generations but did show a growth defect at 20 generations. Dilutions were pinned onto YPD plates in a 5-fold concentration series. The wild-type parental diploid strain is compared to several diploid deletion strains that exhibited a decrease in viability: ddc1Δ, rad24Δ, rad17Δ, mec3Δ, and rad9Δ. In contrast, several of these strains showed little or no decrease in viability at five generations of growth: swi4Δ, yme1Δ, rtf1Δ, and gcs1Δ. This figure underscores the point that, despite the lack of an apparent growth defect in liquid culture, several deletion strains lose viability relatively rapidly when exposed to cisplatin.
Table 3 Comparison of the Average Ranked Sensitivity of Five DNA-Damage-Checkpoint Mutants from Experiments in which the Pooled Homozygous Deletion Collection Was Exposed to 500 μM Cisplatin for Either Five or 20 Generations
In addition to checkpoint-defective strains, a number of other deletion strains exhibited sensitivity only after 20 generations of growth. These strains, however, do not exhibit accelerated growth at five generations but appear to be required only after long-term chronic exposure. These strains are deleted for genes that modify chromatin structure and are involved in the respiratory chain (Table S1). Despite the lack of an observable growth defect at five generations, we hypothesized that some of these strains, depending on the cause of their sensitivity, may have reduced viability. To test this, we plated serial dilutions of individual strains (that exhibited sensitivity after 20 generations, but not after five generations) on yeast extract/peptone/dextrose (YPD) following five generations of growth in 500 μM cisplatin (Figure 2C). Of the four we tested, gcs1Δ, swi4Δ, rtf1Δ, and yme1Δ, none showed an increased loss of viability. Liquid cultures of these strains confirmed these results, as a decrease in growth rate compared to wild-type is only observed at later-generation time points, and is not observed at the five-generation time point.
Interstrand DNA Cross-Linkers
In this study, we examined the effects of a variety of DNA-damaging agents, but placed particular emphasis on compounds that induce cross-links between complementary strands of DNA (the ICLs). Even though each of the six ICL-inducing compounds we profiled cause a variety of structural damage to DNA (see Table 1), the cytotoxic effects of these compounds are attributed primarily to ICLs [36]. These compounds differ in the efficacy with which they induce ICLs, ranging from 30–40% of all lesions for psoralen to 1–5% of all lesions for mechlorethamine [37]. They also differ in their preferred substrates. Mitomycin C, for example, typically cross-links guanines at CpG sequences [38], whereas psoralen predominantly cross-links thymines at TpA sequences [39].
Previous studies have suggested that ICL tolerance in yeast can be attributed to three major pathways: NER, homologous recombination repair (HRR), and post-replication repair (PRR) [18,35,40]. Our results corroborate these findings as strains carrying deletions in genes of the NER (RAD2, RAD4, RAD10, RAD1, and RAD14), HRR (RAD57, RAD55, RAD51, RAD52, RAD54, and RAD59), and PRR (RAD6, RAD18, and RAD5) pathways were found to be highly sensitive to each of the six ICL-inducing compounds we tested. Figure 3 depicts the relative requirement for each of several DNA-repair modules. Although the genes that make up these modules are somewhat arbitrarily assigned, we believe that they are generally acceptable in the DNA-damage community and, moreover, provide a simple visual depiction of some of the important differences between compounds. Figure 3 provides only an overview (for more detail, see Supporting Information). Examining the top 30 sensitive strains for a given compound and the relative sensitivity rankings for these strains gives a snapshot of the first line of defense against the damage induced by that compound. The gene groups with the lowest ranks contain genes that, when deleted, result in the greatest sensitivity to that drug and therefore are presumably most important in conferring resistance. Extending this list to the top 250 sensitive strains then begins to show a more general view of other DDR pathways that are also involved in responding to damage caused by that agent.
Figure 3 The Relative Importance of Well-Characterized DDR Modules for Resistance to Different DNA-Damaging Agents
Detailed examination of strains with mutations in DNA-damage-response genes. Each bar graph represents only strains that were found to be among the top 30 (or 250) most sensitive strains in that compound and are known to be members of a well-characterized DNA-damage-response pathway. The bars represent the median rank for genes in each of the gene groups listed in the visual key. The gene groups were defined in the following way: x-linking genes (PSO2); NER (RAD2, RAD4, RAD10, RAD1, and RAD14); PRR (RAD6, RAD18, and RAD5); error-prone TLS (REV1, REV3, and REV7); HRR (RAD57, RAD55, RAD51, RAD52, RAD54, and RAD59); stalled replication-fork repair (MUS81 and MMS4). Those compounds that form ICLs are labeled with an asterisk.
The genes from these pathways are also important for resistance to additional types of DNA damage, as is evidenced by the profiles of other compounds (Figure 3). Therefore, their sensitivity to cross-linking agents could be due to a combined deficiency in the repair of ICLs as well as other DNA lesions. To identify strains whose sensitivity was entirely due to a lack of ICL repair, we screened the compounds angelicin and 2-DMAEC, monofunctional analogues of psoralen and mechlorethamine, respectively. The most striking difference between these profiles was the relative sensitivity of the pso2Δ deletion strain, which ranked among the most sensitive strains in psoralen and mechlorethamine (and indeed all other cross-linking compounds), but which exhibited no detectable sensitivity to angelicin or 2-DMAEC. These observations are consistent with previous findings for PSO2 [41], whose protein product is thought to facilitate the repair of ICLs as a part of the NER pathway [40]. Further comparisons between ICL and non-ICL-inducing compounds underscored the importance of translesion DNA synthesis (TLS) to the repair of ICLs. This is consistent with previous findings for REV1, REV3, and REV7, which operate in a branch of PRR [42,43]. Unlike PSO2, however, the role of these genes in the DDR is unlikely to be restricted to ICL repair [44].
Of the six ICL-inducing compounds, carboplatin and mitomycin C resistance required a significantly larger number of genes whose biological function was not previously linked to DNA metabolism. Generally speaking, however, with respect to well-characterized DNA-damage-responsive pathways, there was a high level of correlation between the genetic requirements for resistance among the ICL-inducing compounds that we profiled. There were, however, some exceptions to this. Genes encoding the MRX complex (MRE11, RAD50, and XRS2), involved in the repair of DNA double-strand breaks [45], were particularly important for resistance to mechlorethamine (see Supporting Information). Strains deleted for the genes MUS81 and MMS4, whose protein products are thought to form a structure-specific endonuclease important for restarting stalled replication forks [46,47], were highly sensitive to each cross-linker with the exception of psoralen. Finally, resistance to carboplatin appeared less dependent on NER machinery when compared to other cross-linkers, including other members of the platinum family. This is surprising because the mechanism of action and the spectrum of clinical activity of cisplatin and carboplatin are more similar to each other than to oxaliplatin [48].
Non-Cross-Linking DNA-Damaging Agents
We also profiled five compounds that covalently modify DNA but do not induce ICLs. These included the aforementioned angelicin and 2-DMAEC, as well as streptozotocin, MMS, and 4-NQO. Streptozotocin and MMS are both clinically used cytotoxics that methylate DNA primarily at the N-7 position of guanine residues. They also generate, albeit at a lower frequency, the more toxic O-6 methylguanine and N-3 methyladenine adducts [49,50]. 4-NQO reacts with DNA to form several high-molecular-weight adducts including 3-(deoxyguanosine-N2-yl)-4-aminoquinoline-1-oxide, N-(deoxyguanosine-C8-yl)-4-aminoquinoline-1-oxide, and 3-(deoxyadenosine-N6-yl)-4-aminoquinoline-1-oxide [51].
The fitness profiles of these five DNA-modifying compounds were similar in so far as strains deficient in HRR, PRR, and MUS81/MMS4 exhibited sensitivity to all five compounds (Figure 3). This likely reflects a common ability of these compounds to cause lesions that impede the advancement of DNA-replication forks during S-phase. There were also several differences in the five fitness profiles. To remove DNA adducts, the cell can employ one of several excision-repair mechanisms, the selection of which is based on the size of the offending DNA adduct [52]. In agreement with this, we found that NER was important for resistance to compounds that cause bulkier DNA adducts (angelicin and 4-NQO), whereas MAG1, encoding a DNA glycosylase of base excision repair [53,54], was important for resistance to the methylating activities of streptozotocin and MMS. Interestingly, MAG1 appeared to be more important for resistance to MMS than to streptozotocin, based on rank.
A sixth non-cross-linking agent profiled, camptothecin, is distinct from the other compounds in this study because it inhibits the activity of an enzyme rather than directly modifying DNA. Camptothecin inhibits the enzyme topoisomerase I (Top I), and stabilizes a covalent enzyme–DNA intermediate [55]. During replication, collision of the replication machinery with this protein–DNA complex causes double-strand breaks. Our assay revealed a unique sensitivity profile for camptothecin. Strains deficient in the MUS81/MMS4 complex and HRR showed the greatest sensitivity to camptothecin, as expected for a treatment that induces stalled replication forks. Neither PRR nor excision-repair mechanisms appeared to play a prominent role in conferring resistance to camptothecin.
Uncharacterized Genes
Of the 36 array experiments examined in this study, 283 strains scored significantly sensitive in at least one experiment when a 5% FDR threshold was used. Among these 283 strains, 87 are sensitive in three or more independent experiments and four of these strains contain deletions in unnamed genes (Table S2). The most notable of these four strains is YDR291W, a gene encoding a putative DNA helicase that appears to be involved in repairing mechlorethamine-induced lesions. The amino acid sequence of Ydr291w is conserved across many organisms, including mammals (BLASTP E-value = 3.0e−08 to Human RecQ4 [56]), and its proposed helicase function is due to its helicase sequence motifs [57]; the function of this yeast open reading frame is currently being tested. Ymr073c is also conserved from yeast to man (BLASTP E-value = 7.0e−09 to a human cytochrome oxidoreductase), and our results suggest it is involved in resistance to carboplatin and cisplatin. The ylr426wΔ deletion strain showed sensitivity in mechlorethamine and streptozotocin, and its protein product is similar to a human peptide annotated as a dehydrogenase (BLASTP E-value = 6.0e−05). The YKL075C mutant strain was sensitive in both streptozotocin and camptothecin at 20 generations of exposure, and is only conserved with another fungus, Ashbya gossypii. All four of these unnamed genes have detectable protein expression in S. cerevisiae (>50 molecules/cell) [58,59].
The paucity of unnamed genes in our results is likely due to the stringency of the significance cut-off we used to minimize false positives as well as being a consequence of the level of attention that DNA-repair pathways have received. An equally important factor is the finding that many of the named genes do not properly reflect the complexity of their cellular roles. Specifically, many of the sensitive strains that carry three-letter names are poorly characterized and, more importantly, their roles in the DDR are largely unknown. For example, the RMD7 mutant appears sensitive to several cross-linking compounds yet is named for being required for meiotic division [60]. Other genes are better characterized, but thus far are only known to have functions not involved in the DDR (e.g., RTT101 [regulates Ty1 transposition, though other RTT genes are DDR genes, such as ELG1, MMS1, and RTT107], and LTE1 [essential for growth in low temperatures]) [61,62]. Overall, the 283 strains making the sensitivity cut-off in at least one experiment contain 34 strains deleted for genes with no functional annotation (by Gene Ontology [63]), and roughly half are annotated in processes unrelated to DNA repair (Table S2). Our results provide functional data that these genes are in fact involved in the DDR and suggest additional experiments to further characterize them.
Global Analysis
To analyze the 36 array experiments presented here in a global manner, we applied hierarchical clustering techniques to the fitness-defect scores (Figure 4A). The dataset used for clustering was restricted to strains designated as significant (5% FDR) in two or more of the array experiments, and thus represented only the most highly sensitive strains identified in this study. This produced a matrix of fitness-defect scores (141 strains × 36 experiments). An examination of the clusters across the experiment axis showed that, as expected, compounds predicted to have similar mechanisms of action group together, lending confidence in the robustness of the analysis. Members of the platinum family of compounds are highly correlated, as are MMS and streptozotocin, the only two agents profiled that possess DNA-methylating activity. The genome-wide data was highly reproducible, evidenced by the fact that all replicates correlate as nearest neighbors.
Figure 4 Hierarchical Clustering of Genome-Wide Profiles Identifies Mechanistic Relationships Between Drugs and Functional Relationships Between Genes
(A) Clustergram containing all strains significant in two or more array experiments. Raw fitness-defect values were hierarchically clustered using Spearman's rank correlation. Colored bars represent gene clusters of note, including NER (RAD2, RAD4, RAD10, RAD14, and RAD1—blue); error-prone TLS (REV1 and REV3—red); PRR (RAD6, RAD18, and RAD5—yellow); homologous recombination (RAD57, RAD51, and RAD54—green); cell-cycle checkpoint control (RAD9, RAD24, RAD17, DDC1, and MEC3—orange); and a cluster shown in (B) (SHU2, SHU1, CSM2, MPH1, and PSY3—magenta).
(B) Zoom view showing one cluster containing the class I NER genes and a second cluster containing several uncharacterized DNA-repair genes. Four of these five genes (SHU1, SHU2, CSM2, and PSY3) are known to encode proteins that physically interact [65,77,78].
(C) Individual growth curves of single and double deletion strains with MPH1 in 0.002% MMS. In all panels, the growth of wild-type (BY4741) is represented by the black curve and the growth of mph1Δ by the red curve. The growth of eight different deletion strains (shu1Δ, shu2Δ, csm2Δ, psy3Δ, mag1Δ, mus81Δ, rad51Δ, and rad54Δ) are shown in green, and double mutants, in which the MPH1 deletion is added to each of the above, are shown in blue. Double mutants of MPH1, MAG1, and MUS81 show additive or synergistic sensitivity to MMS, whereas double mutants of MPH1, with the four other genes in its cluster, show no additional sensitivity to MMS, suggesting that MPH1 is epistatic with SHU1, SHU2, CSM2, and PSY3.
When the clustered results are viewed along the gene axis, we predicted there would be a high correlation in fitness profiles between strains with deficiencies in genes in the same DNA-repair pathway (Figure 4B), and this prediction was confirmed. This result holds true for several DNA-repair epistasis groups, including NER, PRR, HRR, error-prone TLS, and cell-cycle checkpoint control (see Figure 4A).
The clustering analysis suggested additional novel functional interactions. We focused on one particular cluster within the dataset that included five genes: MPH1, SHU1, SHU2, CSM2, and PSY3. The latter four genes encode proteins that physically interact with one another and belong to a single epistasis group that appears to operate in a branch of HRR [64,65]. MPH1 was also shown to be epistatic with members of the HRR pathway [66], and the Mph1 protein, as predicted by its amino acid sequence, was recently shown to possess DNA helicase activity [67]. To our knowledge, however, no direct functional link between MPH1 and the other four genes has been reported. To test whether such a link exists, as suggested by our clustering analysis, we generated double mutants between MPH1 and each of the other four genes and measured the sensitivity of these strains to MMS. We found that single and double deletion strains exhibited similar sensitivity to MMS, indicating that MPH1 is epistatic with SHU1, SHU2, CSM2, and PSY3 (Figure 4C). Conversely, the addition of an MPH1 deletion to strains carrying deletions known to compromise other defined DNA-damage pathways, such as MUS81 (stalled replication-fork repair) and MAG1 (base excision repair), exacerbated the sensitivity to MMS (Figure 4C).
Discussion
The high level of scrutiny paid to the DDR in yeast and other model organisms has yielded a wealth of data. In this study, we set out to address two aims. First, we sought to collect a sufficient number of fitness-profile signatures for compounds with similar mechanisms of action. This would allow us to test whether we could identify differences in compound profiles that might provide correlative insight into their diverse clinical efficacy. Second, we aimed to develop a standard assay platform, including all protocols and strains so that they can be exported to any laboratory. We believe that such a standard set of protocols will provide a powerful resource to the entire community and will allow for the development of a centralized database for the exchange of information.
While classical forward genetic screens in yeast have been instrumental in identifying important components of the DDR, it is unlikely that saturation across the genome has been achieved. In contrast to traditional genetic analysis, the approach employed here enabled the systematic interrogation of the comprehensive set of all 4,758 homozygous deletion strains, representing nearly all non-essential genes in the yeast genome. Furthermore, because the fitness-defect scores obtained with this method display a continuous range of sensitivity and these sensitivities correlate well with independently measured growth rates (see Figure 1B), genes can be ranked according to their relative importance in conferring resistance to a given compound. This provides a high level of resolution and allows distinctions to be made between functionally related compounds, even in cases where the genetic elements required for fitness largely overlap. For example, cisplatin, oxaliplatin, and carboplatin are structurally related chemotherapeutic drugs and, for each, their cytotoxicity is attributed primarily to their ability to induce ICLs. Predictably, fitness profiling of each compound identified the major pathways important for resistance to DNA cross-linking agents (PSO2/NER, HRR, and PRR). The fitness profile for carboplatin, however, was unique among the platinum-based compounds in that PRR was found to hold far greater importance in conferring resistance than the other pathways (see Figure 3). Carboplatin's distinctive profile is also apparent upon hierarchical clustering of the data (see Figure 4A), and could potentially reflect differences in the spectrum of structural damage to DNA or in cell-cycle-specific reactivity of the drug.
Although DNA-targeting drugs are typically viewed as non-specific, the cellular response to both diverse and functionally similar DNA-damaging agents is clearly distinct as is evidenced by their specific clinical applications [68]. It is important to note that most of these agents were used before their mechanism of action had been defined, and their effectiveness is still defined empirically. Given this fact and the fact that DNA will remain an excellent therapeutic oncology target, particularly as more selective and less toxic compounds are developed, high-throughput assays capable of defining drug action will undoubtedly prove useful. In this study, we used 12 compound profiles (Table 1) to define a benchmark from which the application of this assay to novel DNA-damaging compounds can be compared.
Because the measured sensitivity of each strain in a given profile is a continuum of sensitivity as measured by growth inhibition, it is difficult to gauge the point at which fitness-defect scores cease to reflect significant defects in growth. We therefore sought to confirm the sensitivity detected in competitive pools by growing individual strains +/− the compound mechlorethamine. We found that 88% of the 233 strains most sensitive to mechlorethamine (based on rank) were also found to be sensitive in our individual growth assay (see Table 2). Thus, our assay has the advantage of having a measurable FDR that appears much lower than our calculated FDR [69]. This result is significant on several levels. First, we found 34 uncharacterized genes suggesting, as predicted, that our understanding of the DDR is still incomplete. Second, these results suggest that by choosing a stringent significance cut-off, we have likely missed other DDR genes, and the raw data from these screens should provide a rich data source for further analysis. Added to this complexity is our observation that several strains exhibit compound sensitivity only following extended drug exposure. By conducting our assay for both five and 20 generations of growth, we have gleaned a more comprehensive representation of the DDR. This enabled us to establish definitively a role for five checkpoint genes in conferring resistance to cisplatin (see Figure 2)—an observation for which there has been no consensus in the literature.
Our assay is distinct from expression studies, where an equivalent confirmation assay would represent hundreds of Northern blots or quantitative PCR assays to verify expression changes. In addition, expression arrays do not allow a ranking of genes as it is not clear how or if the magnitude of expression change is biologically significant. The fact that our assay allows a ranking of sensitivities, and confirmation of this ranking, makes it distinct from several colony-based studies using the yeast deletion collection. These solid-media assays for sensitivity are not highly quantitative, though spot-dilution confirmation does allow some measure of relative sensitivities.
In summary, despite significant advances in the breadth and depth of our understanding of the DDR, additional cell-based analyses are required before we can claim to fully understand DNA replication, recombination, and repair [70]. We believe that a new theme has emerged in these studies, however, in that the combined use of chemicals with genetic deletions can define functional groups that, in the absence of compound, would not be uncovered. For example, hierarchical clustering of the most highly sensitive strains identified using a combination of chemical and genetic perturbations successfully predicted novel epistatic interactions between the MPH1 gene and SHU1, SHU2, CSM2, and PSY3. Even though they comprise a major defense mechanism against several DNA-damaging agents, a role for these five gene products in a branch of HRR is only now beginning to emerge [65,66]. Further analysis of this dataset will undoubtedly uncover additional functional relationships whose discovery has thus far remained elusive. Finally, increased use (and increased availability) of chemical inhibitors to probe the yeast deletion collection should provide a more comprehensive, contextual understanding of cellular physiology.
Materials and Methods
Reagents.
Cisplatin, carboplatin, oxaliplatin, psoralen, angelicin, streptozotocin, 2-DMAEC, mechlorethamine, MMS, camptothecin, and 4-NQO were purchased from Sigma-Aldrich (St. Louis, Missouri, United States). Mitomycin C was purchased from Calbiochem (San Diego, California, United States). Each of these compounds was dissolved in DMSO, aliquoted, and stored at −20 °C until use, with the exception of cisplatin, which was used immediately following resuspension, and carboplatin, oxaliplatin, and 2-DMAEC, which were dissolved directly in YPD media at the time of the experiment.
Strains and media.
Yeast was maintained in YPD media [71,72] at 30 °C unless stated otherwise. Strains used for individual analysis in this study are listed in Table S3 and were obtained from the yeast deletion collection or constructed de novo using PCR-based gene replacement [73]. Using this method, the nat or hph genes, conferring resistance to the antibiotics nourseothricin or hygromycin B, respectively [74], were PCR-amplified from plasmids and transformed into the appropriate background strain using a modified LiOAc-based protocol [75,76]. Transformants were selected on YPD containing antibiotic (100 μg/ml nourseothricin or 300 μg/ml hygromycin B), and successful gene-replacement events were verified by PCR.
Deletion-pool construction and growth.
The homozygous deletion pool was constructed as described [22] and stored in 20-μl aliquots at −80 °C. For the five-generation experiments, aliquots of the pool were thawed and diluted in YPD to an optical density at 600 nm (OD600) of 0.0625 and a final volume of 700 μl. Compound was then added and the pool was grown for five generations in a Tecan GENios microplate reader (Tecan, Durham, North Carolina, United States). For the 20-generation experiments, the thawed pool was first recovered for ten generations of logarithmic growth, diluted as above, and grown in the presence of compound over 20 generations. Cells were maintained in logarithmic phase by robotically diluting cultures every five doublings using a Packard Multiprobe II four-probe liquid-handling system (PerkinElmer, Wellesley, California, United States) [23].
Experiments involving the two UVA light-activated compounds (psoralen and angelicin) were performed as described above with the following modifications: the pool was diluted to OD600 of 0.625 and compound was added to 70 μl of this cell suspension, which was then exposed to UVA light via a handheld UVA lamp (Ambion, Austin, Texas, United States) for 15 min. Following irradiation, cells were diluted 10× with YPD to a final volume of 700 μl and grown as described above.
Genomic DNA preparation, PCR, and microarray hybridization.
Genomic DNA preparation, PCR amplification of molecular tags, and microarray hybridization were as described (Datasets S3 and S4; Protocol S3) [23].
Data analysis.
Fitness-defect scores were calculated for each strain in the pool for each experiment. These scores are based on a tag-specific algorithm that takes into account the intensities of each tag on the experimental array and the corresponding intensities on a set of control arrays performed on the pool without compound (the control set). The majority of strains carry four tags that hybridize to the array: upstream tag (uptag), uptag antisense, downstream tag (dntag), and dntag antisense. The tag intensities are log transformed, mean normalized, and the intensities of the two strands of each tag are averaged into a single value for each tag. Next, a mean and standard deviation are calculated for the uptag and dntag intensities for each strain across the set of control arrays. A z-score for upstream and downstream tags for each strain is then calculated by taking the difference of the average intensities between the control and treatment and dividing by the standard deviation of the control-set intensities. The result is two z-score values for the upstream and downstream tags; these are then averaged into a single fitness-defect score for the strain. A significance cut-off for each experiment was determined by calculating the score cut-off required for a 0.05 FDR [28,29].
Clustering analysis.
Hierarchical clustering analysis was performed with Cluster 3.0 (http://bonsai.ims.u-tokyo.ac.jp/~mdehoon/software/cluster/software.htm) in Windows and Linux and visualized using Java Treeview (http://jtreeview.sourceforge.net/) and slcview (http://slcview.stanford.edu). Clustering along both experiment and gene axes was performed on the calculated fitness-defect scores (as described above) using average linkage and Spearman's rank correlation as a distance metric (Dataset S5).
Confirmations of individual strain growth rates.
Individual yeast strains were first grown to saturation for approximately 20 h. Cells were then diluted to an OD600 of 0.02 in a final volume of 100 μl using a Biomek FX Laboratory Automation Workstation (Beckman Coulter, Allendale, New Jersey, United States). Normalized cultures were grown in 96-well plates (Nunc, Rochester, New York, United States) in Tecan GENios microplate readers (Tecan) for up to 30 h. The growth rate of each culture was monitored by measuring the OD600 every 15 min and calculating the average doubling time (AvgG). AvgG was calculated by recording the time (Δt) from the start of growth until the optical density of the culture reached the calibrated five-generation point (OD5g) and dividing this by the number of generations, i.e., five. If the culture did not reach OD5g, AvgG was calculated by using a binomial search to determine the fractional generations of the optical density at the end of the growth (ODf), assuming an exponential growth rate, then dividing the time to ODf by the generations calculated.
We defined the sensitivity to mechlorethamine of each individual strain as the difference between the AvgG in mechlorethamine and in DMSO (ΔAvgG = AvgGmech − AvgGDMSO). Individual deletion strains were scored as sensitive to mechlorethamine if this difference was greater than that of wild-type +/− the standard deviation calculated from 18 wild-type replicates.
Viability assays.
Selected strains were grown for five generations in the presence of compound and cells collected. Cells were transferred to 96-well plates and 5-fold dilutions were prepared. These dilutions were then “stamped” onto YPD plates using a pintool calibrated to deliver 5 μl. Cells were then allowed to form colonies for 2 d at 30 °C.
Supporting Information
The supporting files are also available in a searchable format at http://chemogenomics.stanford.edu.
Dataset S1 Analyzed Microarray Data
Calculated fitness-defect scores for each microarray experiment are presented.
(7.2 KB XLS)
Click here for additional data file.
Dataset S2 Complete Set of Individual Strain Growth-Curve Data
Each strain has three growth curves corresponding to growth in 1% DMSO, 31.3 μM mechlorethamine, and 62.5 μM mechlorethamine, respectively.
(1.7 MB PNG)
Click here for additional data file.
Dataset S3 Unanalyzed, Raw Microarray Data from Affymetrix Software, Part 1
This file contains 18 .cel files output from Affymetrix GeneChip operating system; these files constitute half of the total microarray dataset.
(9.2 MB ZIP)
Click here for additional data file.
Dataset S4 Unanalyzed, Raw Microarray Data from Affymetrix Software, Part 2
This file contains 18 .cel files output from Affymetrix GeneChip Operating System; these files constitute half of the total microarray dataset.
(8.5 MB ZIP)
Click here for additional data file.
Dataset S5 Files Containing Cluster Output Used in Generating Figure 4A
These text files can be opened by clustergram-viewing software to browse the cluster shown in Figure 4A.
(24 KB TXT)
Click here for additional data file.
Protocol S1 Document Describing the Data Included in Dataset S1
(26 KB DOC)
Click here for additional data file.
Protocol S2 Document Describing the Data Included in Dataset S2
(24 KB DOC)
Click here for additional data file.
Protocol S3 Document Describing the Data Included in Datasets S3 and S4
(84 KB DOC)
Click here for additional data file.
Table S1 Strains Significantly Sensitive in Two or More 20-Generation Experiments that Exhibit Greater Sensitivity in Long-Term Drug Exposure
(61 KB DOC)
Click here for additional data file.
Table S2 Number of Times Strains Were Calculated as Significantly Sensitive for Each Drug Condition and Length of Drug Exposure
(80 KB XLS)
Click here for additional data file.
Table S3 Individual Strains Used in this Study
(39 KB DOC)
Click here for additional data file.
Accession Numbers
Swiss-Prot (http://www.ebi.ac.uk/swissprot) accession numbers for the strains discussed in this paper are as follows: CSM2 (P40465), DDC1 (Q08949), ELG1 (Q12050), LTE1 (P07866), MAG1 (P22134), MEC3 (Q02574), MMS1 (Q06211), MMS4 (P38257), MPH1 (P40562), MRE11 (P32829), MUS81 (Q04149), PSO2 (P30620), PSY3 (Q12318), RAD1 (P06777), RAD2 (P07276), RAD4 (P14736), RAD5 (P32849), RAD6 (P06104), RAD9 (P14737), RAD10 (P06838), RAD14 (P28519), RAD17 (P48581), RAD18 (P10862), RAD24 (P32641), RAD50 (P12753), RAD51 (P25454), RAD52 (P06778), RAD54 (P32863), RAD55 (P38953), RAD57 (P25301), RAD59 (Q12223), REV1 (P12689), REV3 (P14284), REV7 (P38927), RMD7 (P40056), RTT101 (P47050), RTT107 (P38850), SHU1 (P38751), SHU2 (P38957), YDR291W (Q05549), YKL075C (P36083), YLR426W (Q06417), YMR073C (Q04772), and XRS2 (P33301).
We thank Grant Brown for critical comments and advice on the manuscript. WL was supported by the Stanford Genome Training Program. RPS was supported by a postdoctoral fellowship from the Canadian Institutes of Health Research. AA and PF were funded by a grant from the Howard Hughes Medical Institute. This work was supported by a grant from the National Cancer Institute and the National Institute of Biomedical Imaging and Bioengineering.
Competing interests. The authors have declared that no competing interests exist.
Author contributions. WL, RPS, CN, and GG conceived and designed the experiments. WL and RPS performed the experiments. PF and WL analyzed the data. MP designed the robotic software and built the robotics platform. APA and MIJ provided the analysis tools and assistance. WL, RPS, CN, and GG wrote the paper. RWD provided a nurturing environment and valuable intellectual insights.
Abbreviations
DDRDNA-damage response
2-DMAEC2-dimethylaminoethyl chloride
DMSOdimethyl sulfoxide
FDRfalse-discovery rate
HRRhomologous recombination repair
ICLinterstrand cross-link
MMSmethyl methanesulfonate
NERnucleotide excision repair
4-NQO4-nitroquinoline-1-oxide
PRRpost-replication repair
TLStranslesion DNA synthesis
YPDyeast extract/peptone/dextrose
==== Refs
References
Friedberg EC 2001 How nucleotide excision repair protects against cancer Nat Rev Cancer 1 22 33 11900249
Zhou BB Elledge SJ 2000 The DNA damage response: Putting checkpoints in perspective Nature 408 433 439 11100718
Kastan MB Bartek J 2004 Cell-cycle checkpoints and cancer Nature 432 316 323 15549093
Kraus E Leung WY Haber JE 2001 Break-induced replication: A review and an example in budding yeast Proc Natl Acad Sci U S A 98 8255 8262 11459961
Michel B Grompone G Flores MJ Bidnenko V 2004 Multiple pathways process stalled replication forks Proc Natl Acad Sci U S A 101 12783 12788 15328417
Rajagopalan H Lengauer C 2004 Aneuploidy and cancer Nature 432 338 341 15549096
Loeb LA 1994 Microsatellite instability: Marker of a mutator phenotype in cancer Cancer Res 54 5059 5063 7923117
Vogelstein B Kinzler KW 2004 Cancer genes and the pathways they control Nat Med 10 789 799 15286780
Longley DB Johnston PG 2005 Molecular mechanisms of drug resistance J Pathology 205 275 292
Prakash L Prakash S 1977 Isolation and characterization of MMS-sensitive mutants of Saccharomyces cerevisiae
Genetics 86 33 55 195865
Cox BS Parry JM 1968 The isolation, genetics and survival characteristics of ultraviolet light-sensitive mutants in yeast Mutat Res 6 37 55 5708072
Chang M Bellaoui M Boone C Brown GW 2002 A genome-wide screen for methyl methanesulfonate-sensitive mutants reveals genes required for S phase progression in the presence of DNA damage Proc Natl Acad Sci U S A 99 16934 16939 12482937
Birrell GW Giaever G Chu AM Davis RW Brown JM 2001 A genome-wide screen in Saccharomyces cerevisiae for genes affecting UV radiation sensitivity Proc Natl Acad Sci U S A 98 12608 12613 11606770
Bennett CB Lewis LK Karthikeyan G Lobachev KS Jin YH 2001 Genes required for ionizing radiation resistance in yeast Nat Genet 29 426 434 11726929
Birrell GW Brown JA Wu HI Giaever G Chu AM 2002 Transcriptional response of Saccharomyces cerevisiae to DNA-damaging agents does not identify the genes that protect against these agents Proc Natl Acad Sci U S A 99 8778 8783 12077312
Hanway D Chin JK Xia G Oshiro G Winzeler EA 2002 Previously uncharacterized genes in the UV- and MMS-induced DNA damage response in yeast Proc Natl Acad Sci U S A 99 10605 10610 12149442
Parsons AB Brost RL Ding H Li Z Zhang C 2004 Integration of chemical-genetic and genetic interaction data links bioactive compounds to cellular target pathways Nat Biotechnol 22 62 69 14661025
Wu HI Brown JA Dorie MJ Lazzeroni L Brown JM 2004 Genome-wide identification of genes conferring resistance to the anticancer agents cisplatin, oxaliplatin, and mitomycin C Cancer Res 64 3940 3948 15173006
Giaever G Chu AM Ni L Connelly C Riles L 2002 Functional profiling of the Saccharomyces cerevisiae genome Nature 418 387 391 12140549
Winzeler EA Shoemaker DD Astromoff A Liang H Anderson K 1999 Functional characterization of the S. cerevisiae genome by gene deletion and parallel analysis Science 285 901 906 10436161
Steinmetz LM Scharfe C Deutschbauer AM Mokranjac D Herman ZS 2002 Systematic screen for human disease genes in yeast Nat Genet 31 400 404 12134146
Giaever G Flaherty P Kumm J Proctor M Nislow C 2004 Chemogenomic profiling: Identifying the functional interactions of small molecules in yeast Proc Natl Acad Sci U S A 101 793 798 14718668
Giaever G Chu AM Ni L Connelly C Riles L 2002 Functional profiling of the Saccharomyces cerevisiae genome Nature 418 387 391 12140549
Deutschbauer AM Jaramillo DF Proctor M Kumm J Hillenmeyer ME 2005 Mechanisms of haploinsufficiency revealed by genome-wide profiling in yeast Genetics 169 1915 1925 15716499
Parsons AB Geyer R Hughes TR Boone C 2003 Yeast genomics and proteomics in drug discovery and target validation Prog Cell Cycle Res 5 159 166 14593709
Begley TJ Rosenbach AS Ideker T Samson LD 2002 Damage recovery pathways in Saccharomyces cerevisiae revealed by genomic phenotyping and interactome mapping Mol Cancer Res 1 103 112 12496357
Birrell GW Brown JA Wu HI Giaever G Chu AM 2002 Transcriptional response of Saccharomyces cerevisiae to DNA-damaging agents does not identify the genes that protect against these agents Proc Natl Acad Sci U S A 99 8778 8783 12077312
Efron B 2004 Large-scale simultaneous hypothesis testing: The choice of a null hypothesis J Am Stat Assoc 99 96 104
Benjamini Y Hochberg Y 1995 Controlling the false discovery rate—A practical and powerful approach to multiple testing J R Stat Soc B Methodological 57 289 300
Symington LS 2002 Role of RAD52 epistasis group genes in homologous recombination and double-strand break repair Microbiol Mol Biol Rev 66 630 670 12456786
Weinert TA Hartwell LH 1990 Characterization of RAD9 of Saccharomyces cerevisiae and evidence that its function acts posttranslationally in cell cycle arrest after DNA damage Mol Cell Biol 10 6554 6564 2247073
Paciotti V Lucchini G Plevani P Longhese MP 1998 Mec1p is essential for phosphorylation of the yeast DNA damage checkpoint protein Ddc1p, which physically interacts with Mec3p EMBO J 17 4199 4209 9670034
Majka J Burgers PM 2003 Yeast Rad17/Mec3/Ddc1: A sliding clamp for the DNA damage checkpoint Proc Natl Acad Sci U S A 100 2249 2254 12604797
Simon JA Szankasi P Nguyen DK Ludlow C Dunstan HM 2000 Differential toxicities of anticancer agents among DNA repair and checkpoint mutants of Saccharomyces cerevisiae
Cancer Res 60 328 333 10667584
Grossmann KF Ward AM Matkovic ME Folias AE Moses RE 2001
S. cerevisiae has three pathways for DNA interstrand crosslink repair Mutat Res 487 73 83 11738934
Lawley PD Phillips DH 1996 DNA adducts from chemotherapeutic agents Mutat Res 355 13 40 8781575
Dronkert ML Kanaar R 2001 Repair of DNA interstrand cross-links Mutat Res 486 217 247 11516927
Tomasz M 1995 Mitomycin C: Small, fast and deadly (but very selective) Chem Biol 2 575 579 9383461
Gamper H Piette J Hearst JE 1984 Efficient formation of a crosslinkable HMT monoadduct at the Kpn I recognition site Photochem Photobiol 40 29 34 6091158
Henriques JA Moustacchi E 1981 Interactions between mutations for sensitivity to psoralen photoaddition (pso) and to radiation (rad) in Saccharomyces cerevisiae
J Bacteriol 148 248 256 7026532
Ruhland A Kircher M Wilborn F Brendel M 1981 A yeast mutant specifically sensitive to bifunctional alkylation Mutat Res 91 457 462 7027037
Lawrence CW Christensen R 1976 UV mutagenesis in radiation-sensitive strains of yeast Genetics 82 207 232 770231
Broomfield S Hryciw T Xiao W 2001 DNA postreplication repair and mutagenesis in Saccharomyces cerevisiae
Mutat Res 486 167 184 11459630
Kunz BA Straffon AF Vonarx EJ 2000 DNA damage-induced mutation: Tolerance via translesion synthesis Mutat Res 451 169 185 10915871
Lewis LK Resnick MA 2000 Tying up loose ends: Nonhomologous end-joining in Saccharomyces cerevisiae
Mutat Res 451 71 89 10915866
Kaliraman V Mullen JR Fricke WM Bastin-Shanower SA Brill SJ 2001 Functional overlap between Sgs1-Top3 and the Mms4-Mus81 endonuclease Genes Dev 15 2730 2740 11641278
Haber JE Heyer WD 2001 The fuss about Mus81 Cell 107 551 554 11733053
Wang D Lippard SJ 2005 Cellular processing of platinum anticancer drugs Nat Rev Drug Discov 4 307 320 15789122
Bolzan AD Bianchi MS 2002 Genotoxicity of streptozotocin Mutat Res 512 121 134 12464347
Beranek DT 1990 Distribution of methyl and ethyl adducts following alkylation with monofunctional alkylating agents Mutat Res 231 11 30 2195323
Galiegue-Zouitina S Bailleul B Loucheux-Lefebvre MH 1983 In vitro DNA reaction with an ultimate carcinogen model of 4-nitroquinoline-1-oxide: The 4-acetoxyaminoquinoline-1-oxide. Enzymatic degradation of the modified DNA Carcinogenesis 4 249 254 6299600
Friedberg EC Walker GC Siede W 1995 DNA repair and mutagenesis Washington, DC ASM Press
Xiao W Chow BL Hanna M Doetsch PW 2001 Deletion of the MAG1 DNA glycosylase gene suppresses alkylation-induced killing and mutagenesis in yeast cells lacking AP endonucleases Mutat Res 487 137 147 11738940
Chen J Derfler B Maskati A Samson L 1989 Cloning a eukaryotic DNA glycosylase repair gene by the suppression of a DNA repair defect in Escherichia coli
Proc Natl Acad Sci U S A 86 7961 7965 2682633
Hsiang YH Lihou MG Liu LF 1989 Arrest of replication forks by drug-stabilized topoisomerase I–DNA cleavable complexes as a mechanism of cell killing by camptothecin Cancer Res 49 5077 5082 2548710
Balakrishnan R Christie KR Costanzo MC Dolinski K Dwight SS 2005 Fungal BLAST and model organism BLASTP best hits: New comparison resources at the Saccharomyces Genome Database (SGD) Nucleic Acids Res 33 D374 D377 15608219
Shiratori A Shibata T Arisawa M Hanaoka F Murakami Y 1999 Systematic identification, classification, and characterization of the open reading frames which encode novel helicase-related proteins in Saccharomyces cerevisiae by gene disruption and Northern analysis Yeast 15 219 253 10077188
Huh WK Falvo JV Gerke LC Carroll AS Howson RW 2003 Global analysis of protein localization in budding yeast Nature 425 686 691 14562095
Sickmann A Reinders J Wagner Y Joppich C Zahedi R 2003 The proteome of Saccharomyces cerevisiae mitochondria Proc Natl Acad Sci U S A 100 13207 13212 14576278
Enyenihi AH Saunders WS 2003 Large-scale functional genomic analysis of sporulation and meiosis in Saccharomyces cerevisiae
Genetics 163 47 54 12586695
Scholes DT Banerjee M Bowen B Curcio MJ 2001 Multiple regulators of Ty1 transposition in Saccharomyces cerevisiae have conserved roles in genome maintenance Genetics 159 1449 1465 11779788
Wickner RB Koh TJ Crowley JC O'Neil J Kaback DB 1987 Molecular cloning of chromosome I DNA from Saccharomyces cerevisiae : Isolation of the MAK16 gene and analysis of an adjacent gene essential for growth at low temperatures Yeast 3 51 57 3332963
Ashburner M Ball CA Blake JA Botstein D Butler H 2000). Gene Ontology: Tool for the unification of biology. The Gene Ontology Consortium Nat Genet 25 25 29 10802651
Ito T Chiba T Ozawa R Yoshida M Hattori M (2001 A comprehensive two-hybrid analysis to explore the yeast protein interactome Proc Natl Acad Sci U S A 98 4569 4574 11283351
Shor E Weinstein J Rothstein R 2005 A genetic screen for top3 suppressors in Saccharomyces cerevisiae identifies SHU1, SHU2, PSY3, and CSM2—Four genes involved in error-free DNA repair Genetics 169 1275 89 15654096
Schurer KA Rudolph C Ulrich HD Kramer W 2004 Yeast MPH1 gene functions in an error-free DNA damage bypass pathway that requires genes from homologous recombination, but not from postreplicative repair Genetics 166 1673 1686 15126389
Prakash R Krejci L Van Komen S Anke Schurer K Kramer W 2005
Saccharomyces cerevisiae MPH1 gene, required for homologous recombination-mediated mutation avoidance, encodes a 3′ to 5′ DNA helicase J Biol Chem 280 7854 7860 15634678
Hurley LH 2002 DNA and its associated processes as targets for cancer therapy Nat Rev Cancer 2 188 200 11990855
Tusher VG Tibshirani R Chu G 2001 Significance analysis of microarrays applied to the ionizing radiation response Proc Natl Acad Sci U S A 98 5116 5121 11309499
Alberts B 2003 DNA replication and recombination Nature 421 431 435 12540917
Guthrie C Fink GR 1991 Guide to yeast genetics and molecular biology New York Academic Press
Rose MD Winston F Heiter P 1990 Methods in yeast genetics: A laboratory manual Cold Spring Harbor, New York Cold Spring Harbor Laboratory Press
Erdeniz N Mortensen UH Rothstein R 1997 Cloning-free PCR-based allele replacement methods Genome Res 7 1174 1183 9414323
Goldstein AL McCusker JH 1999 Three new dominant drug resistance cassettes for gene disruption in Saccharomyces cerevisiae
Yeast 15 1541 1553 10514571
Gietz RD Schiestl RH Willems AR Woods RA 1995 Studies on the transformation of intact yeast cells by the LiAc/SS-DNA/PEG procedure Yeast 11 355 360 7785336
Chen DC Yang BC Kuo TT 1992 One-step transformation of yeast in stationary phase Curr Genet 21 83 84 1735128
Breitkreutz BJ Stark C Tyers M 2003 The GRID: The General Repository for Interaction Datasets Genome Biol 4 R23 12620108
Breitkreutz BJ Stark C Tyers M 2003 Osprey: A network visualization system Genome Biol 4 R22 12620107
Hardman JG Limbird LE Gilman AG 2001 Goodman and Gilman's The pharmacological basis of therapeutics, 10th edition New York McGraw-Hill
Ruhland A Fleer R Brendel M 1978 Genetic activity of chemicals in yeast: DNA alterations and mutations induced by alkylating anti-cancer agents Mutat Res 58 241 250 370576
Komura J Ikehata H Hosoi Y Riggs AD Ono T 2001 Mapping psoralen cross-links at the nucleotide level in mammalian cells: Suppression of cross-linking at transcription factor- or nucleosome-binding sites Biochemistry 40 4096 4105 11300790
Drablos F Feyzi E Aas PA Vaagbo CB Kavli B 2004 Alkylation damage in DNA and RNA—Repair mechanisms and medical significance DNA Repair (Amst) 3 1389 1407 15380096
Szkudelski T 2001 The mechanism of alloxan and streptozotocin action in B cells of the rat pancreas Physiol Res 50 537 546 11829314
Lu LJ Baxter JR Wang MY Harper BL Tasaka F 1990 Induction of covalent DNA modifications and micronucleated erythrocytes by 4-nitroquinoline 1-oxide in adult and fetal mice Cancer Res 50 6192 6198 2119250
|
16121259
|
PMC1189734
|
CC BY
|
2021-01-05 08:00:24
|
no
|
PLoS Genet. 2005 Aug 19; 1(2):e24
|
utf-8
|
PLoS Genet
| 2,005 |
10.1371/journal.pgen.0010024
|
oa_comm
|
==== Front
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0034
Special Topics in Public Health
Featured Abstracts
Featured Abstracts from the 18th National Conference on Chronic Disease Prevention and Control
4 2004
15 3 2004
1 2 A09
==== Body
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0034a
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedState-Community Partnerships: Eliminating Health Disparities Through Coalition-driven, Asset-based, Community-focused Interventions
Acosta Michael Health Program Administrator New York State Department of Health, Office of Minority Health
Empire State Plaza, Corning Tower, Room 780, Albany, NY 12237 518-474-2180 [email protected]
4 2004
15 3 2004
1 2 A09
The objective of this state-community partnership initiative was to mobilize community action to eliminate health disparities through coalition-driven, asset-based, neighborhood-specific program designs. Participants included 16 African American, Hispanic, Asian American, and other underserved neighborhoods throughout New York State.
From April 2000 through March 2003, the New York State Department of Health funded the Minority Health Community Partnerships initiative, establishing 16 coalitions to address health disparities. Coalitions were funded for 3 years and received training on coalition development and the asset-based community development model. Interventions were designed around the strengths and resources of coalition members and community assets — individuals, associations, and institutions. Disparities addressed included asthma, cardiovascular disease, diabetes, HIV/AIDS, oral health, and access to care. Peer education, provider education, case management, media messages, and community-wide outreach strategies were used. Success was measured based on the extent to which intervention objectives were met, community assets utilized, and coalition members engaged.
A total of 400 organizations (e.g., faith-based, educational, health care, commercial, financial, media) and residents engaged in the process. Three hundred peer educators were trained to deliver prevention messages, and 300 providers were trained on prevention strategies. A total of 60,000 community residents were reached with prevention messages. A total of $2 million in additional funding was leveraged.
A coalition-driven, asset-based approach to addressing health disparities creates opportunities for underserved populations to make collective decisions about community health and to implement strategies that build skill sets and competencies useful to the community.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0034b
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedImplementation of the Coordinated Approach to Child Health (CATCH) Program in Texas
Barroso Cristina MPH Graduate Research Assistant University of Texas-Houston School of Public Health, Center for Health Promotion and Prevention Research
7000 Fannin, 2610 A, Houston, TX 77030 713-500-9603 [email protected]
Hoelscher DM
Kelder SH
McCullum C
Ward JL
Cribb P
Murray N
4 2004
15 3 2004
1 2 A09
Implementation of the Coordinated Approach to Child Health (CATCH) program was evaluated by child nutrition (CN) and physical education (PE) specialists who attended CATCH trainings in Texas during 2001 and 2002.
Coordinated school health programs offer an effective means of providing consistent health promotion messages. CATCH is a school-based nutrition and physical activity program designed to reduce risk factors — unhealthy dietary intake, physical inactivity, and smoking, for example — for chronic diseases among elementary school children.
CN and PE specialists who attended trainings completed a mail survey that assessed factors influencing implementation of CATCH at their schools. A cross-sectional study design was used: response rates were 38.7% for PE specialists and 39.9% for CN specialists.
The mean score for the percentage of PE lessons using CATCH PE activities was 44.7%, and the mean score for the percentage of CN specialists implementing CATCH Eat Smart guidelines was 80.4%. Likert scales were used to score satisfaction with the CATCH program, where 1 = strongly disagree and 5 = strongly agree, and mean scores were calculated. Both PE and CN specialists reported that the CATCH program helped them meet PE and food service goals (mean of 3.8 for each). PE and CN specialists were satisfied with the CATCH program (mean scores were 4.1 for PE specialists and 3.9 for CN specialists). PE and CN specialists would recommend the CATCH program to others (mean of 4.1 for each).
Results indicate that CATCH is being implemented and is viewed positively by both CN and PE specialists in schools in which personnel were trained using a coordinated school health model.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0034c
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedIdentifying Walking and Trail Use Supports and Barriers Through Focus-Group Research
Burroughs Ericka MA, MPH Program Coordinator University of South Carolina, Arnold School of Public Health, Prevention Research Center
730 Devine St, Columbia, SC 29208 803-436-2182 [email protected]
Fields RM
Granner ML
Sharpe PA
4 2004
15 3 2004
1 2 A09
Walking and trail use supports and barriers in a South Carolina county were identified. As part of a community-based participatory research project, focus groups were conducted to develop social marketing activities.
Twelve focus groups were conducted to identify themes related to physical activity, walking, and trail use. Questions covered preferred walking location, social support for physical activity, preferred incentives, and characteristics of trails. Discussions were recorded, transcribed verbatim, and analyzed using NVivo software.
Results from the focus groups revealed that, while concern for safety was the primary walking barrier cited by women and older adults, walking groups were a potential support for them. There were differences in preferences between active and inactive groups and male and female participants. The secluded nature of some trails had positive and negative aspects. Furthermore, awareness of existing trails was low..
This focus-group research indicated that the social marketing intervention in this community should address safety concerns and emphasize walking groups for women and older adults. The lack of awareness of available walking trails indicated a need to publicize trails as part of the intervention.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0034d
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedMoving Right Along: A Creative Partnership to Engage Older Adults in Physical Activity and Nutrition Programs
Chapel Denise MPH, MS, RD Program Coordinator Community Nutrition, Missouri Department of Health and Senior Services
290 Wildwood Avenue, Jefferson City, MO 65102 573-751-6183 [email protected]
McCulla MM
Reinsch B
Warren C
4 2004
15 3 2004
1 2 A09
A creative partnership was established among Tri-Parish Nursing Ministries, The Arthritis Foundation, Missouri Extension Services, and the Saint Louis County Department of Health to promote quality of life for older adults through physical activity and health education messages.
Twenty-nine participants met twice a week in a local church to exercise for one hour with a trainer from PACE, or People with Arthritis Can Exercise, and then to receive 30 minutes of nutrition education from a Saint Louis County registered dietitian using the Missouri Extension's Health for Everybody program.
A pilot study design was used to engage Saint Louis North County older adults in a physical activity program. Participants were recruited from doctors' offices, local churches, grocery stores, and libraries, using flyers and brochures. Newspaper ads were also used. Outcomes desired included a minimum participation rate of 50% at the end of the 6-week sessions, increased flexibility, and positive feelings reported in Health for Everybody evaluation forms.
The participation rate for the program was 62% at the end of 6 weeks. Mean age of participants was 74 years. All participants reported positive feedback, indicating that they enjoyed sessions and "learned something new." Mean Healthy Eating Index score was 68.6. Feedback revealed that participants desired more ways to exercise outside of class, so Tai Chi instruction and pedometers will be incorporated into future programs.
Nutrition education and structured exercise classes in a safe setting, tailored to varying levels of ability, engaged older adults for the 6-week session.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0034e
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedImproving Care for the Homeless Population Using the Chronic Care Model
Choucair Bechara MD Health Care for the Homeless, Medical Director, Crusader Central Clinic Association
1200 W State St, Rockford, IL 611022 815-490-1729 [email protected]
Palmer T
4 2004
15 3 2004
1 2 A09
The objective of this study was to adapt the chronic care model to improve outcomes among the homeless population in the Health Care for the Homeless Program at Crusader Clinic in Rockford, Ill.
A major goal of Healthy People 2010 is to eliminate health disparities. Obtaining baseline data to assess clinical quality is a necessary step toward identifying areas for eliminating health disparities between the homeless population and the general population.
Diabetes, hypertension, and asthma were the chronic conditions chosen for study because of their prevalence among the Rockford, Ill, homeless population. Staff members were divided into 3 teams, each addressing one of the chronic conditions. A registry was established to track the outcome of each condition. The Patient Electronic Care System (PECS), provided by the Health Disparities Collaborative, was used to track outcomes.
The number of hypertension patients with a blood pressure of less than 140/90 mm Hg increased from a baseline of 31% to 48%, despite the addition of newly diagnosed patients. Homeless clients with diabetes had an average HbA1c of 7.9%. Among asthma patients, 25% had a severity assessment documented, and 50% had received an influenza vaccine.
Using the chronic care model and intensive follow-up, improvements in outcomes can be significant, despite many barriers to optimal care among the homeless population.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0034f
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedUtah K-8th Grade Height and Weight Measurement Project
Coats Karen MD Program Coordinator Utah Department of Health, Heart Disease and Stroke Prevention Program
PO Box 142107, Salt Lake City, UT 84114-2107 801-538-6227 [email protected]
Friedrichs MD
Ware JL
4 2004
15 3 2004
1 2 A09
Baseline height and weight data were established on more than 10,000 Utah children in kindergarten through eighth grade.
The prevalence of childhood overweight in the United States has tripled since 1966. In Utah, the Behavioral Risk Factor Surveillance System is used to track adult overweight and obesity, and the Youth Risk Behavior Survey is used to track adolescent overweight. However, no data source exists to track childhood overweight.
Height and weight data were collected on 10,041 students at randomly selected elementary and middle schools in Utah. School nurses assisted with recruiting volunteers, scheduling, and measuring. The Heart Disease and Stroke Prevention Program provided staff, equipment, and measurement protocol. Screens were used to protect children's privacy. Gender, birth date, height, and weight were recorded.
One fourth of Utah students in kindergarten through eighth grade — 28% of boys and 23% of girls — were overweight or at risk of becoming overweight. The percentage of students considered overweight was 12.2% — 14% for boys and 10% for girls. Since 1993, overweight among third-grade boys has increased by 119%; overweight among third-grade girls has increased by 40%.
Utah mirrors the national trend of increasing rates of childhood overweight; therefore, surveillance of kindergarten through eighth-grade students is important. With assistance from school nurses and parent and teacher volunteers, data can be collected in a timely manner. Consistent methodology is recommended, and children's privacy should be considered.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv12_03_0034g
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedBC Walks: The Use of Mass Media and Community Programming to Promote Walking
Fell Patricia RN, MS Director Community Health Services, United Health Services Hospitals, Wilson Medical Center
33-57 Harrison St, Johnson City, NY 13790 607-763-6159 [email protected]
Fisher B
Reger B
Spicer D
4 2004
15 3 2004
1 2 A09
The objective of this study was to increase the number of people who meet the recommended standard for moderate-intensity physical activity through walking and to assess the use of mass media and community programming to increase walking awareness and behaviors in individuals. The goal was to have 10,000 people pledge to start walking 10 minutes a day, working up to 30 minutes a day.
BC Walks media campaign, public relations, and public health activities were conducted in Broome County, NY, during May and June 2003. Broome County's total population is 200,500.
BC Walks promoted walking among the residents of Broome County through a mass media campaign including television, radio, cable, and print. Additional programming included walking events such as "Walk With a Doc" and a walk with a county executive downtown at lunchtime. Preprinted prescription pads for walking were given out by health care providers to their patients. A speakers' bureau was developed to provide speakers for service groups, a program that reached more than 1000 people with the BC Walks message. Additionally, a Web site was developed for individuals to log their minutes and see the latest activities of BC Walks. Worksite challenges using pedometers as incentives were conducted in 30 businesses representing more than 4000 employees. Five schools introduced programs that encouraged walking among students, staff, and parents. Community programs on walkable communities were conducted in 4 local municipalities.
During the 2-month program, 10,800 people signed pledge cards to start walking 30 minutes a day. This mass media and community program demonstrated an increase in the number of people who pledged to meet the recommended standard for moderate physical activity.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv04_03_0034h
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedUsing CDC’s School Health Index to Improve the Physical Activity and Nutrition Environments in 15 Michigan Public Schools
Drzal Nicholas MPH, RD Nutrition Education Consultant Michigan Department of Education, Office of School Excellence
608 W Allegan St, PO Box 30008, Lansing, MI 48933 517-335-1730 [email protected]
Grost L
Coke-Haller E
Murphy A
4 2004
15 3 2004
1 2 A09
The childhood obesity crisis was addressed in Michigan by implementing the Centers for Disease Control and Prevention's SHI: School Health Index in 15 schools during the 2002–2003 school year.
The Michigan departments of education and community health have partnered to encourage schools to improve their physical activity and nutrition environments by using the School Health Index. Fifteen Michigan public schools agreed to implement School Health Index and report their results during the 2002–2003 school year in return for $1000.
Each school was required to form a Coordinated School Health Team (CSHT), designate a team leader (on-site coordinator), and work with a trained School Health Index implementation facilitator. Facilitators assisted the on-site coordinator and CSHT through School Health Index implementation.
Evaluation results indicated that using the School Health Index encouraged schools to create and maintain building-level CSHTs, increased on-site coordinators' familiarity with physical activity and nutrition policies, and encouraged physical activity and nutrition promotion activities. Moreover, this process increased staff and student opportunities to be physically active and, in addition, increased the number of nutrition learning opportunities during school hours. Despite these successes, some schools experienced difficulty maintaining their teams over the year and lacked administrative support.
In summary, the School Health Index is a valuable, free tool for a committed school staff member or public health agency representative to use in mobilizing a school to offer more physical activity and nutrition education opportunities and to serve healthier food.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv04_03_0034i
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedHealthy Weight in Schools: Supporting Schools Through Partnerships
Oleksyk Shannon Carney MS, RD Nutrition Consultant Michigan Department of Community Health, Division of Chronic Disease and Injury Control/Cardiovascular Health, Nutrition & Physical Activity
3423 N MLK Jr Blvd, PO Box 30195, Lansing, MI 48909 517-335-9373 [email protected]
Haller EC
4 2004
15 3 2004
1 2 A09
In 2002, statewide consensus recommendations — The Role of Michigan Schools in Promoting Healthy Weight — were developed under the leadership of the state departments of education and public health. Thirty statewide organizations participated in the consensus process. Many of these original partners and additional new organizations are actively collaborating to identify and implement initiatives to help schools implement the consensus guidelines.
Five specific initiatives have evolved: 1) a development, printing, and distribution plan for family education materials; 2) grants to schools to implement the Centers for Disease Control and Prevention's SHI: School Health Index self-assessment and planning tool and the U.S. Department of Agriculture's Changing the Scene kit; 3) a Web site for easy access to resources and assessment tools; 4) a statewide conference on healthy school environments; and 5) leadership and development of resources for the Michigan Action for Healthy Kids Coalition.
Preliminary evaluation results suggest that schools find these tools useful in creating healthy school environments and promoting healthy weight.
Schools are struggling to address obesity during a time of limited resources. It is imperative that states do everything possible to help them in their efforts. Partnerships reduce duplication and produce innovative products and tools.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv04_03_0034j
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedSHI: School Health Index: Implementing Changes in the Third Edition
Harrykissoon Samantha MPH School Health Education Specialist Centers for Disease Control and Prevention, Research Application Branch/Division of Adolescent and School Health
4770 Buford Hwy NE, Mail Stop K-12, Atlanta, GA 30341 770-488-6128 [email protected]
Wechsler H
4 2004
15 3 2004
1 2 A09
The Centers for Disease Control and Prevention's SHI: School Health Index: A Self-Assessment and Planning Guide is designed to help schools assess and improve their physical activity, healthy eating, tobacco use, and unintentional injury and violence prevention policies and programs in the context of a coordinated school health program.
In addition to reviewing the history, purposes, and structure of the School Health Index, contributors to the guide describe the process of developing the third edition and identify changes that have been made to the new edition, which will be released in 2004. Changes include the addition of items for assessing a school's injury prevention policies and programs and the revision of other items based on feedback from public health practitioners. The third edition also will feature a new interactive on-line version that allows users to tailor the School Health Index according to health and safety topics — physical activity, nutrition, tobacco-use prevention, and injury and violence prevention, for example.
The presenters share information collected from education agencies, health departments, schools, and other programs across the nation about how they used the School Health Index, including descriptions of innovative strategies for promoting its use and increasing its impact.
Contributors to the guide will continue their interactive tradition by encouraging School Health Index users to share their experiences in using or promoting the School Health Index, as well as to suggest ideas for improving the guide or maximizing its effectiveness.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv04_03_0034k
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedPolicy and Environmental Change Strategies to Reduce Obesity: Action Packets
Kaley Lori MS, MSB, RD, LD Coordinator Coordinator, Community Health Initiatives, University of Southern Maine, Edmund S. Muskie School of Public Service, Institute for Public Sector Innovation
295 Water St, Augusta, ME 04330 207-626-5258 [email protected]
Wigand D
Whalen K
Root> A
4 2004
15 3 2004
1 2 A09
Action packets were developed to help communities, schools, and work sites in Maine develop policies and produce changes in the environment that would lead to improved nutrition and increased physical activity.
The Maine Cardiovascular Health Program and the Maine Nutrition Network collaborated to develop policy and environmental change strategies with a focus on increasing physical activity and improving nutrition. Evidence-based findings and the social/ecological model were used in determining strategies. Strategies were presented to state- and community-level stakeholders. Action packets are being used to implement strategies.
A concept and framework development team, action packet workgroups, and potential users collaborated to produce action packets. Two action packets are Promote Trail Development and Use of Safe Community Routes for Walking and Biking and Develop Policies that Support Healthy Eating at Group Events. Two additional action packets in production are Enhance Access to Places for Physical Activity and Develop Policies that Support Health Options in Vending Machines. Action packets include case studies, action steps (including outcome evaluation), advocacy materials, Web resources, and references.
Regional action packet educational sessions were provided. An educational session evaluation report indicated that a majority of participants would incorporate strategies into their action plans. Hard copies of action packets have been distributed and are also available on-line. Ongoing process evaluation is being completed.
Policy and environmental change initiatives consume resources. Providing groups in multiple settings with needed resources in the form of action packets supports implementation of initiatives that address risk factors for obesity and other chronic diseases.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv04_03_0034l
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedHeart-healthy and Stroke-free: Making the Business Case to Employers and Purchasers for Preventing Heart Disease and Stroke
Matson-Koffman Dyann DrPH, MPH Public Health Educator/Behavioral Scientist Centers for Disease Control and Prevention, Cardiovascular Health Branch, Division of Adult and Community Health, National Center for Chronic Disease Prevention and Health Promotion
4770 Buford Hwy NE, Mail Stop K-47, Atlanta, GA 30341-3717 770-488-8002 [email protected]
Anwuri VA
Orenstein D
Shore> K
Agin L
Garfinkel SA
Sokler> LA
Watkins NB
Mensah> GA
4 2004
15 3 2004
1 2 A09
The objective of this study was to highlight 2 Centers for Disease Control and Prevention (CDC) initiatives aimed at educating employers about health benefits and interventions that will have the greatest impact on preventing heart disease and stroke and reducing associated costs.
The CDC conducted a literature review and met with the National Business Group on Health to present effective interventions and promising practices for controlling heart disease and stroke and related risk factors. The CDC is also working with the American Institute of Research to develop a toolkit for states that will contain similar information.
We conducted a literature review using the Internet and ABI/Inform, LexisNexis, Medline, OVID, and PubMed databases. We identified 55 articles for health care and 22 for work site settings.
Findings suggest that the most promising interventions for improving the prevention and control of high blood pressure and high blood cholesterol in health care settings include quality care teams and protocols that follow national treatment and prevention guidelines, the use of physician and patient reminders via automated record systems, and patient education combined with quality improvement goals. In the work site setting, the most promising interventions are individual counseling and follow-up, combined with environmental supports such as health risk appraisals, wellness communications, health education classes, and access to healthy food choices and exercise facilities. On the basis of information from 9 organizations, the return-on-investment estimates ranged from $1.40 to $4.90 in savings per dollar spent for work site health management interventions.
To have the greatest impact on preventing heart disease and stroke, employers should consider individual risk-reduction counseling for high-risk employees within the context of a comprehensive systems-level approach and the most promising environmental health promotion interventions.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv04_03_0034m
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedChronic Disease Risk Factors by Ethnicity and Border Residence in Arizona: 10 Years of Behavioral Risk Factor Surveillance System Data
McGorray Matt MS Senior Research Specialist Southwest Center for Community Health Promotion, Arizona College of Public Health, Epi-Biostats
1145 N Campbell Ave, PO Box 210228, Tucson, AZ 85721-0228 520-318-7270 [email protected]
Veazie MA
4 2004
15 3 2004
1 2 A09
We estimated chronic disease risk factor prevalence by border county residence and Hispanic ethnicity along the U.S.-Mexico border. Hispanics living along the U.S.-Mexico border suffer from high rates of chronic diseases. Ongoing surveillance of chronic disease risk factors in this population is lacking.
We combined 10 years of Arizona Behavioral Risk Factor Surveillance System (BRFSS) data (1992–2001). The prevalence of selected risk factors was calculated for Hispanics and non-Hispanics by border county residence. After adjusting for the effects of survey design, age, and sex, we estimated the interaction between Hispanic ethnicity and residence for each risk factor in counties with cities on the border. We also mapped statewide telephone coverage among Hispanics by census tract.
Of the 20,409 respondents, 3.1%were border Hispanics, 12.3% were non-border Hispanics, 6.6% were border non-Hispanics, and 78.1% were non-border non-Hispanics. When border Hispanics are compared with non-border non-Hispanics, the age- and sex-adjusted odds ratios were 2.85 (95% Confidence Interval [CI] = 1.92, 4.23) for diabetes; 0.33 (CI = 0.25, 0.44) for self-reported good or excellent health; 1.48 (CI = 1.11, 1.98) for obesity; 0.52 (CI = 0.42, 0.64) for recommended physical activity; and 0.48 (CI = 0.38, 0.60) for current smoking.
Arizona Hispanics (represented by BRFSS survey participants) suffer from a higher prevalence of chronic disease risk factors than non-Hispanics, regardless of whether they live on the border. Results document the need to oversample on the border and to address the issue of low telephone coverage.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv04_03_0034n
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedCommunities Combating Chronic Disease: The Kate B. Reynolds SELF Improvement Program
Sauer Maggie MS, MHA Administrative Program Manager Department of Community and Family Medicine, Division of Community Health
DUMC 2914, Durham, NC 27 919-681-3086 [email protected]
Michener JL
Yaggy SD
4 2004
15 3 2004
1 2 A09
The objective of this study is to prevent chronic disease in at-risk populations through the development of community-based service delivery networks using community partnerships to create and sustain behavior change.
Administered by Duke University Medical Center's Department of Community and Family Medicine (CFM), on behalf of the Kate B. Reynolds Charitable Trust, the program provides funding to 16 low-income community partnerships across 21 counties in North Carolina over a 5-year period. Partners include health departments, health care organizations, schools, park and recreation departments, economic development offices, and other civic groups.
The Kate B. Reynolds Smoking, Education, Lifestyle and Fitness (SELF) Improvement Program is focused on 3 main health problems: tobacco use, inadequate nutrition, and physical inactivity. Projects are required to demonstrate effective and long-term change in the most at-risk populations through collaboration with local providers and community-based organizations as a support base for sustainability.
CFM provides technical assistance to grantees on a variety of issues, including project management and evaluation, health promotion, and service delivery.
The SELF Improvement Program is a statewide initiative that enables North Carolina communities to effect positive health changes through community-level interventions.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv04_03_0034o
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedTRAILS, a School-based Walking Program
Moore William PhD Assistant Professor of Research University of Oklahoma Prevention Research Center
800 NE 15th St, Room 532, Oklahoma City, OK 73104 405-271-2330 [email protected]
Wilson T
Stephens A
Eichner J
4 2004
15 3 2004
1 2 A09
The effects of a high school walking program, The Robust American Indian Lifestyle Study (TRAILS), were evaluated by measuring changes in lipid profile, aerobic capacity, and body composition. The walking program was implemented at Anadarko Public High School in southwest Oklahoma. Approximately 55% of the school's students are Native American, 34% are white, and 4% are African American. Ethnically, 7% of the students are Hispanic or Latino.
The intervention consisted of daily (Monday through Friday) self-paced walking for 35 to 40 minutes for 11 weeks. Pre- and post-intervention assessments of non-fasting lipid profile, non-fasting plasma glucose, aerobic capacity, and body composition were performed.
Twenty-five students completed the pre- and post-intervention assessments. The mean age of the students was 16.7 years (± 1.4). Sixty percent of the students were of normal weight, 4% were at risk for overweight, and 36% were overweight. The mean miles walked was 37.6 (± 10.7). Using paired t-tests, statistically significant improvements were seen in total cholesterol (174.8 to 149.4 mg/dL, P < .001); low-density lipoprotein cholesterol (106.4 to 85.1 mg/dL, P < .001); non-high-density lipoprotein cholesterol, calculated by subtracting high-density lipoprotein cholesterol from total cholesterol (128.5 to 105.9 mg/dL, P < .001); and non-fasting plasma glucose (103.2 to 82.4 mg/dL, P < .001).
A school-based walking program may have a positive impact on lipids and non-fasting plasma glucose.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv04_03_0034p
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedLessons Learned From Global Reviews of Mass Media Campaigns Designed to Reduce Smoking and Exposure to Secondhand Smoke
Schar Elizabeth President Healthcare POV
5961 Masters Blvd, Orlando, FL 32819 513-703-5887 [email protected]
Gutierrez KK
4 2004
15 3 2004
1 2 A09
The Centers for Disease Control and Prevention (CDC) is conducting several reviews of mass media campaigns to determine the kinds of campaign elements that contribute most to success. The purpose of the reviews is to aid states and other countries in developing their own campaigns to reduce smoking and exposure to secondhand smoke. The first 2 reviews (of campaigns to promote adult smoking cessation and youth tobacco use prevention) have been completed. Preliminary findings for the third review (of campaigns to reduce exposure to secondhand smoke) are available.
Media campaigns are an effective component of a comprehensive tobacco control program. Programs must determine ways to make their limited funds work most efficiently to change attitudes and behaviors related to smoking and secondhand smoke via media campaigns.
Data and results were solicited through a variety of channels, including CDC networks, GLOBALink, and the World Health Organization. Qualitative and quantitative data as well as published and unpublished data were analyzed to understand both study results and insights into target audiences. Key measurements used to determine campaign effectiveness included changes in awareness, relevant knowledge, attitudes, and behavior.
Each campaign review produced unique findings on the following effective campaign elements: 1) a carefully targeted audience; 2) an effective message or combination of messages; 3) appropriate tone and format, including the use of emotion; 4) publicity and promotion through news media coverage; 5) sufficient media presence (reach, frequency, and duration); 6) thorough evaluation (formative, process, and outcome); and 7) synergy between the campaign and other elements of a comprehensive tobacco control program. Differences exist between the strategies and tactics needed for campaigns focused on individual change and campaigns focused on institutional or policy change.
The complete reviews illustrate key findings with examples of advertisements used in countries around the world.
Although limited by the incomplete and imperfect data collected globally, the findings provide a clear sense of direction to readers planning campaigns to encourage adult smoking cessation and youth tobacco use prevention and to reduce exposure to secondhand smoke.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
15670441
PCDv04_03_0034q
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedCost-effective, Community-based Strategies Targeting Cardiovascular Disease and Diabetes Risk Factors Among African American Women in Faith-based Environments
Taylor Cheryl RN, MN, PhD Principal Investigator National Black Women's Health Imperative, REACH 2010: At the Heart of New Orleans
1515 Poydoas, Suite 1020, New Orleans, LA 70112 504-680-2810 [email protected]
Cole L
Ferdinand D
Arline S
4 2004
15 3 2004
1 2 A09
The objective of this study was to examine the differences in heart-health behaviors of African American women in faith-based environments and the effects of faith-based interventions on behaviors.
Eliminating disparities among racial and ethnic groups and increasing quality of life and years of healthy life are the major goals of REACH 2010 research demonstration projects. African Americans in Louisiana have alarming rates of death and disability due to heart disease and diabetes. Limited data are available on faith-based interventions among African American women.
Forty churches with memberships ranging from 100 to 15,000 each were randomized into 3 groups. The Community Health Assessment Program Survey (CHAPS), similar to the Behavioral Risk Factor Surveillance System, was conducted among 1100 African American women 18 years and older. A profile from each church was used to collect baseline aggregate data. Each group received varied doses of intervention. All churches received free annual blood pressure, body mass index, and lipid screening and counseling; screening resources were provided by community board members. Follow-up survey data were collected to determine the effects of interventions on reducing risk factors. A comparative analysis was conducted of baseline CHAPS data on 1100 women. Annual follow-up surveys are still in progress.
Participants have reported a need for chronic disease interventions that address a continuum of risk reduction, including response to emergency events such as heart attacks and strokes and referrals for rehabilitation resources. Also, the presence of respected community leaders and collaborative relationships has empowered REACH 2010: At the Heart of New Orleans to share resources and expand its reach to communities across the state.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv04_03_0034r
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedUsing Exercise for Risk Reduction in African American Breast Cancer Survivors: A Community-based Pilot Study
Wilson Diane EdD, MS, RD Associate Professor Virginia Commonwealth University School of Medicine, Internal Medicine/Massey Cancer Center
1200 E Broad St, West Hospital W10-402, PO Box 980306, Richmond, VA 23298-0306 804-828-9891 [email protected]
Porter JS
Parker G
Smith TJ
Kilpatrick J
4 2004
15 3 2004
1 2 A09
The objective of this study was to pilot test a low-impact exercise program, Walking Counts!, for its effect on steps walked per day and body mass index (BMI) in a population of African American breast cancer survivors.
The Massey Cancer Center in Richmond, Va, partnered with community centers to offer a walking intervention, designed by the study's primary investigator, to high-risk breast cancer survivors. More than 60% of women report weight gain after breast cancer diagnosis, increasing their risk of cancer recurrence and other co-morbidities. Few studies have tested cognitive/behavioral healthy lifestyle interventions in cancer survivors. This study was designed to measure the impact of Walking Counts! on steps per day, BMI, and related measures by providing skills, knowledge, and self-assessment for African American women who have had breast cancer.
An 8-week intervention was held at community locations for African American breast cancer survivors (n = 23) aged 30 to 70 years. Pedometers, a walking scheduler/tracker, and informational/motivational sessions were provided to participants to help them achieve 10,000 steps per day. Data were collected at 3 points to examine changes in walking; BMI; body fat percentage; waist, hip, and forearm circumferences; attitudes toward exercise; cancer stress; and related demographic measures.
Pre- and post-intervention impact included statistically significant increases in steps per day (P = .001), as well as decreases in BMI (P = .004), body weight (P = .006), percent body fat (P = .002), and waist (P = .035) and forearm (P = .005) circumferences. Increased positive perception of exercising was also reported. Follow-up data, including 3-month post-intervention data, will be presented to identify characteristics related to successful outcomes.
Increasing walking for exercise, without making other changes, can improve attitudes and anthropometric measures, which may help reduce risk of cancer recurrence. The high retention rate (95%), along with positive study outcomes, indicate that breast cancer survivors are motivated to improve their health habits.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv04_03_0034s
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedImproving Diabetes Care With the Collaborative Model: The First North Carolina Diabetes Collaborative
Wolf Marti RN, MPH Co-Director North Carolina Diabetes Collaborative, North Carolina Primary Health Care Association
875 Walnut St, Suite 150, Cary, NC 27511 919-469-5701 [email protected]
Reaves J
Porterfield DS
Carlyle RM
Wang A
4 2004
15 3 2004
1 2 A09
The planning, implementation, and outcomes of the North Carolina Diabetes Collaborative, modeled after the Bureau of Primary Health Care's Health Disparities Collaborative, is described.
The North Carolina Diabetes Collaborative is the result of a partnership between the North Carolina Diabetes Prevention and Control Program and the North Carolina Primary Health Care Association. An advisory council made up of strategic statewide partners complements this partnership. Fourteen teams from various health care settings across North Carolina were recruited to participate in this intervention, which focuses on improving the management of diabetes.
Participants receive technical assistance that includes learning sessions, monthly conference calls, distribution lists, and feedback on monthly reports. Teams address system-level changes and use monthly reports to track improvements in delivery of care and health outcomes. An electronic database helps the teams to identify effective interventions and to track outcomes using 8 required measures.
Improvements have been documented in the 8 required measures. Improvement has been most remarkable in the number of patients who have had their HbA1c levels checked twice per year, the number of patients with blood pressure levels below 135/85 mm Hg, the number of patients receiving foot exams, and those with low-density lipoprotein cholesterol levels below 100 mg/dL. During the first 6 months, the teams entered data on 907 patients into their disease management registries.
The North Carolina Diabetes Collaborative shows promise of increasing the quality of care for patients in participating sites. Next steps include obtaining funding for future collaboratives; expanding content area to include cardiovascular disease; incorporating feedback to improve the collaborative; and expanding the number of participants.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
15634371
PCDv04_03_0034t
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedWhose Choice Is It? Understanding HIV Risk Among African American Women
Yancey Elleen PhD Center Director Morehouse School of Medicine, Prevention Research Center
720 Westview Drive SW, Atlanta, GA 30310 404-752-1511 [email protected]
Goodin LM
Wang M
4 2004
15 3 2004
1 2 A09
The types and prevalence of human immunodeficiency virus (HIV) risk behaviors among African American women ages 17 to 44 years were identified and an intervention was developed to reduce the risk of HIV infection by addressing culture and gender issues specific to these women.
In this intervention, we identified communities with high incidences of HIV infection and acquired immunodeficiency syndrome (AIDS) among African American women.
Before and after the intervention, an HIV Risk Reduction Survey was administered to 422 women to assess risk behavior variables. Focus groups were conducted. An intervention was developed and conducted that consisted of 7 weekly sessions. The intervention used this project's research findings and incorporated the theoretical underpinnings of 2 concepts: Ntu (an Africentric model of spiritual beliefs, practices, culture, and interpersonal relationships) and the Theory of Gender and Power (a social theory about sexual inequities, gender and power, and balances).
Intervention and control group comparisons before and after the intervention indicate a significant increase in HIV knowledge among women in the intervention group, based on the 12-item HIV Knowledge Scale in the Morehouse School of Medicine HIV Reduction in African American Women Survey: Intervention group mean scores pre-intervention vs post-intervention were 8.66 vs 10.01; control group mean scores pre-intervention vs post-intervention were 8.41 vs 8.42 (P = .01). Intention to use condoms increased among women in the intervention group but decreased among women in the control group, based on the 4-item Condom Barrier Beliefs construct (using a Likert scale of 1 to 4) in the Morehouse Survey: Intervention group mean pre-intervention vs post-intervention was 1.64 vs 1.69; control group mean pre-intervention vs post-intervention was 1.64 vs 1.61 (P = .05). Personal risk perceptions increased in both groups (using a 1-item Likert scale of 1 to 5), although less in the intervention group: Intervention group mean pre-intervention vs post-intervention was 1.95 vs 2.01; control group mean pre-intervention vs post-intervention was 1.96 vs 2.33 (P = .05).
Interventions to reduce the risk of HIV infection among African American women should help them understand relationships, facilitate increased knowledge about HIV, and support attitude and behavior changes within the context of their culture and environment. Women in this study showed an interest in seeking information on reducing their risk of HIV infection and possibly initiating steps toward behavior change. A sustained and protracted effort might be needed to help this population move from increased understanding to sustained behavior change.
Prev Chronic Dis
Preventing Chronic Disease
1545-1151
Centers for Disease Control and Prevention
PCDv04_03_0034u
Special Topics in Public Health
Original Research: Featured Abstract from the 18th National Conference on Chronic Disease Prevention and Control
Peer ReviewedThe Spatial Analysis of CVD Mortality in a Tri-county Area of Mississippi
Zhang Lei Business Systems Analyst II Mississippi State Department of Health
570 E Woodrow Wilson Ave, Jackson, MS 39215 601-576-7112 [email protected]
Penman A
Haydel C
Sutton V
Kamali V
Fos P
4 2004
15 3 2004
1 2 A09
The geographic distribution and spatial pattern of cardiovascular disease (CVD) mortality was investigated in Hinds, Rankin, and Madison counties in Mississippi from 1997 to 2000. The analysis of geographic distribution of disease mortality has an important role to play in public health and epidemiological studies.
The 1997–2000 CVD mortality data in Hinds, Rankin, and Madison counties with residential addresses were obtained from Mississippi vital statistics and geocoded to the census block groups using ArcView software. Because of the small number of CVD deaths in each block group, the geocoded records were aggregated to the related census tract. Kernel density estimator was used to calculate annual CVD mortality for each census tract. A geographically weighted regression method was used to analyze the spatial pattern of CVD mortality in the tri-county area.
For most census tracts, the changes in population density did not explain the changes in CVD death density, suggesting that the high rates were real and not an artifact of population change. The changes in CVD mortality over time were not significant for most census tracts.
These results may be useful in suggesting hypotheses for further study related to environmental factors and socioeconomic status.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Acosta MA. State-community partnerships: eliminating health disparities through coalition-driven, asset-based, community-focused interventions [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034a.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Barroso CS, Hoelscher DM, Kelder SH, McCullum C, Ward JL, Cribb P, et al. Implementation of the Coordinated Approach to Child Health (CATCH) program in Texas [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034b.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Burroughs EL, Fields RM, Granner ML, Sharpe PA. Identifying walking and trail use supports and barriers through focus-group research [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034c.htm
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Chapel DL, McCulla MM, Reinsch B, Warren C. Moving right along: a creative partnership to engage older adults in physical activity and nutrition programs [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034d.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Choucair B, Palmer T. Improving care for the homeless population using the chronic care model [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034e.htm
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Coats KA, Friedrichs MD, Ware JL. Utah K-8th grade height and weight measurement project [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034f.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article:Fell PE, Fisher B, Reger B, Spicer D. BC Walks: the use of mass media and community programming to promote walking [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034g.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Grost L, Coke-Haller E, Murphy A, Drzal N. Using CDC's School Health Index to improve the physical activity and nutrition environments in 15 Michigan public schools [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL:http://www.cdc.gov/pcd/issues/2004/apr/03_0034h.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article:Haller EC, Oleksyk SC. Healthy weight in schools: supporting schools through partnerships [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034i.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Haller EC, Oleksyk SC. Healthy weight in schools: supporting schools through partnerships [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034i.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Kaley L, Wigand D, Whalen K, Root A. Policy and environmental change strategies to reduce obesity: action packets [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034k.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Matson-Koffman DM, Anwuri VA, Orenstein D, Shore K, Agin L, Garfinkel SA, et al. Heart-healthy and stroke-free: making the business case to employers and purchasers for preventing heart disease and stroke [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034l.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: McGorray MC, Veazie MA. Chronic disease risk factors by ethnicity and border residence in Arizona: 10 Years of Behavioral Risk Factor Surveillance System data [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034m.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article:: Michener JL, Sauer ML, Yaggy SD. Communities combating chronic disease: the Kate B. Reynolds SELF Improvement Program [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034n.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Moore W, Wilson T, Stephens A, Eichner J. TRAILS, a school-based walking program [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034p.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Schar EH, Gutierrez KK. Lessons learned from global reviews of mass media campaigns designed to reduce smoking and exposure to secondhand smoke [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034p.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Taylor C, Cole L, Ferdinand D, Arline S. Cost-effective, community-based strategies targeting cardiovascular disease and diabetes risk factors among African American women in faith-based environments [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034q.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Wilson DB, Porter JS, Parker G, Smith TJ, Kilpatrick J. Using exercise for risk reduction in African American breast cancer survivors: a community-based pilot study [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034r.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Wolf MS, Reaves J, Porterfield DS, Carlyle RM, Wang A. Improving diabetes care with the collaborative model: the first North Carolina Diabetes Collaborative [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034s.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Yancey EM, Goodin LM, Wang M. Whose choice is it? Understanding HIV risk among African American women [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034t.htm.
The opinions expressed by authors contributing to this journal do not necessarily reflect the opinions of the U.S. Department of Health and Human Services, the Public Health Service, Centers for Disease Control and Prevention, or the authors' affiliated institutions. Use of trade names is for identification only and does not imply endorsement by any of the groups named above.
Suggested citation for this article: Zhang L, Penman A, Haydel C, Sutton V, Kamali V, Fos P. The spatial analysis of CVD mortality in a tri-county area of Mississippi [abstract]. Prev Chronic Dis [serial online] 2004 Apr [date cited]. Available from: URL: http://www.cdc.gov/pcd/issues/2004/apr/03_0034u.htm.
|
15663884
|
PMC1189735
|
CC BY
|
2022-01-27 23:20:34
|
no
|
Prev Chronic Dis. 2004 Mar 15; 1(2):A09
|
utf-8
|
Prev Chronic Dis
| 2,004 |
nan
|
oa_comm
|
==== Front
BMC BiochemBMC Biochemistry1471-2091BioMed Central London 1471-2091-6-131608078810.1186/1471-2091-6-13Research ArticleArchazolid and apicularen: Novel specific V-ATPase inhibitors Huss Markus [email protected] Florenz [email protected] Brigitte [email protected] Rolf [email protected] Heinrich [email protected] Gudrun [email protected] Axel [email protected] Helmut [email protected] Universität Osnabrück, Fachbereich Biologie/Chemie, Abteilung Tierphysiologie, 49069 Osnabrück, Germany2 Gesellschaft für Biotechnologische Forschung, Bereich Naturstoffe, 38124 Braunschweig, Germany3 Universität Göttingen, Fakultät für Chemie, Institut für Organische und Biomolekulare Chemie, 37077 Göttingen, Germany2005 4 8 2005 6 13 13 7 3 2005 4 8 2005 Copyright © 2005 Huss et al; licensee BioMed Central Ltd.2005Huss et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
V-ATPases constitute a ubiquitous family of heteromultimeric, proton translocating proteins. According to their localization in a multitude of eukaryotic membranes, they energize many different transport processes. Since their malfunction is correlated with various diseases in humans, the elucidation of the properties of this enzyme for the development of selective inhibitors and drugs is one of the challenges in V-ATPase research.
Results
Archazolid A and B, two recently discovered cytotoxic macrolactones produced by the myxobacterium Archangium gephyra, and apicularen A and B, two novel benzolactone enamides produced by different species of the myxobacterium Chondromyces, exerted a similar inhibitory efficacy on a wide range of mammalian cell lines as the well established plecomacrolidic type V-ATPase inhibitors concanamycin and bafilomycin. Like the plecomacrolides both new macrolides also prevented the lysosomal acidification in cells and inhibited the V-ATPase purified from the midgut of the tobacco hornworm, Manduca sexta, with IC50 values of 20–60 nM. However, they did not influence the activity of mitochondrial F-ATPase or that of the Na+/K+-ATPase. To define the binding sites of these new inhibitors we used a semi-synthetic radioactively labelled derivative of concanamycin which exclusively binds to the membrane Vo subunit c. Whereas archazolid A prevented, like the plecomacrolides concanamycin A, bafilomycin A1 and B1, labelling of subunit c by the radioactive I-concanolide A, the benzolactone enamide apicularen A did not compete with the plecomacrolide derivative.
Conclusion
The myxobacterial antibiotics archazolid and apicularen are highly efficient and specific novel inhibitors of V-ATPases. While archazolid at least partly shares a common binding site with the plecomacrolides bafilomycin and concanamycin, apicularen adheres to an independent binding site.
==== Body
Background
Vacuolar-type ATPases (V-ATPases) are ubiquitous proton pumps in the endomembrane system of all eukaryotic cells and in plasma membranes of many animal cells where they energize transport processes across the membrane or regulate the pH of corresponding compartments [1]. They are heteromultimeric enzymes consisting of a membrane bound, proton translocating Vo complex and a catalytic V1 complex which is oriented towards the cytosol. In recent years it became more and more evident that malfunction of the V-ATPase is correlated with a multitude of diseases such as osteopetrosis, male infertility or renal acidosis [2-4]. Therefore the V-ATPase turned out to be a subject for biomedical research and even was considered as a potential target for cancer drug therapy [5]. In order to understand the development of these diseases and to design efficient drugs for their therapy it is necessary to gain a most comprehensive knowledge of the mode of action of the enzyme as well as of known V-ATPase inhibitors on the one hand, and, on the other hand, to search for novel potent and specific inhibitors with different inhibition characteristics.
The best examined and established specific V-ATPase inhibitors are the plecomacrolides bafilomycin [6] and concanamycin [7], which both take effect in nanomolar concentrations by binding to the Vo subunit c [8-10]. Recently various new inhibitors of V-ATPases such as the benzolactone enamides [11] or chondropsines [12] have been described (reviewed in [13]) but so far in no case the binding site has been determined. Only for the benzolactone enamide salicylihalamide it was shown that its binding site is different to that of plecomacrolides [10] and may reside somewhere between the Vo and the V1 complex [14]. In the present report we introduce two types of antibiotics produced by myxobacteria, apicularens, new benzolactone enamides [15,16] and archazolids, a novel class of macrolactones [17] which both represent highly potent and specific V-ATPase inhibitors, however, with different modes of action and different binding sites.
Results and Discussion
Archazolid and apicularen influence the viability of mammalian cell-lines
The novel antibiotics archazolids and apicularens (Fig. 1) were checked for their impact on the cell growth of a variety of mammalian cell lines from different tissues (Tab. 1). For nearly all of them IC50 values were in the nanomolar range, comparable with the IC50 values for concanamycin A and bafilomycin A1. Apicularen B was the only exception, with an average IC50 value two orders of magnitude higher. Growth inhibition of the multidrug-resistant cell line KB-V1 was also measured in the presence of verapamil. As this compound inactivates the Pgp efflux pump, a comparison of the IC50 values obtained in the presence and in the absence of verapamil revealed to which extent the compounds were pumped out of the cells by the MDR1 Pgp. The data in Tab 1. show, that unlike the archazolids, the apicularens are poor substrates of Pgp. To visualize the impact of the antibiotics, PtK2 (potoroo kidney) cells were incubated with the inhibitors and stained for intact acidic lysosomes (Fig. 2). Evidently, in the presence of apicularen A and archazolid A as well as in the presence of concanamycin A and of bafilomycin A1 (not shown) the red staining, indicating acidic lysosomes, disappeared compared to cells which had not been treated by drugs. The same observation was made with KB-3-1 cells (data not shown). These results provided the first indication that the novel antibiotics, like concanamycin or bafilomycin, are interfering with the V-ATPase which is the obligatory acidifier of lysosomes. Unlike the cell lines listed in Tab. 1, the growth of PtK2 cells was not completely stopped. The same holds true for A-498 cells (human kidney carcinoma). Both cell lines grow epithelial-like. There was obviously a differential reaction of cell lines to V-ATPase inhibitors. For A-431 cells (human epidermoid carcinoma) a dependence on the expression of the EGF receptor was shown [18].
Table 1 Growth inhibition of different mammalian cell lines by antibiotics
Cell line Origin ArcA ArcB ApiA ApiB BafA1 ConA
IC50 (nM)
L-929 murine connective tissue 0.81a 1.1a 4.5 620 3.2a 0.23a
3Y1 rat, embryogenic fibroblast cell line 0.95 1.1 3.2 310 5.6 1.4
KB-V1b human cervix carcinoma 48.0 35.0 23.0 1600 7.2 28.0
KB-V1b (in presence of 11 μM verapamil) 2.3 1.5 7.9 540 4.0 2.7
A-594 human lung carcinoma 0.54 0.69 0.23c 310c 0.4 0.17
M1 mouse, embryogenic fibroblast cell line 0.27 0.35 1.4 190 2.6 0.56
a [17]; b mdr cell line; c [15]
DSMZ: L-929, KB-V1, A-594
3Y1: [33]
M1: [30]
Figure 1 Structure of the antibiotics. A, I-concanolide A (9-O-[p-(trifluoroethyldiazirinyl)-benzoyl]-21,23-dideoxy-23-[125I]iodo-concanolide A). B, archazolid A and B. C, apicularen A and B.
Figure 2 Inhibition of lysosomal acidification by the novel inhibitors. Potoroo kidney cells (PtK2) were treated with V-ATPase inhibitors for 4 hours and stained for lysosomes (red) with the acidotropic reagent LysoTracker, for mitochondria with MitoTracker (A, green) or for nuclei with Hoechst 33258 (B, blue). The control cells show many red vesicles indicating acidic lysosomes whereas in cells treated with the inhibitors only few red spots can be observed.
The V-ATPase is highly sensitive to archazolid and apicularen
To verify our assumption, we tested the inhibitory efficacy of archazolid A and B as well as of apicularen A and B on the purified V-ATPase holoenzyme from the midgut of the tobacco hornworm. As shown in Fig. 3, both archazolids inhibited the purified enzyme half-maximally at a concentration of ca. 20 nM, equivalent to an IC50 value of ca. 0.8 nmol per mg of protein. Apicularen A and B inhibited the purified enzyme half-maximally at concentrations of ca. 20 nM and 60 nM, respectively, equivalent to IC50 values of ca. 0.8 nmol and 2.4 nmol per mg of protein. These inhibitory values were in the same concentration range as those measured for the plecomacrolides concanamycin A, bafilomycin A1 and B1, and of the benzolactone enamide salicylihalamide which all exhibited a half-maximal inhibition at ca. 10 nM and an IC50 of ca. 0.5 nmol per mg of protein [10]. Thus the novel compounds are highly efficient V-ATPase inhibitors.
Figure 3 Inhibition of the V1/Vo holoenzyme activity by the antibiotics. Values represent the means ± S.D. of three independent enzyme preparations. Archazolid A (open circles), archazolid B (solid circles), apicularen A (open triangles) and apicularen B (solid triangles). The specific enzyme activity of the controls without inhibitors was 1.5 ± 0.2 μmol*mg-1 *min-1.
These findings confirmed our assumption that growth inhibition and the morphological changes induced by archazolid and by apicularen were due to their inhibitory effect on the V-ATPase. Apicularen B was slightly lower active compared to the other drugs, but the difference is much less than expected from the cell culture assays. This difference may result from a lower membrane permeability of apicularen B due to its additional N-acetyl-β-D-glucosamine residue.
While the macrolactones (archazolid A and B) and the benzolactone enamides (apicularen A and B, salicylihalamide) were shown to be potent growth inhibitors of cultured mammalian cells, they all failed to inhibit bacterial growth, while only archazolid showed weak activity against fungi; in addition, the benzolactone enamide salicylihalamide effectively inhibited mammalian V-ATPases but not at all V-ATPases from fungi such as Saccharomyces cerevisiae or Neurospora crassa [11,15,17]. Since the tobacco hornworm V-ATPase is very sensitive to all these antibiotics and available in milligram amounts, it appears to be an appropriate model for further investigations on the mechanism of inhibition by archazolid and the benzolactone enamides.
Archazolid and apicularen do not appear to inhibit F-and P-ATPases
Having shown the high sensitivity of the V-ATPase to the two novel antibiotics, we also wanted to estimate their inhibitory effect on F-and P-type ATPases. In various cell cultures growth was inhibited at nanomolar concentrations (see above). Therefore we asked, whether a concentration of 1 μM which is fairly enough to knock out V-ATPase activity and which is clearly higher than the IC50 values in the growth experiments, would also affect F-ATPases such as the mitochondrial ATP-synthase or P-ATPases such as the plasma membrane Na+/K+-ATPase. Thus, we prepared mitochondria rich crude membranes from mouse heart and submitochondrial particles from beef heart as well as highly purified, Na+/K+-ATPase containing plasma membranes from pig kidney, and tested the inhibitory potential of archazolid and of apicularen. To detect F-ATPase activity we used its specific inhibitors, azide or oligomycin [19], and ouabain or vanadate were used as specific inhibitors of the Na+/K+-ATPase [20,21]. Whereas the ATPase activity in the mouse and beef heart preparations was reduced to values around 25% by the specific F-ATPase inhibitors azide or oligomycin, the new antibiotics archazolid and apicularen as well as the established specific V-ATPase inhibitors concanamycin and bafilomycin reduced the activity only slightly to approximately 80% (Fig. 4). An increase of the inhibitor concentration to 10 μM for archazolid and apicularen had no substantial effect. As shown in Tab. 2, Na+/K+-ATPase activity was completely inhibited by 1 mM vanadate and by 1 mM ouabain, whereas it was only slightly affected by apicularen A and by archazolid A. A slight inhibition of comparable size was also observed for the established inhibitor of V-ATPases, concanamycin A, corroborating the well known fact that plecomacrolidic antibiotics start to inhibit P-ATPases at concentrations in the micromolar range [6,7]. Taken together our results with F-and P-ATPases strongly suggest that archazolid and apicularen are specific V-ATPase inhibitors.
Table 2 Inhibition of Na+/K+-ATPase from pig kidney by antibiotics
specific activity μmol*mg-1 *min-1
experiment 1 experiment 2
control without inhibitors 19.0 15.9
vanadate 0.1 mM 2.0 n. d.
vanadate 1 mM 0 0
ouabain 0.1 mM 5.2 n. d.
ouabain 1 mM 0 0
concanamycin A 1 μM 16.7 12.7
apicularen A 1 μM 17.3 13.2
archazolid A 1 μM 17.1 12.2
Figure 4 Inhibition of F-ATPase activity in crude membranes from mouse heart and submitochondrial particles from beef heart. Values represent the means ± S. D. of three (mouse heart) or four (beef heart) independent experiments., respectively. The specific ATPase activity without inhibitor was 0.16 ± 0.05 μmol*mg-1 *min-1 in mouse heart crude membranes (blue columns) and 6.1 ± 0.35 μmol*mg-1 *min-1 in submitochondrial particles of beef heart (red columns). Solitary columns indicate that the conditions were only tested either for mouse or for beef heart.
Archazolid and apicularen bind to different parts of the V-ATPase
To investigate the mode of V-ATPase inhibition in more detail, we used the semi synthetic derivative of concanamycin, 9-O-[p-(trifluoroethyldiazirinyl)benzoyl]-21,23-dideoxy-23-[125I]iodo-concanolide A (I-concanolide A, Fig. 1) which already had been used successfully to identify the binding site of plecomacrolides in V-ATPases unambiguously [10]. In our previous study we had shown that the Vo subunit c is the only subunit of the V-ATPase which is labelled by I-concanolide A and that this labelling could be prevented by pre-incubation with the plecomacrolides concanamycin A, bafilomycin A1 and B1. In a comparable UV-radiation experiment labelling of subunit c by I-concanolide A was completely abolished by pre-incubation with archazolid A (Fig. 5), thus indicating that plecomacrolides and archazolid A share, at least partially, the same binding site in subunit c.
Figure 5 Protection of the binding site for I-concanolide A by plecomacrolidic antibiotics. Tricine-SDS-PAGE gels. Lane 1, stained with Coomassie Blue; lane 2–8, autoradiography of the gel after exposition to a phosphoscreen. Samples of 20 μg V-ATPase were preincubated for 60 min on ice with 100 μM or 10 μM bafilomycin B1 (lanes 3 and 4), 100 μM or 10 μM apicularen A (lanes 5 and 6) and 100 μM or 10 μM archazolid (lanes 7 and 8), respectively. I-concanolide A was then added to give a final concentration of 10 μM. The mixture was incubated for another 60 min on ice and then treated with UV light. Control with pre-incubation, but without effectors (lane 2).
Unlike archazolid, the benzolactone enamide apicularen A did not appear to interfere with labelling of subunit c by the concanolide A derivative because the signal obtained was almost as strong as in the control (Fig. 4). Therefore we suggest that apicularen binds to a site which is largely different from that for the plecomacrolides. This result is in agreement with our former observation that the benzolactone enamide salicylihalamide does not bind to subunit c and supports our conclusion that the sites and mechanisms of inhibition for benzolactone enamides are different from those for plecomacrolides [10].
Conclusion
The novel antibiotics archazolid and apicularen are highly efficient and specific novel inhibitors of V-ATPases. Despite the different structures of archazolid and the plecomacrolides they probably have a similar mode of inhibition and their binding sites in the V-ATPase have at least a considerable overlap. In contrast, apicularen which demonstrates the same inhibition efficacy does not interfere with I-concanolide A, thus suggesting a mode of inhibition different from that of the plecomacrolides.
Methods
Enzyme preparations
The V-ATPase holoenzyme was purified as published elsewhere [10]. Preparation of highly purified membranes containing Na+/K+-ATPase from pig kidney followed the protocol of Jørgensen [22] with the three main steps of differential centrifugation, incubation with SDS in the presence of ATP and sucrose density gradient centrifugation in a fixed angle rotor, and led to a specific enzyme activity which was in the same range as that reported by the authors. The resulting sample was stored frozen at -20°C. To prepare crude membrane extracts, the hearts of three mice were washed three times with an ice-cold buffer of pH 8.1 consisting of 0.25 M sucrose, 5 mM Tris-HCl, 5 mM EDTA, and 10 mM Pefabloc SC (Biomol), homogenized in 5 ml of this buffer and centrifuged at 4°C for 25 min at 233,000 × gmax in a fixed angle rotor. The resulting pellet was resuspended in 5 ml of a buffer of pH 7.5, consisting of 5 mM Tris-MOPS (3-morpholinopropanesulfonic acid), 10 mM NaCl, 10 mM Pefabloc SC, 9.6 mM 2-mercaptoethanol, 0.53 mM EGTA and 0.1 % Triton X-100. After an aliquot for protein determination had been taken, 10% bovine serum albumin (final concentration) was added to the suspension which was then stored on ice until the activity assays were run. Beef heart mitochondria were isolated by differential centrifugation, following the protocol of Smith by using a blender to homogenize the heart mince [23]. The initial homogenization buffer consisted of 250 mM saccharose, 10 mM KH2PO4, 10 mM Tris, 2 mM EGTA, 2 mM MgCl2, pH 7.4, and further isolation procedures were carried out in the same medium without EGTA [24]. Submitochondrial particles were obtained by ultrasonic treatment of the mitochondria.
Antibiotics
Labelled I-concanolide A was synthesized as described elsewhere [25]. Archazolid A and B, apicularen A and B, concanamycin A and bafilomycin A1 and B1 were isolated according to published procedures [15-17,26]. To avoid freeze-thaw cycles which have significant influence on the stability of the substances, aliquots of stock solutions in dimethyl sulfoxide were stored at -70°C and thawed only once immediately before use. The actual concentrations of the stock solutions were determined spectrophotometrically.
ATPase assays
Standard V-ATPase assays with a final volume of 160 μl and a pH of 8.1 consisted of 3–4 μg of protein, 50 mM Tris-MOPS, 3 mM 2-mercaptoethanol, 1 mM MgCl2, 20 mM KCl, 0.003% C12E10, 20 mM NaCl, and 3 mM Tris-HCl. After 5 min of pre-incubation at 30°C with or without inhibitors, 1 mM Tris-ATP was added and after incubation for 2 min the reaction was stopped by placing the tube in liquid nitrogen. Assays using Na+/K+-ATPase were performed in 160 μl at pH 7.5 and contained 0.5 μg of protein, 50 mM Tris-MOPS, 5 mM imidazole, 0.2 mM EDTA, 4 mM MgCl2, 20 mM KCl and 100 mM NaCl. After 5 min of pre-incubation at 37°C with or without inhibitors, 3 mM Tris-ATP was added and after 1 min of incubation the reaction was stopped by placing the tube in liquid nitrogen. Assays using crude membranes from mouse heart had a volume of 160 μl and a pH of 7.5, and consisted of 10 μg of membrane protein, 50 mM Tris-MOPS, 3 mM 2-mercaptoethanol, 1 mM MgCl2, 20 mM KCl, 100 mM NaCl, 0.02% Triton X-100 and 0.3 mg/ml BSA. After 5 min of pre-incubation at 30°C with or without inhibitors, 1 mM Tris-ATP was added and after an incubation time of 10 min the reaction was stopped by placing the tubes in liquid nitrogen. ATPase assays with submitochondrial particles from beef heart were performed in a final volume of 1 ml and a pH of 8.0. The samples contained 9–12 μg protein, 50 mM Tris, 50 mM KCl and 2.5 mM MgCl2. After 5 min of preincubation with or without inhibitors, 5 mM ATP was added, and after an additional incubation time of 15 min the reaction was stopped by the addition of 0.4 ml of 20 % TCA.
Inorganic phosphate produced in the assays of V-ATPase, Na+/K+-ATPase and mouse heart F-ATPase was measured according the protocol of Wieczorek et al. [27], while the determination of inorganic phosphate produced in the assays of beef heart F-ATPase followed the method of Fiske and Subarrow [28] using ascorbic acid as reducing agent.
Labelling
Twenty micrograms of the samples were pre-incubated with the inhibitors for 1 h on ice. The labelled I-concanolide A was then added to give a final concentration of 10 μM. Controls were run without effectors. Mixtures with volumes of 40 μl were incubated for an additional 1 h on ice and then treated for 3 min with UV light (366 nm) on ice. After UV irradiation, 10 μl of 5-fold sample buffer [10] was added, the mixture was heated for 45 s at 95°C, cooled on ice, and subjected to Tricine-SDS-PAGE (16.5% T, 3% C separating gel and 10%T, 3% C spacer gel [29]), followed by Coomassie staining. The gels were sealed in plastic wrap before they were exposed to a phosphoscreen for up to 72 h and analyzed with the aid of a phosphoimager (Molecular Dynamics).
Cell culture and growth inhibition assay
Cell lines were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ). Cell lines 3Y1 and M1 (embryonal fibroblasts from 129/Ola × C57BL/6 mice, [30]) were a generous gift from Dr. S. Miyamoto, Fukoka, Japan, and Dr. P. F. Mühlradt, Braunschweig, Germany. All cell lines were cultivated under conditions recommended by the supplier. Growth inhibition was measured in microtiterplates. Aliquots of 120 μl of the suspended cells (50,000/ml) were given to 60 μl of a serial dilution of the inhibitor. After 5 days, growth was determined using the MTT assay [31].
Cell staining
PtK2 (ATCC CCL-56) or KB-3-1 cells were grown on glass coverslips (13 mm diameter) in four-well-plates. Exponentially growing cells were incubated with the inhibitors for 4 hours and stained for lysosomes with 50 nM LysoTracker Red DND-99 and for mitochondria with 75 nM MitoTracker Green FM (both from Molecular Probes) at 37°C for 30 min. The nuclei were stained using Hoechst 33258 (5 μg/ml). The coverslips were mounted upside down in PBS, fixed with nail polish, and observed under the fluorescent microscope.
Other procedures
Fifth instar larvae of M. sexta (Lepidoptera, Sphingidae), weighing 6–8 g, were reared under long day conditions (16 h of light) at 27°C using a synthetic diet modified according to Bell et al [32].
Authors' contributions
MH purified the V-ATPase and the Na+/K+-ATPase, carried out the enzyme and labelling assays, and drafted the manuscript. FS did the fermentation of the archazolids and carried out the cell culture studies. BK did the fermentation of the apicularens and tested F-ATPase activity in beef heart mitochondria. RJ isolated apicularens. HS isolated archazolids. GI and AZ purified the plecomacrolides and synthesized the I-concanolide A. HW participated in the conception and design of the study and drafted the manuscript. All authors read and approved the final manuscript.
Acknowledgements
We like to thank Nina Dankers, Martin Dransmann, Petra Haunhorst, Birte Engelhardt and Bettina Hinkelmann for their excellent technical assistance. This work was supported by grants from the Deutsche Forschungsgemeinschaft (SFB 431: H.W.) and the Fonds der Chemischen Industrie (A.Z.).
==== Refs
Wieczorek H Brown D Grinstein S Ehrenfeld J Harvey WR Animal plasma membrane energization by proton motive V-ATPases Bioessays 1999 21 637 648 10440860 10.1002/(SICI)1521-1878(199908)21:8<637::AID-BIES3>3.0.CO;2-W
Kornak U Schulz A Friedrich W Uhlhaas S Kremens B Voit T Hasan C Bode U Jentsch TJ Kubisch C Mutations in the a3 subunit of the vacuolar H+-ATPase cause infantile malignant osteopetrosis Human Molecular Genetics 2000 9 2059 2063 10942435 10.1093/hmg/9.13.2059
Brown D Smith PJ Breton S Role of V-ATPase-rich cells in acidification of the male reproductive tract J Exp Biol 1997 200 257 262 9050233
Karet FE Finberg KE Nelson RD Nayir A Mocan H Sanjad SA Rodriguez-Soriano J Santos F Cremers C Di Pietro A Hoffbrand BI Winiarski J Bakkaloglu A Ozen S Dusunsel R Goodyer P Hulton SA Wu DK Skvorak AB Morton CC Cunningham MJ Jha V Lifton RP Mutations in the gene encoding B1 subunit of H+-ATPase cause renal tubular acidosis with sensorineural deafness Nature Genetics 1999 21 84 90 9916796 10.1038/5022
Torigoe T Izumi H Ise T Murakami T Uramoto H Ishiguchi H Yoshida Y Tanabe M Nomoto M Kohno K Vacuolar H+ -ATPase: functional mechanisms and potential as a target for cancer chemotherapy Anticancer Drugs 2002 13 237 243 11984067 10.1097/00001813-200203000-00005
Bowman EJ Siebers A Altendorf K Bafilomycins: A class of inhibitors of membrane ATPases from microorganisms, animal cells, and plant cells Proc Natl Acad Sci USA 1988 85 7972 7976 2973058
Dröse S Bindseil KU Bowman EJ Siebers A Zeeck A Altendorf K Inhibitory effect of modified bafilomycins and concanamycins on P- and V-type adenosinetriphosphatases Biochemistry 1993 32 3902 3906 8385991 10.1021/bi00066a008
Bowman BJ Bowman EJ Mutations in subunit c of the vacuolar ATPase confer resistance to bafilomycin and identify a conserved antibiotic binding site J Biol Chem 2002 277 3965 3972 11724795 10.1074/jbc.M109756200
Bowman EJ Graham LA Stevens TH Bowman BJ The bafilomycin/concanamycin binding site in subunit c of the V-ATPases from Neurospora crassa and Saccharomyces cerevisiae J Biol Chem 2004 279 33131 33138 15180988 10.1074/jbc.M404638200
Huss M Ingenhorst G König S Gassel M Dröse S Zeeck A Altendorf K Wieczorek H Concanamycin A, the specific inhibitor of V-ATPases, binds to the V(o) subunit c J Biol Chem 2002 277 40544 40548 12186879 10.1074/jbc.M207345200
Boyd MR Farina C Belfiore P Gagliardi S Kim JW Hayakawa Y Beutler JA McKee TC Bowman BJ Bowman EJ Discovery of a novel antitumor benzolactone enamide class that selectively inhibits mammalian vacuolar-type (H+)-atpases J Pharmacol Exp Ther 2001 297 114 120 11259534
Bowman EJ Gustafson KR Bowman BJ Boyd MR Identification of a new chondropsin class of antitumor compound that selectively inhibits V-ATPases J Biol Chem 2003 278 44147 44152 12944415 10.1074/jbc.M306595200
Beutler JA McKee TC Novel marine and microbial natural product inhibitors of vacuolar ATPase Curr Med Chem 2003 10 787 796 12678782 10.2174/0929867033457827
Xie XS Padron D Liao X Wang J Roth MG De Brabander JK Salicylihalamide A inhibits the V0 sector of the V-ATPase through a mechanism distinct from bafilomycin A1 J Biol Chem 2004 279 19755 19763 14998996 10.1074/jbc.M313796200
Kunze B Jansen R Sasse F Höfle G Reichenbach H Apicularens A and B, new cytostatic macrolides from Chondromyces species (myxobacteria): production, physico-chemical and biological properties J Antibiot (Tokyo) 1998 51 1075 1080 10048565
Jansen R Kunze B Reichenbach H Höfle G Apicularen A and B, cytotoxic 10-membered lactones with a novel mechanism of action from Chondromyces species (Myxobacteria): Isolation, structure elucidation, and biosynthesis European Journal Of Organic Chemistry 2000 913 919 10.1002/(SICI)1099-0690(200003)2000:6<913::AID-EJOC913>3.3.CO;2-C
Sasse F Steinmetz H Höfle G Reichenbach H Archazolids, new cytotoxic macrolactones from Archangium gephyra (Myxobacteria). Production, isolation, physico-chemical and biological properties J Antibiot (Tokyo) 2003 56 520 525 12931860
Yoshimoto Y Imoto M Induction of EGF-dependent apoptosis by vacuolar-type H(+)-ATPase inhibitors in A431 cells overexpressing the EGF receptor Exp Cell Res 2002 279 118 127 12213220 10.1006/excr.2002.5588
Ebel RE Lardy HA Stimulation of rat liver mitochondrial adenosine triphosphatase by anions J Biol Chem 1975 250 191 196 124729
Robinson JD Flashner MS The (Na+/K+)-activated ATPase. Enzymatic and transport properties Biochim Biophys Acta 1979 549 145 176 224922
Macara IG Vanadium - An element in search of a role Trends In Biochemical Sciences 1980 5 92 94 10.1016/0968-0004(80)90256-X
Jørgensen PL Purification of Na+,K+-ATPase: enzyme sources, preparative problems, and preparation from mammalian kidney Methods Enzymol 1988 156 29 43 2835612
Smith AL Preparation, properties, and conditions for assay of mitochondria: Slaughterhouse material, small-scale Methods Enzymol 1967 10 81 86
Thierbach G Reichenbach H Myxothiazol, a new inhibitor of the cytochrome b-c1 segment of the respiratory chain Biochim Biophys Acta 1981 638 282 289 6274398
Ingenhorst G Bindseil KU Boddien C Dröse S Gassel M Altendorf K Zeeck A Synthesis of a doubly labelled concanamycin derivative for ATPase binding studies Eur J Org Chem 2001 23 4525 4532 10.1002/1099-0690(200112)2001:23<4525::AID-EJOC4525>3.3.CO;2-J
Dröse S Boddien C Gassel M Ingenhorst G Zeeck A Altendorf K Semisynthetic derivatives of concanamycin A and C, as inhibitors of V- and P-type ATPases: structure-activity investigations and developments of photoaffinity probes Biochemistry 2001 40 2816 2825 11258892 10.1021/bi001759q
Wieczorek H Cioffi M Klein U Harvey WR Schweikl H Wolfersberger MG Isolation of goblet cell apical membrane from tobacco hornworm midgut and purification of its vacuolar-type ATPase Methods Enzymol 1990 192 608 616 2150092
Fiske CH Subbarow Y The colorimetric determination of phosphorus J Biol Chem 1925 66 375 400
Schägger H von Jagow G Tricine-sodium dodecyl sulfate-polyacrylamide gel electrophoresis for the separation of proteins in the range from 1 to 100 kDa Anal Biochem 1987 166 368 379 2449095 10.1016/0003-2697(87)90587-2
Morr M Takeuchi O Akira S Simon MM Muhlradt PF Differential recognition of structural details of bacterial lipopeptides by toll-like receptors Eur J Immunol 2002 32 3337 3347 12432564
Mosmann T Rapid colorimetric assay for cellular growth and survival - Application to proliferation and cyto-toxicity assays J Immunol Methods 1983 65 55 63 6606682 10.1016/0022-1759(83)90303-4
Bell RA Borg TK Ittycheriah PI Neurosecretory cells in the frontal ganglion of the tobacco hornworm, Manduca sexta J Insect Physiol 1974 20 669 678 4833353 10.1016/0022-1910(74)90189-9
Nakagawa H Miyamoto S Actin-filaments localize on the sorting endosomes of 3Y1 fibroblastic cells Cell Struct Funct 1998 23 283 290 9872569
|
16080788
|
PMC1190152
|
CC BY
|
2021-01-04 16:26:24
|
no
|
BMC Biochem. 2005 Aug 4; 6:13
|
utf-8
|
BMC Biochem
| 2,005 |
10.1186/1471-2091-6-13
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1621598750410.1186/1471-2105-6-162Methodology ArticleTheme discovery from gene lists for identification and viewing of multiple functional groups Pehkonen Petri [email protected] Garry [email protected]örönen Petri [email protected] Department of Neurobiology, A.I. Virtanen-Institute, University of Kuopio P.O. Box 1627, FIN-70211 Kuopio, Finland2 Department of Computer Science, University of Kuopio P.O. Box 1627, FIN-70211 Kuopio, Finland3 Bioinformatics Group, Institute of Biotechnology, P.O. Box 56, 00014 University of Helsinki, Finland2005 29 6 2005 6 162 162 18 3 2005 29 6 2005 Copyright © 2005 Pehkonen et al; licensee BioMed Central Ltd.2005Pehkonen et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
High throughput methods of the genome era produce vast amounts of data in the form of gene lists. These lists are large and difficult to interpret without advanced computational or bioinformatic tools. Most existing methods analyse a gene list as a single entity although it is comprised of multiple gene groups associated with separate biological functions. Therefore it is imperative to define and visualize gene groups with unique functionality within gene lists.
Results
In order to analyse the functional heterogeneity within a gene list, we have developed a method that clusters genes to groups with homogenous functionalities. The method uses Non-negative Matrix Factorization (NMF) to create several clustering results with varying numbers of clusters. The obtained clustering results are combined into a simple graphical presentation showing the functional groups over-represented in the analyzed gene list. We demonstrate its performance on two data sets and show results that improve upon existing methods. The comparison also shows that our method creates a more simplified view that aids in discovery of biological themes within the list and discards less informative classes from the results.
Conclusion
The presented method and associated software are useful for the identification and interpretation of biological functions associated with gene lists and are especially useful for the analysis of large lists.
==== Body
Background
Recent developments in biosciences have created a dramatic change from the analysis of a few genes to large gene lists. These lists are usually selected at the genomic level by criteria such as activity in a stress treatment [1], importance to cell survival in a specific growth condition [2], or as a result of clustering genes by expression profiles [3]. As current high throughput methods produce a vast amount of data as gene lists, the subsequent analysis tends to be a bottleneck due the size of the data set and the high probability of false positive genes among the lists.
One solution to analyse a gene list is to draw information either from the existing literature or from the databases representing whole genome [4,5] or proteome annotations [6,7], and then using these to guide the analysis. Most of these databases simplify the analysis by classifying genes to the biological categories or classes that present their function, localization, or partnership in some protein complex. A further step is to estimate the statistical significance of associations between the classes and genes of the obtained list. Several applications have been recently reported for such analysis [8,9]. Most of these applications compare the frequency of gene classes in the user supplied gene list, obtained by various criteria, to the remaining genes that did not fulfill the criteria. The latter often includes the rest of the genes from the whole genome. The usual outcome from these methods is a sorted list of biological classes considered important. These methods have been beneficial to data analysis by guiding the process towards the most important features in the gene list [10-13]. In addition, the observation of multiple genes from the same functional class increases confidence in results obtained from high throughput methods.
While these methods are useful, several weaknesses are associated with this approach. A gene list can have a heterogeneous structure with multiple dissimilar gene groups such as stress response, a specific metabolic pathway, and protein degradation. The basic statistics used by the previously mentioned methods are often insufficient to reveal this kind of heterogeneity from the associated functional classes. Rather, they have a tendency to be biased toward the gene sub-group associated with the most over-represented functional classes within the analyzed list of genes. This overwhelms many important, but less over-represented, classes that are associated with the rest of the genes in the list. Therefore, it could be hypothesized that there exists other interesting biological functions among the genes that are not members of the best scoring classes. As such, the existing methods do not address this question and thus there is a need for an approach that would concentrate on the possible heterogeneity in the gene list. In the current work, we propose the clustering of a gene list for finding gene groups that differ in functional class annotations.
Results
Principle of the method
Our method takes, as input, the user given gene list chosen by some selection criteria. The selected list is referred to as a sample gene list, and the gene list that did not meet the criteria is referred to as a reference gene list. The aim is then clustering the sample gene list for finding gene groups with different functional class annotations. The clustering is solely based on the gene associations with functional classes obtained from Gene Ontology (GO) database [14], and the measurements like gene expression level or sequence similarity are not used. As a clustering method, we use Non-negative Matrix Factorization (NMF) [15] to create a k-means like partition. The well known weakness with this type of clustering approach is the requirement to select the number of clusters and the initialization for the algorithm. We circumvent this weakness by using a non-nested hierarchical clustering scheme, which allows parallel visualization of several different clustering results. Here, a gene list is repeatedly divided into a growing number of clusters by clustering from random starting initializations. The different clustering results are presented in consecutive levels ordered with the number of clusters, with the first level presenting the gene list without any clustering. Strongly correlating clusters between the consecutive levels are connected by edges forming a non-nested hierarchy (see figures 1, 2, 3). The output graph highlights the clusters that stay similar through the different clustering levels despite the varying number of divisions and different random starting initializations. The resulting visualization can be used either for obtaining suitable grouping for a gene list, or identifying individual clusters that are of interest.
Figure 1 Graphical results from the analysis of H2O2 dataset. The figure shows the non-nested hierarchical clustering tree obtained from GENERATOR with the H2O2 dataset. Each layer presents one clustering solution and each box a single cluster. Boxes show the two best scoring functional classes and the colour of the box corresponds to the over-representation of the best scoring functional class. Best correlating clusters between the consecutive clustering layers are connected with lines. A thicker line indicates a stronger correlation. The correlation value is indicated beside each line. The lines between the first and second level (marked with asterisks) do not present any value as the correlation measure is not defined here. Section A presents a view where two functional classes that contributed most to the cluster formation are shown for each cluster. Section B shows more informative visualization, the default view of GENERATOR, where two classes that were most over-represented in both the original sample list and in the cluster in question are shown. Note the conserved clusters across the different clustering results. We have marked them with Roman numerals.
Figure 2 Replications of non-nested hierarchical clustering tree with H2O2 dataset. The figure presents the four replications for the non-nested hierarchical clustering graph for H2O2 dataset. We have marked the conserved gene clusters with the same Roman numerals as in figure 1. Notice that most clusters (especially I, II and III) can be observed over several levels in each cluster tree.
Figure 3 Graphical results from the analysis of itraconanzole dataset. The figure shows the non-nested hierarchical clustering tree obtained from GENERATOR with the itraconanzole dataset. Section A shows the tree with functional classes that contributed most to the formation of each cluster. Section B shows the default view of GENERATOR with the highest over-represented functional classes in the original list and in the cluster in question. The details of the presentation are explained in text for figure 1. Also in this figure we highlight some conserved clusters with roman numbers.
In the non-nested clustering hierarchy, the cluster contents are described with the most representative functional classes. For this, a combination of three different measures was used to show over-represented classes within each cluster. The measures are positive/negative signed ten based logarithmic transforms [10] of p-values calculated with Fisher's test [16,17], which compares class frequencies between two sets of genes. The first measure, "Original log(p)" (denoted by O.log(p)), makes a comparison between the whole user given sample and reference gene lists. It reports class over-representation that was observed before any clustering. Because of the wide usage of this measure reported in the literature [10,11,18], it is suitable for method comparison. As a comparison, the second measure, "Sample log(p)" (denoted by S.log(p)), concentrates fully on clustered sample gene list by comparing a single cluster against the other genes in the sample list. It highlights the classes that contributed most to the formation of the cluster. The third measure, "Complete log(p)" (denoted by C.log(p)), compares a single cluster against the other genes of the sample gene list and reference. It takes into account both the contribution to the formation of a cluster and the over-representation in the sample list before clustering, and thus we use it for reporting the contents of a cluster. C.log(p) is partly dependent on the preceding clustering, and thus can report some classes that are not over-represented in the whole user given sample gene list, which we are aiming to analyze. Therefore, such hits are filtered by excluding the classes with weak O.log(p) from the report. Similarly, classes that have not contributed to the formation of the analyzed cluster are removed by discarding the classes that do not show even slight over-representation with S.log(p). As a result of filtering, the remaining classes are over-represented in both the analyzed cluster and in the original sample list. In this description, only O.log(p) gives statistically analyzable results because C.log(p) and S.log(p) are both based on the same data with the preceding clustering. Nevertheless, the latter two are suitable for highlighting the classes that are over-represented within the clusters. A more detailed description of the non-nested clustering scheme is given in the Methods section.
Software implementation
In order to make the method applicable for others, we have developed an end-user program called GENERATOR (GENElist Research Aimed Theme-discovery executOR) for the Windows 2000/XP environments. It takes, as input, the sample and reference lists of genes that can be comprised of gene names or identifiers supported by GO database. The list of available species and allowable naming systems are described more in GENERATOR user manual [21] and in GO web site [22]. Alternatively, GENERATOR can be used to analyze existing binary data matrices like in-house created functional gene classifications or other similar binary data analysis problems consisting of sample and reference groups. The first outcome from the program is a non-nested hierarchical clustering tree, which shows the discovered gene sub-groups from the user given sample gene list. The content of each cluster is described by the two most over-represented classes. A more detailed analysis is also possible for each cluster by viewing the sorted list of over-represented classes or by viewing the clustered genes. The program can create multiple cluster trees, produce statistical evaluations for clustering divisions and single clusters, and provide flexibility in changing the parameters for clustering execution and visualizations. Results can be saved as graph figures and tab-delimited files describing different gene groups or class contents within them. These functions are further described in the program manual. GENERATOR will be updated twice in a year including the GO database within it and is freely available [21].
Analysis with GENERATOR
Gene list from yeast under H2O2 stress
We have analyzed the data obtained from growing yeast deletion strains during oxidative stress [2]. Yeast deletion strains have deletions in genes not needed in normal growth conditions (non-essential genes). The research aims to find new genes and functionalities that are important for the cells to survive and grow in the presence of oxidative stress. We limit the analysis to the gene list obtained from hydrogen peroxide stress (H2O2 stress). This was used as a sample list for GENERATOR and it included 117 genes of which 109 were recognized by the GO database. The remaining 4589 non-essential yeast genes were used as a reference list of which 4115 were recognized by the database. The use of a whole genome as a reference here might cause some error in the results as it is natural to assume that different functional groups have different proportions of non-essential genes. The principal observation when analyzing the results as one group in the original article is the clear association with mitochondrion [2].
Clustering was done with 2 to 6 groups. In the first step the obtained clusters were analyzed against the other clusters using S.log(p) values to determine which functional classes contributed most to the formation of each cluster. The obtained graphical view is shown in figure 1A. The figure shows a cluster of ribosome genes that forms the clearest separate group (marked with I) and remains although the number of clusters changes from 2 to 6. The strong link between the different clustering results (thick lines showing correlations higher than 0.9) highlights this. Similarly, a cluster of genes with RNA associated function (marked with II) is clearly separated and is shown on several levels. Also, a small cluster of 'mitochondrial inner membrane' genes (marked with III), a cluster of genes with unknown function (marked with V), and a cluster associated with 'transcription regulation' and 'nucleus' (marked with IV) can be seen. All of these five clusters stay similar over many levels of the visualization despite the changing number of clusters and different random starting points. The whole cluster tree step was also replicated four times, each showing similar results. These replications are detailed below.
The previous information obtained by S.log(p) explains the clustering, but it does little to help understand the original sample list. This is due to the exclusion of the reference list from the analysis. For example, the previous results do not provide emphasis on mitochondrial functions although it is the most significant theme when analyzing the data as one group (see table 2). Figure 1A also presents 'molecular function unknown' class, although it is under-represented in the original sample list. Therefore, the second step of the analysis is to take the reference gene list into account. Here, classes are sorted with C.log(p) values and O.log(p) and S.log(p) are used as cut-offs to remove non-relevant information. The rationale of using the cut-offs and the purpose of the different values is discussed more in the Methods (Description of the cluster contents). This is also the default view of GENERATOR. The resulting graph is presented in figure 1B. Now the obtained view is different showing 'mitochondrial ribosome' cluster (I, previous ribosome cluster), 'tRNA ligase' cluster (II, previous RNA associated cluster), 'mitochondrial inner membrane' cluster (III) and 'transcription regulation' cluster (IV, previous transcription and nucleus cluster) and a 'mitochondrial genome maintenance' cluster (V, previous cluster of unknown genes). The clusters are the same as the ones shown in the figure 1A but now each one of the clusters shows the functional classes, over-represented in the original sample list, that are associated with the clustered genes. The over-represented classes for clustering with 5 clusters from figure 1 are shown in table 1. In order to see how robust the results are, the non-nested hierarchical clustering was replicated four times. The replications are in figure 2 and show that similar clusters can be obtained with each.
Table 1 Results from GENERATOR with H2O2 dataset using five clusters. The table shows the reported classes for five clusters from GENERATOR clustering shown in figure 1. The three over-representation values described in Methods section and figure 5 are reported for each class. Results are also compared to graphical output from SGD GO term finder (figure 8 [see Additional file 3]). Abbreviations in this column are: MF, molecular function; CC, cellular component; BP, biological process. Classes reported as 'missing' were not observed in the SGD GO term finder graphs. A more detailed view of the data is available in table 5 [see Additional file 5]. Classes with S.log(p) < 1 (marked with -) are not included to analysis although they are still shown here.
CLUSTER FUNCTIONAL CLASS C.log(p) O.log(p) S.log(p) SGD
I organellar ribosome 46.7 26.1 21 CC
mitochondrial ribosome 46.7 26.1 21 CC
mitochondrial matrix 41.6 30.5 14.9 CC
Ribosome 36.9 14.5 23.7 CC
structural constituent of ribosome 35.9 14.8 22.3 MF
II RNA ligase activity 13.3 6.68 6.64 MF
tRNA ligase activity 13.3 6.68 6.64 MF
ligase activity, forming aminoacyl-tRNA and related compounds 13.3 6.68 6.64 MF
ligase activity, forming carbon-oxygen bonds 13.3 6.68 6.64 MF
ligase activity, forming phosphoric ester bonds 13 6.42 6.64 MF
III mitochondrial membrane 13.9 3.66 11 CC
inner membrane 13.1 3.99 9.6 CC
mitochondrial inner membrane 13.1 3.99 9.6 CC
Mitochondrion 8.91 40.1 1.03 -
respiratory chain complex III 8.03 4.32 3.74 CC
IV transcription regulator activity 12.3 2.52 11.5 MF
nucleobase, nucleoside, nucleotide and nucleic acid metabolism 8.99 2.15 7.51 BP
mediator complex 8.97 5.15 3.88 Missing
general RNA polymerase II transcription factor activity 7.94 4.16 3.88 MF
DNA-directed RNA polymerase II, holoenzyme 7.81 4.04 3.88 Missing
V mitochondrial genome maintenance 7.22 5.97 2.02 Missing
mitochondrion organization and biogenesis 6.74 4.06 3.05 Missing
mitochondrial chromosome 4.23 3.18 1.05 - (CC)
soluble fraction 3.1 2.52 1.09 -
helicase activity 3 3.35 0.79 -
Table 2 Comparison of sorted class list against GENERATOR clustering with H2O2 dataset. The table shows the sorted list of over-represented functional classes for H2O2 dataset. Only the 25 best scoring classes are shown to limit the size. Columns indicate the obtained log-p-values (O.log(p)), class names and its rank in the list, and the corresponding GENERATOR cluster number, if the class was included into the obtained GENERATOR result. Notice that the most of the functional classes are associated with mitochondrial ribosome proteins. Detailed results are shown in table 7 [see Additional file 7].
Rank Class name O.log(p) In cluster
1 Mitochondrion 40.06 first level
2 mitochondrial matrix 30.45 II
3 organellar ribosome 26.09 II
4 mitochondrial ribosome 26.09 II
5 protein biosynthesis 21.69 II
6 organellar large ribosomal subunit 18.61 II
7 mitochondrial large ribosomal subunit 18.61 II
8 macromolecule biosynthesis 15.6 not reported
9 structural constituent of ribosome 14.84 II
10 Ribosome 14.53 II
11 Biosynthesis 13.15 not reported
12 structural molecule activity 12.87 not reported
13 ribonucleoprotein complex 12.52 II
14 protein metabolism 12.34 not reported
15 Metabolism 11.75 not reported
16 large ribosomal subunit 11.33 not reported
17 organellar small ribosomal subunit 7.94 not reported
18 mitochondrial small ribosomal subunit 7.94 not reported
19 aerobic respiration 7.3 IV
20 cellular respiration 7.03 not reported
21 Cytoplasm 7.02 not reported
22 RNA ligase activity 6.68 III
23 tRNA ligase activity 6.68 III
24 ligase activity, forming aminoacyl-tRNA and related compounds 6.68 III
25 ligase activity, forming carbon-oxygen bonds 6.68 III
Analysis of the results in figures 1 and 2 (result summary shown in tables 1 and 5 [see Additional file 5]) shows that within the group of genes that first seem homogeneous, there are sub-groups differently associated with the mitochondrial functionality. The strongest feature in the obtained results is the group of mitochondrial ribosome clusters that stays similar whether clustering from 2 to 6 clusters. Analysis of this cluster actually reveals that there are two genes (YNR036C and YPL183W-A) that are reported as hypothetical mitochondrial ribosome proteins. The fact that the mitochondrial ribosome proteins are strongly over-represented in the dataset support the notion that they are mitochondrial ribosomal proteins.
One small group, not mentioned in the original analysis [2], is the group of tRNA ligases (cluster II). Although this group only includes 6 members, its O.log(p) was 6.64 making the over-representation significant. A more detailed analysis reveals that the genes in question are mitochondrion associated tRNA ligases and one of them is a hypothetical mitochondrial tRNA ligase. Again its importance for the growth of yeast cells in oxidative stress further confirms its association with mitochondrial function. The rank of these ligase associated categories starts at 23 in the sorted class list for the original sample gene list (see table 2) and therefore this group can go easily unnoticed if the sample list is analyzed without clustering. The rest of the cluster II (in fig. 1, when using five clusters) includes proteins that link to RNA processing and to translation, for example, NAM1, two mitochondrial elongation factors, and YDR194C.
Cluster V shows 'mitochondrion organization' and 'genome maintenance' (5 and 7 genes, with O.log(p) 4.06 and 5.97) but the analysis of the cluster content shows no clear common theme. Instead, most of the genes have no known function, and therefore this cluster does not seem to contribute to the analysis. Indeed the unknown function was associated to this cluster in figure 1A. A separate cluster of unknown genes is an expected behavior for our method as these genes have highly different GO classification profiles from the known genes. We have also observed it regularly with other datasets. Still, this cluster was able to highlight the small group of genes associated with mitochondrion genome maintenance.
Cluster IV shows nucleus-associated functionalities ('transcription regulator activity', 'regulator complex', 'general RNA polymerase II transcription factor activity'). When the actual cluster content is analyzed, the cluster includes: RNA polymerase II holoenzymes, transcription factors, and transcription regulators. This cluster of genes was unexpected and seems to show a link from nucleus driven functionalities to mitochondrial functionalities. Clusters II and IV show nicely mitochondrion linked functions elsewhere in the cell, but at the same time these groups are harder to detect when analyzing the data as one group (see tables 2 and 7 [see Additional file 7]). In summary, GENERATOR has shown that within the mitochondrion associated gene list, the main members are mitochondrion ribosomal proteins, mitochondrion membrane genes, tRNA ligases, unknown genes, and genes associated with transcription regulation.
Gene list from drug treated yeast
Another dataset that was analyzed includes the gene expression differences in yeast during itraconanzole treatment, a drug known to affect sterol biosynthesis and normal growth [20]. Both up and down regulated genes were used for the analysis. These contained 255 genes of which 248 were recognized by the GENERATOR GO database. The remaining 6102 non-regulated yeast genes constituted the reference list of which 5369 were recognized by our database. When the obtained gene list is analyzed normally with the sorted class list, the most significant feature observed is the massive over-representation of the 'aminoacid biosynthesis' and related functional classes (table 4).
Table 4 Comparison of sorted class list against GENERATOR clustering with itraconanzole dataset. The table shows the sorted list of over-represented functional classes for itraconanzole dataset. Only the 25 best scoring classes are shown to limit the size. Columns are same as in table 2. Notice that most of the functional classes are associated with amino acid biosynthesis. Detailed results are shown in table 8 [see Additional file 8].
Rank Class O.log(p) In cluster
1 amino acid biosynthesis 19.07 I
2 amine biosynthesis 18.08 I
3 carboxylic acid metabolism 14.77 I
4 organic acid metabolism 14.77 I
5 amino acid metabolism 14.70 I
6 amino acid and derivative metabolism 13.75 I
7 amine metabolism 13.02 I
8 arginine biosynthesis 11.28 I
9 steroid metabolism 8.89 II
10 nitrogen metabolism 8.79 not reported
11 urea cycle intermediate metabolism 8.66 not reported
12 arginine metabolism 8.66 not reported
13 Biosynthesis 8.52 I, II
14 transaminase activity 8.32 not reported
15 transferase activity, transferring nitrogenous groups 8.32 not reported
16 glutamine family amino acid biosynthesis 8.16 not reported
17 sterol metabolism 8.02 II
18 glutamine family amino acid metabolism 7.99 I
19 sterol biosynthesis 7.77 II
20 steroid biosynthesis 7.69 II
21 ergosterol biosynthesis 6.22 II
22 ergosterol metabolism 6.22 II
23 aromatic compound metabolism 5.85 not reported
24 branched chain family amino acid metabolism 5.61 not reported
25 cyclohydrolase activity 5.50 not reported
Similar to the previous analysis, two steps were used and the classes that contributed most to the clustering were monitored first. The results show the 'carboxylic acid biosynthesis' associated cluster (marked with I) a cluster associated with 'cellular process' class (III); a 'macromolecule biosynthesis' associated cluster (II); and a cluster associated with unknown functionality (IV). With a larger number of clusters, 'nucleobase metabolism' and 'transcription' associated cluster (V) can be seen.
When the clustering view is changed to show the over-representation reported by C.log(p) (figure 3B), the previous clusters obtain different annotations (result summary shown in tables 3 and 6 [see Additional file 6]). This analysis step was again repeated four times to see how similar the results remained (figure 6 [see Additional file 1]). Cluster I, that showed in fig. 3A carboxylic acid biosynthesis, is now associated with amino acid and carboxylic acid biosynthesis. It forms the most stable cluster and it is seen regularly on several clustering levels also in the replications. Cluster II (macromolecule biosynthesis) is now associated with steroid biosynthesis. Genes in the cluster represent sterol biosynthesis associated functions and other macromolecule biosynthesis functions (for example synthesis of phospholipids). Steroid synthesis is a known target of the drug and that it is now nicely separated from other functionalities that are likely more secondary responses to the drug. Third, a regularly seen cluster is one enriching the plasma membrane and cell wall associated functionalities (III). The genes in this cluster show many membrane associated functions, like transporting activities. Unexpectedly, another cluster, associated with cell wall (cluster IV) can be regularly observed. A detailed analysis of these clusters still reveals that they are different. Cluster III is associated strongly with 'plasma membrane' and 'cell wall'. The other cell wall cluster (cluster IV) is more connected to unknown cellular component than to cell wall and the connection to cell wall is also very weak. Even a slight raise of the cut-off for S.log(p) would filter this link. A more detailed analysis of the cluster IV reveals that 55 out of the 65 genes in the cluster have biological process unknown. Moreover, molecular function is unknown for 58 of these genes. Therefore this cluster does not contribute to the analysis of the gene list. Cluster V does not seem as stable as the earlier clusters. Still, it is observed in most of the replications (figure 6 [see Additional file 1]). It groups together genes associated with nucleobase metabolism and transcription. Detailed analysis shows transcription factors associated with regulation of transcription from the Pol II promoter. Among these genes, some of them are reported to be important for drug resistance (YLR266C, YCR106W) and to stress response (YFL031W, YMR037C) and to two associated with copper uptake (YGL166W, YMR021C). In summary, we observed with GENERATOR an amino acid biosynthesis associated group, steroid and lipid biosynthesis associated group, a group of unknown genes, and genes associated to membrane and transport.
Table 3 Results from GENERATOR with itraconanzole dataset using five clusters. The table presents the reported classes from GENERATOR clustering with six clusters shown in figure 3. Columns and abbreviations are the same as in table 1. A more detailed view is presented in table 6 [see Additional file 6]. We omit the classes with S.log(p) smaller than 1 from the comparison with SGD (-). One outlier cluster is also omitted (shown in more detailed view).
CLUSTER FUNCTIONAL CLASS C.log(p) O.log(p) S.log(p) SGD
I amino acid metabolism 51.48 14.7 38.32 BP
carboxylic acid metabolism 50.64 14.77 36.64 BP
organic acid metabolism 50.64 14.77 36.64 BP
amino acid and derivative metabolism 50.36 13.75 38.32 BP
amino acid biosynthesis 49.98 19.07 31.6 BP
II steroid metabolism 19.83 8.89 11.36 BP
lipid biosynthesis 17.88 5.24 14.07 BP
lipid metabolism 17.87 5.05 13.81 BP
sterol metabolism 17.3 8.02 9.61 BP
steroid biosynthesis 16.95 7.69 9.61 BP
III plasma membrane 9.08 2.58 7.5 Missing
cell wall (sensu Fungi) 4.18 4.37 1.43 CC
cell wall 4.18 4.37 1.43 CC
external encapsulating structure 4.18 4.37 1.43 CC
structural constituent of cell wall 3.97 2.23 1.8 Missing
IV cell wall (sensu Fungi) 2.4 4.37 0.54 -
cell wall 2.4 4.37 0.54 -
external encapsulating structure 2.4 4.37 0.54 -
acid phosphatase activity 1.25 3.1 0.22 -
V specific RNA polymerase II transcription factor activity 10.18 3.24 7.48 MF
nucleobase metabolism 3.26 2.85 1.52 Missing
purine base metabolism 2.56 2.58 1.09 Missing
aromatic compound metabolism 2.54 5.85 0.72 -
heterocycle metabolism 2.22 2.65 1.02 Missing
Comparison with competing methods
Sorted class list
GENERATOR was also compared to existing methods. One of the simplest ways of analyzing a gene list is to take it as one single group, analyze how over-represented different classes are, and to report the results as a sorted list. Sorting is based on the p-values calculated for the observed over-representation in order to show the best results at the top of the list. This method does not take into consideration the heterogeneity in the list, but otherwise it is similar to analysis done with each of the GENERATOR clusters. Actually, the first level of the GENERATOR cluster tree graph does this analysis. Therefore we compared GENERATOR clustering to the sorted class list using the results from the first level. We changed the default settings so that the number of reported functional classes was not limited.
The comparison used the two previously analyzed data sets. The results from sorted class list were compared to GENERATOR clustering summaries shown in tables 1 and 3. When the number of classes was limited only by the p-value, an immediately observed drawback of the sorted list method was the amount of information (number of classes) obtained. For the H2O2 dataset, we obtained 75 classes and for itraconanzole 76 with -log(p-value) > 2 (55 and 43 with -log(p-value) > 3). The resulting sorted lists are shown in tables 7 [see Additional file 7] and 8 [see Additional file 8]. This can be corrected by raising the cut-off for the included genes. This is also reasonable as we have not used here any correction for increased risk of false positives due to multiple testing. Strong filtering with p-values or limiting the number of reported classes leaves the most over-represented functional classes. In the example datasets, the most over-represented functional classes were all associated with the same gene group. With H2O2, the first 18 functional classes were associated with mitochondrial ribosome proteins (see table 2). With itraconanzole, the first 19 classes (except classes 9 and 17) show functions associated with amino acid biosynthesis (see table 4).
When the GENERATOR results were compared to a sorted class list, many classes were omitted from the results. With default settings, GENERATOR shows at maximum ten classes for each cluster in the output text file. This filters out the repetitive occurrences of functional classes associated with the same gene group. In the H2O2 dataset, classes like macromolecule biosynthesis, protein metabolism, and large ribosomal subunit were excluded in this way. This seems acceptable as many similar classes are shown in the results by cluster I. The omitted classes can be still viewed with the sorted list available for each cluster. Another group of classes that are not reported by GENERATOR with H2O2 were very broad classes, such as intracellular, cell, or physiological process. These contribute very little information to the analysis. Similar observations were also seen with the itraconanzole dataset, where many amino acid biosynthesis associated classes were excluded from GENERATOR clustering results. As an exception, itraconanzole showed some broad classes in the results (plasma membrane, cell wall).
Direct acyclic graph
Another way to analyze the obtained gene list is to map the over-represented functional classes into a tree like structure that is behind the GO classes and visualize the results as a graph structure. The benefit to the sorted list presentation is that the hierarchical structures are now visible, highlighting the over-represented functional classes occurring repetitively in the same part of the GO graph. Also, if there are different branches showing over-represented functional classes in the GO structure, they are clearly separated. The major drawback is the large size of the obtained visualization. The graph obtained from AMIGO server [23] using the whole list of over-represented classes from H2O2 dataset was simply too large for analysis (figure 7 [see Additional file 2]). Instead we selected a graphical output from GO term finder at Saccharomyces Genome Database [8] for comparison. The GO term finder adds color coding to show which of the classes showed strongest over-representation. It also tries to make the obtained graph smaller by discarding some branches. As the graph for each ontology is obtained separately, we combined the obtained three graphs to the same picture for a better view. We used GENERATOR clustering summaries shown in tables 1 and 3 for comparison.
In order to compare the obtained GO graphs with the GENERATOR results, we flagged each class that was reported significant if it was included in the GENERATOR result table (figures 8 [see Additional file 3] and 9 [see Additional file 4]). We first observed, in the comparison, that the graphs obtained from SGD GO term finder are still large for analysis. Also, the important features are scattered over three graphs, in comparison to the single table from GENERATOR. It was observed that some classes in the H2O2 data were not shown in the SGD GO graph even though their log-p-value results were highly significant (tables 1 and 3, classes marked as 'missing'). Some of these classes were: aerobic respiration (O.log(p) 7.3), cellular respiration (7.03), and mitochondrial genome maintenance (5.97). This might be an artifact caused by the limited size of the GO graph. SGD graph, on the other hand, showed classes that were not reported by GENERATOR. These classes were the same classes discussed when comparing GENERATOR with the sorted lists. Some of the differences between the results might be explained by the usage of binomial test for calculating significance of the functional classes in GO term finder. It should be noted that the Fisher's exact test used by GENERATOR is a more correct method [8], although we observed similar p-values with both methods. Also the whole genome is always used as a population by GO term finder, which might also cause bias in the results with some datasets (see analysis of H2O2 dataset above).
Comparison to GOToolBox
During the preparation of this manuscript, we also observed another method that performs similar GO clustering. GO-Proxy in GOToolBox [19], takes the user given sample gene list, creates the GO classifications for each gene and clusters the obtained matrix by using czekanowski-dice distance and hierarchical clustering. The reported clusters (called classes) are selected from the different levels of tree with two parameters, defined by the user. One parameter defines how similar genes have to be inside the cluster and the other defines the minimum size for the cluster. The principal difference between the methods is that GENERATOR (with default parameters) reports only the GO-classes that display over-representation in both the original sample gene list and in the obtained cluster, whereas GOToolBox concentrates its analysis to the obtained cluster. Also, GENERATOR gives an overview of the clustered data with visualization.
In the analysis for H2O2 and itraconanzole datasets, GOToolBox, with default parameters, created more and smaller clusters when compared to GENERATOR (tables 9 [see Additional file 9] and 10 [see Additional file 10] for results with each ontology). The cluster number is probably larger because the same clusters with minor changes are selected from different levels of the hierarchical clustering tree which causes repetition in the results. The small clusters in GOToolBox results tend to give a scattered view of the data but could be also useful when analyzing details from the obtained gene list. However, by setting a larger minimum cluster size they can be filtered. With larger clusters GOToolBox reported nonspecific functional classes like cellular process, cell, or metabolism in addition to the same GO-classes that were previously reported by GENERATOR (mitochondrial ribosome classes, tRNA classes etc.). With the default settings, GOToolBox found also some small clusters that were not reported by GENERATOR (clusters associated with 'abiotic stress', 'RNA metabolism' etc.). These clusters were quite small and the most associated functional classes did not show any over-representation in the original sample list (see table 7 [see Additional file 7]) as GOToolBox does not filter the results with O.log(p). GENERATOR could be also run with a larger maximum cluster number in order to obtain similar smaller clusters.
Discussion
We have presented a method that groups a user provided gene list into functionally dissimilar gene clusters. The grouping is done with varying numbers of clusters, which are used to create a tree-like graphic visualization. Despite the emphasis on clustering, our method also analyzes the gene list as a single entity (result with one cluster). The obtained graph presents the main output of the method showing the most important simultaneous gene groups that occur in the data in a single figure. The graph can be created multiple times to see how stable it remains when different random initializations are used for clustering. Our results from clustering replications show that the most visible gene groups remain, thus increasing confidence in the method.
There are two alternative methods previously used to obtain an overview of the over-represented functional categories. Methods like EASE analyze the gene list as one entity and output the functional categories as a sorted list according to the significance of the over-representation. Other methods, like SGD GO term finder, give the over-represented functional categories as a directed tree-like graph by using the hierarchical structure of GO. Graph methods create a much more complex representation with the danger of overwhelming the user with unimportant details. The sorted list gives an impression of a homogenous gene group. As an example, we showed the results from SGD GO term finder, AMIGO visualization, and the sorted list of functional classes for the gene list as one entity. These methods do not group the gene list before analyzing it. A positive unexpected observation was that results from the other methods seemed more informative after we marked them with the corresponding GENERATOR clusters. For example classes in a sorted list can be marked according to which cluster they belong to (see tables 2 and 4). Marking the corresponding clusters enables the opportunity to combine GENERATOR clustering results and results from other methods.
We also compared the GENERATOR results to another gene clustering tool, GOToolBox. The principal difference in methods is that GENERATOR provides the cluster description by using filtering procedure which discards the GO-classes with no over-representation in the original sample gene list and with weak association to the genes of the cluster. GENERATOR includes also visualization for viewing the optional clustering results. Despite the differences we were able to obtain also similar GO-classes with both methods when analyzing the H2O2 and itraconanzole datasets.
Since partitive clustering has an inherent weakness in the initialization, we present a novel solution. Instead of selecting a single clustering number, we monitor the results with a range of clustering numbers. As a result, we obtain correlations between the clusters that highlight those features that can be obtained even though the cluster number would change. The replication of the whole cluster tree visualization was done in order to further highlight those features that are conserved. It should be noted that these ideas could also be used with other clustering applications. Similar work was done by Heger and Holm [24] by replicating NMF many times and looking for the conserved features in the obtained matrix factorizations and by Brunet et al [25] where optimal cluster number was selected by replicating NMF clustering many times.
We analyzed the obtained clusters by concentrating on those functional categories that were over-represented in the cluster when compared to the rest of the gene list and also in the original list of genes when compared to a reference list of genes. If the over-representation in the cluster only would be monitored, the obtained cluster would be well explained, but the drawback would be that the obtained categories could at worst be such that they were under-represented in the original gene list and therefore produce erroneous conclusions. If the over-representation in the original list would be only monitored, the clustering would not be informative to the analysis. The current way of combining these two over-representations highlights those features that are common between the original list and the obtained cluster. As the data is grouped to separate clusters, each of them will represent different features from the list of over-represented functional classes for the original gene list. The reporting method therefore separates those functional categories from the original gene list that are not associated to the same genes and groups together those functional categories from the original gene list that are connected to the same genes. A good example of genes that were associated to the same function were the members of the same protein complex that were often seen as a separate cluster.
The selection of the reported functional categories requires the definition of the cut-off for the significant over-representation. Here the threshold was purposely selected to be liberal (p-value < 0.01, O.log(p) > 2.0). This is known to be too weak a threshold when the analysis includes multiple testing as it increases the possibility of the false positives. Therefore the emphasis was placed in the later analysis on those functional categories that showed clearly stronger over-representation than what the cut-off was and the p-values larger than 0.001 were monitored with caution. Similarly we also discarded classes with S.log(p) < 1 from our analysis. The P-value borders could be selected more precisely by doing repetitive testing with a similar sized sample list with randomly selected members (permutation analysis). The evaluation of the results using runs with randomized samples from the analyzed data is one of the planned additions to the GENERATOR software.
The associated software uses a reference list to calculate over-representation for the original cluster. Although the whole genome for the organism could be used, the reference list will ensure that the biases towards some functional groups in the test situation do not affect the analysis.
The method demonstrates that a drugs primary target can be identified within a separate group among different regulated genes and different cellular functions. Work shown here was done with yeast allowing the use of detailed annotation of the yeast genome. Still, we have also obtained encouraging results from human cell line and C. elegans gene expression datasets (manuscript in preparation). As more information is being gathered from the gene functions, this method should be able to perform even better. Nonetheless, accuracy in the used gene annotations is the weak link for our method. This should not necessarily be a hindrance, as the randomly classified genes should distribute randomly also among the observations. Another limitation is the recognition of the analyzed genes. Gene identifiers can be problematic when working with different naming systems that originate from various databases or high throughput methods, like gene chips. These are also the problems faced by other methods.
The presented software includes the possibility of using it also with binary matrices. The reference group can be given as a binary matrix or as a vector that represents a number of members of each category and also the size of the reference group. This should enable the analysis of other similar binary data sets, like SNP datasets, word occurrences in abstract texts etc. These are being currently tested as future applications.
Conclusion
We have presented an analysis method and associated software, GENERATOR, for analysis of large gene lists. Our aim has been to fulfill the need for an analysis tool to separate and identify functional gene groups from gene lists that would otherwise be difficult to find. The method should be useful especially as larger and more complex gene lists are produced due to the increased use of high throughput genomic methods.
Methods
Data representation
The associations between genes and functional classes in the sample and reference gene lists must be represented as a binary matrix to enable the analysis (see figure 4, steps A and a). As functional classes, we use annotations from the April 2004 delivery of Gene Ontology (GO) database [14]. GO includes three principal sub-hierarchies, representing biological processes, cellular components and molecular functions for a gene. We combine the information from all these three hierarchies in the clustering process.
Figure 4 Flow diagram of the method. The gene associations with the GO functional classes in the sample and reference gene lists are transformed into binary matrices (A and a) and a sum vector (b). The sample set is clustered with NMF based method (B) into a varying number of sub-groups producing a non-nested hierarchical tree (C). Contents of the clusters are described with the over-represented classes within them (c and D).
The gene and functional class associations are transformed into binary matrix where rows represent genes and columns represent classes. Association between gene and class is denoted by one and lack of association with zero in a matrix cell. In addition to directly associated classes, a gene is also denoted to associate with its ancestors in the hierarchical GO structure to assure maximal information for analysis. The obtained matrix for sample gene list is inputted for the clustering process whereas the matrix for the reference list is summed into an occurrence vector (figure 4, step b) which is used later for analyzing the over-represented classes within the obtained clusters.
Clustering technique for binary data
A binary matrix is used as data when clustering the user given sample gene list into a fixed number of groups (see figure 4, step B). Many traditional clustering methods obtain weak results with such data due to its non-continuous nature (see for example [26,27]) and the small proportion of non-zero entries (sparse matrix). Therefore we have selected a clustering procedure based on Non-negative Matrix Factorization (NMF, [15]) that has shown good performance with binary data in the 'topic finding' literature ([15,28]).
NMF aims to reduce the dimensions of multivariate data by factorization X ≈ WH where X represents the binary matrix obtained from the associations of n genes and m classes in the user given sample gene list. Given the fixed number for r, two matrices W (size n × r) and H (size r × m) are produced as a result, representing the input data X in compressed form of r factors. The first of the matrices describes the loadings of the genes on r factors and is further used in clustering. In the clustering process, the genes are deposited into clusters by using a winner-takes-all approach that finds the factor with the highest loading for each gene from matrix W. The relation between the highest loading and sum of all loadings is used to measure the fitness of a gene in a cluster. In the visualization (see next chapter) the fitness is used to present genes in a sorted order for each cluster. More detailed descriptions concerning clustering binary data with NMF are given in [15,28]. We use the NMF algorithm presented in [29] which minimizes the least squares error (LSE) between the input data and resulting factorization.
Non-nested hierarchical clustering scheme
The core of the proposed method is a non-nested hierarchical clustering tree, which is shown in figure 4, step C. There the user given sample gene list is repeatedly clustered into r number of groups, where r grows gradually from two into a user given number. Each partitive clustering is executed from a random starting initialization using NMF, producing an independent division level to the visualization. The levels are placed consecutively in the growing order of r starting from r = 1, which represents the sample gene list without any clustering. In the visualization, each level is shown with a bar of constant size that is split into r sections. Each section represents a single cluster, the size of which is indicated by the width of the section. Correlations between each cluster in level r and all clusters of previous level r-1 are calculated by comparing cluster memberships of genes with a correlation measure between two binary classifications presented in [10]. The strongest correlation for each cluster is denoted by a line between the corresponding clusters. The width of the line indicates the magnitude of the correlation. The lines between the first and second levels present only the proportions of the genes, as the binary correlation with the first level can not be defined. Together the edges and sections form a non-nested hierarchical tree that visualizes the underlying heterogeneity in the gene and class association data.
Description of cluster contents
We have developed a procedure for describing the contents of gene clusters (figure 4, steps c and D) resulting from the non-nested hierarchical clustering scheme introduced above. There, a combination of three measures is applied to find informative classes by studying their over-representation in the sample gene list with and without clustering. By definition, the over-representation means a greater frequency of classes in the collected set of genes than in the rest of the population. A robust way to test this is the calculation of p-values from a hypergeometric distribution with Fisher's test [16,17], that we apply. Classes with low p-values are highly over- or under-represented in the gene set and thus interesting. Nevertheless, the significant p-values are small numbers that are difficult to handle and visualize. They neither distinguish the over- and under-represented classes. Thus, we use signed logarithmic transform of p-value introduced before [10] which has negative or positive sign depending on the under- or over-representation and suitable scale for visualization.
In our method, we study the over-representation for multiple purposes. We calculate the p-values for each biological class (description in figure 5):
Figure 5 The measures for studying over-representation of classes. Over-representation of classes is measured by using A) the whole sample gene list as a sample and the reference gene list as a remainder population, O.log(p); B) a single cluster as a sample and the rest of the sample gene list and the reference gene list as a remainder population, C.log(p); and C) a single cluster as a sample and the rest of the sample gene list as a remainder population excluding the reference gene list, S.log(p). In each situation, Fisher's exact test f(x, M, n, k) [16] is used to determine the over-representation. O.log(p) presents the original over-representation of sample gene list without clustering. C.log(p) highlights the classes that are over-represented in the original sample gene list and in individual cluster. S.log(p) reports the contribution to the formation of cluster structure.
A) From original sample gene list without any clustering using user given reference gene list as a rest of the population. This is denoted by O.log(p).
B) From each individual cluster using other clusters of sample gene list and reference gene list as a rest of the population. This is denoted by C.log(p).
C) From each individual cluster using other clusters of the sample gene list as a rest of the population and excluding the user given reference gene list. This is denoted by S.log(p).
In the default view of GENERATOR, these measures are used to show the over-represented functional classes in the clustering tree. In each cluster description, the basic over-representation measure C.log(p) is used to sort the classes. As C.log(p) is dependent on the clustering, it ranks high in some classes that are over-represented when measured from the cluster, but not over-represented when measured from the sample gene list without clustering. This is caused by the clustering process, when for example a tight group of genes is associated with the classes that are under-represented in the non clustered list. Since we aim at interpreting the whole list, such classes would be misleading and have to be removed. Therefore we filter them by using O.log(p), which is fully independent on the clustering. Another problem is that C.log(p) can rank high in some classes that have not contributed to formation of the analyzed cluster. These classes are under-represented in the cluster when comparing only to the rest of the sample gene list. Still they are so strongly over-represented in the whole sample gene list that C.log(p) shows over-representation. As these classes are uninformative for the analysis of the cluster, we filter them by using S.log(p). By default, the classes with O.log(p) < 2.0 or S.log(p) < 0.0 are filtered, although we encourage also the use of stricter cut-offs like 3 and 1, which we have found to work better especially with S.log(p). In the non-nested tree visualization, the two best classes from this filtered list are shown to describe cluster contents, and the longer list is available through the user interface (see Software Implementation in Results). The cluster is coloured according to the C.log(p) value of its most over-represented class with the strongest red for largest over-representation.
Analysis protocol
We study two clustering views for each data set in our analysis. In addition to previously discussed default view, which shows the over-represented classes in the clustering tree, we first study the cluster formation. For that, we sort the classes by S.log(p) within the clusters, which excludes the reference gene list and fully concentrates on clustered data. The outcomes from this setting are shown in fig. 1A and fig. 3A in Results. Similarly, the outcomes from the default view are shown in fig. 1(B) and fig. 3B. In our manual analysis (results shown in tables), we further filtered the results from default view by emphasizing the classes with O.log(p) > 3.0 and S.log(p) > 1.0. In addition to two different clustering views, we also study the stability of our clustering scheme with both datasets in figures 2 and 6 [see Additional file 1]. This helps us to detect random and non-random outcomes in similar way as with single clustering levels explained above.
List of abbreviations
GENERATOR GENElist Research Aimed Theme-discovery executOR
NMF Non-negative Matrix Factorization
GO Gene Ontology
SGD Saccharomyces Genome Database
MIPS Munich Information center for Protein Sequences
DAG Direct Acyclic Graph
LSE Least Squares Error
log(p) Signed 10 based Logarithmic Transform of p-value
O.log(p) Original log(p)
C.log(p) Complete log(p)
S.log(p) Sample log(p)
Authors' contributions
The original idea of the approach was introduced by PT. Design of the method and the associated software was contributed equally by both PT and PP. PP implemented and tested the software and executed the analysis for biological datasets. GW reviewed the results and provided advice and guidance on improving the analysis and performing comparison to other methods. The obtained results were interpreted by PT.
Supplementary Material
Additional File 5
GENERATOR detailed class output for H2O2 dataset. Table 5 The table presents a detailed view for the results obtained from GENERATOR with H2O2 data using five clusters. The table has two components. The first columns present the reported functional classes. The presented functional classes are selected with GENERATOR default view so that they highlight functional classes over-represented in the original list of genes and also in the cluster in question. In addition this part also shows the three reported p-values. These are the same as in the table 1. The table also includes number of class members in the cluster (inner size), number of class members in the original sample list (original inner size) and the number of class members in both sample and reference list (total size). These values can be used to analyse the proportion of the class that was included to cluster and the proportion of the class from the reported cluster. The reported classes can be also viewed as DAG with the included link to AMIGO www server. This enables the analysis of the hierarchy structure of the reported classes. The second part of table presents the list of genes for each cluster. Genes are presented in the sorted order so that the ones at the top of the list have always the strongest membership to the analyzed cluster.
Click here for file
Additional File 7
Detailed sorted class list for H2O2 dataset. Table 7 The table presents a detailed sorted list of over-represented classes obtained with H2O2 dataset. Classes with p-value < 0.01 (O.log(p) > 2.0) are shown.
Click here for file
Additional File 6
GENERATOR detailed class output for itraconanzole dataset. Table 6 The table presents a detailed view for the results obtained from GENERATOR with itraconanzole data using six clusters. One outlier cluster has not been taken in our analysis. Table has three main components. These are similar to the first three components in the table 5 [see Additional file 5].
Click here for file
Additional File 1
Four replications of non-nested hierarchical cluster tree with itraconanzole dataset. Figure 6 The figure shows four replications for the non-nested hierarchical clustering graph for itraconanzole dataset. We have marked the conserved gene clusters with the same Roman numerals as in figure 1. Notice again the conserved clusters observed over several levels in each cluster tree.
Click here for file
Additional File 8
Detailed sorted class list for itraconanzole dataset. Table 8 The table presents a detailed sorted list of over-represented classes obtained with itraconanzole dataset. Classes with p-value < 0.01 (O.log(p) > 2.0) are shown.
Click here for file
Additional File 2
H2O2 dataset analysed with Amigo DAG View. Figure 7 The figure presents the DAG view of all the reported classes shown in table 7 [see Additional file 7]. These classes had p-value < 0.01 (O.log(p) > 2.0). Figure was obtained from AMIGO server. The obtained figure was considered too complex for manual analysis.
Click here for file
Additional File 3
H2O2 dataset analysed with SGD DAG View. Figure 8 The figure presents the three DAG tree figures obtained from SGD GO term finder with the H2O2 data. The reported classes are colour coded according the reported p-value. We have marked the classes that were reported by some cluster in GENERATOR results by adding the number of corresponding cluster. Note that many classes that were not reported by GENERATOR are usually close in the hierarchy to already reported classes.
Click here for file
Additional File 4
Itraconanzole data analysed with SGD DAG View. Figure 9 Figure presents the three DAG tree figures obtained from SGD GO term finder using itraconanzole data. The reported classes are colour coded according the reported p-value. We have marked the classes that were reported by some cluster in GENERATOR results by adding the number of corresponding cluster. Note that many classes that were not reported by GENERATOR are usually close in the hierarchy to already reported classes.
Click here for file
Additional File 9
Summary of GOToolBox results with H2O2 dataset. Table 9 The table shows a summary of H2O2 dataset analysis with GOToolBox. Results are shown separately for three sub-ontologies of GO. Columns show GOToolBox cluster number (Cluster nb), number of gene products within each cluster (Number of genes), and obtained class description (Class names). The raw output from GOToolBox is available in table 10 [see Additional file 10].
Click here for file
Additional File 10
GOToolBox outputs from analysis with H2O2 and itraconanzole datasets. Table 10 Files include the clustering results for H2O2 and itraconanzole datasets from GOToolBox.
Click here for file
Acknowledgements
We would like to thank Dr. Aleksei Krasnov for comments. Jouni K. Seppänen and Ella Bingham provided the NMF Matlab code for preliminary testing of the method. PT and PP would like to thank the Finnish Cultural Foundation for financial support.
==== Refs
Gasch AP Spellman PT Kao CM Carmel-Harel O Eisen MB Storz G Botstein D Brown PO Genomic expression programs in the response of yeast cells to environmental changes Mol Biol Cell 2000 11 4241 4257 11102521
Thorpe GW Fong CS Alic N Higgins VJ Dawes IW Cells have distinct mechanisms to maintain protection against different reactive oxygen species: oxidative-stress-response genes Proc Natl Acad Sci U S A 2004 101 6564 6569 15087496 10.1073/pnas.0305888101
Eisen MB Spellman PT Brown PO Botstein D Cluster analysis and display of genome-wide expression patterns Proc Natl Acad Sci U S A 1998 95 14863 14868 9843981 10.1073/pnas.95.25.14863
Weng S Dong Q Balakrishnan R Christie K Costanzo M Dolinski K Dwight SS Engel S Fisk DG Hong E Issel-Tarver L Sethuraman A Theesfeld C Andrada R Binkley G Lane C Schroeder M Botstein D Michael Cherry J Saccharomyces Genome Database (SGD) provides biochemical and structural information for budding yeast proteins Nucleic Acids Res 2003 31 216 218 12519985 10.1093/nar/gkg054
Mewes HW Frishman D Guldener U Mannhaupt G Mayer K Mokrejs M Morgenstern B Munsterkotter M Rudd S Weil B MIPS: a database for genomes and protein sequences Nucleic Acids Res 2000 28 37 40 10592176 10.1093/nar/28.1.37
Apweiler R Bairoch A Wu CH Barker WC Boeckmann B Ferro S Gasteiger E Huang H Lopez R Magrane M Martin MJ Natale DA O'Donovan C Redaschi N Yeh LS UniProt: the Universal Protein knowledgebase Nucleic Acids Res 2004 32 D115 119 14681372 10.1093/nar/gkh131
Boeckmann B Bairoch A Apweiler R Blatter MC Estreicher A Gasteiger E Martin MJ Michoud K O'Donovan C Phan I Pilbout S Schneider M The SWISS-PROT protein knowledgebase and its supplement TrEMBL in 2003 Nucleic Acids Res 2003 31 365 370 12520024 10.1093/nar/gkg095
Boyle EI Weng S Gollub J Jin H Botstein D Cherry JM Sherlock G GO::TermFinder – open source software for accessing Gene Ontology information and finding significantly enriched Gene Ontology terms associated with a list of genes Bioinformatics 2004 20 3710 5 15297299
Hosack DA Dennis G JrSherman BT Lane HC Lempicki RA Identifying biological themes within lists of genes with EASE Genome Biol 2003 4 R70 14519205 10.1186/gb-2003-4-10-r70
Toronen P Selection of informative clusters from hierarchical cluster tree with gene classes BMC Bioinformatics 2004 5 32 15043761 10.1186/1471-2105-5-32
Wu LF Hughes TR Davierwala AP Robinson MD Stoughton R Altschuler SJ Large-scale prediction of Saccharomyces cerevisiae gene function using overlapping transcriptional clusters Nat Genet 2002 31 255 265 12089522 10.1038/ng906
Knuuttila J Törönen P Castren E Effects of antidepressant drug imipramine on gene expression in rat prefrontal cortex Neurochem Res 2004 29 1235 1244 15176480 10.1023/B:NERE.0000023610.89631.d6
Lähteinen S Pitkänen A Knuuttila J Törönen P Castren E Brain-derived neurotrophic factor signaling modifies hippocampal gene expression during epileptogenesis in transgenic mice Eur J Neurosci 2004 19 3245 54 15217381 10.1111/j.0953-816X.2004.03440.x
Ashburner M Ball CA Blake JA Botstein D Butler H Cherry JM Davis AP Dolinski K Dwight SS Eppig JT Harris MA Hill DP Issel-Tarver L Kasarskis A Lewis S Matese JC Richardson JE Ringwald M Rubin GM Sherlock G Gene ontology: tool for the unification of biology. The Gene Ontology Consortium Nat Genet 2000 25 25 29 10802651 10.1038/75556
Lee DD Seung HS Learning the parts of objects by non-negative matrix factorization Nature 1999 401 788 791 10548103 10.1038/44565
Fisher RA On the interpretation of c2 from contingency tables, and on the calculation of P J R Stat Soc 1922 85 81 94
Agresti A A Survey of Exact Inference for Contingency Tables Stat Sci 1992 1 131 177
Segal E R Y Koller D Genome-wide discovery of Transcriptional Modules from DNA Sequence and Gene Expression Bioinformatics 2003 19 i273 282 12855470 10.1093/bioinformatics/btg1038
Martin D Brun C Remy E Mouren P Thieffry D Jacq B GOToolBox: functional analysis of gene datasets based on Gene Ontology Genome Biol 2004 5 R101 15575967 10.1186/gb-2004-5-12-r101
Hughes TR Marton MJ Jones AR Roberts CJ Stoughton R Armour CD Bennett HA Coffey E Dai H He YD Kidd MJ King AM Meyer MR Slade D Lum PY Stepaniants SB Shoemaker DD Gachotte D Chakraburtty K Simon J Bard M Friend SH Functional discovery via a compendium of expression profiles Cell 2000 102 109 126 10929718 10.1016/S0092-8674(00)00015-5
The GENERATOR web page
The Gene Ontology web page
The AMIGO web server
Heger A Holm L Sensitive pattern discovery with 'fuzzy' alignments of distantly related proteins Bioinformatics 2003 19 i130 7 12855449 10.1093/bioinformatics/btg1017
Brunet JP Tamayo P Golub TR Mesirov JP Metagenes and molecular pattern discovery using matrix factorization Proc Natl Acad Sci U S A 2004 101 4164 9 15016911 10.1073/pnas.0308531101
Fränti P Xu M Kärkkäinen I Classification of binary vectors by using DeltaSC-distance to minimize stochastic complexity Pattern Recogn Lett 2003 24 65 73 10.1016/S0167-8655(02)00189-7
Cai L Huang H Blackshaw S Liu JS Cepko C Wong WH Clustering analysis of SAGE data using a Poisson approach Genome Biol 2004 5 R51 15239836 10.1186/gb-2004-5-7-r51
Seppänen JK Bingham E Mannila H Nada Lavrac, Dragan Gamberger, Hendrik Blockeel, Ljupco Todorovski A simple algorithm for topic identification in 0–1 data Knowledge Discovery in Databases: PKDD 2003; Cavtat-Dubrovnik, Croatia 2003 Springer 423 434
Lee DD Seung HS Todd K Leen, Thomas G Dietterich, Volker Tresp Algorithms for non-negative matrix factorization Adv Neural Info Proc Syst: 2001; Vancouver, British Columbia, Canada 2001 MIT Press 556 562
|
15987504
|
PMC1190153
|
CC BY
|
2021-01-04 16:27:25
|
no
|
BMC Bioinformatics. 2005 Jun 29; 6:162
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-162
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1681599847010.1186/1471-2105-6-168SoftwareHigh-Throughput GoMiner, an 'industrial-strength' integrative gene ontology tool for interpretation of multiple-microarray experiments, with application to studies of Common Variable Immune Deficiency (CVID) Zeeberg Barry R [email protected] Haiying [email protected] Sudarshan [email protected] Margot [email protected] Hong [email protected] David W [email protected] Mark [email protected] Robert M [email protected] David [email protected] Stanley K [email protected] Eldad [email protected] Danielle M [email protected] Thomas A [email protected] Charlotte [email protected] Donn M [email protected] David [email protected] John N [email protected] Genomics and Bioinformatics Group, Laboratory of Molecular Pharmacology, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892, USA2 Metabolism Branch, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892, USA3 SRA International, 4300 Fair Lakes CT, Fairfax, VA 22033, USA4 Advanced Biomedical Computing Center, National Cancer Institute at Frederick, SAIC Frederick, PO Box B, Frederick, MD, 21702, USA5 Laboratory of Parasitic Disease, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD 20892, USA6 The Mount Sinai Medical Center, 1425 Madison Avenue, New York, NY 10029, USA2005 5 7 2005 6 168 168 23 2 2005 5 7 2005 Copyright © 2005 Zeeberg et al; licensee BioMed Central Ltd.2005Zeeberg et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
We previously developed GoMiner, an application that organizes lists of 'interesting' genes (for example, under-and overexpressed genes from a microarray experiment) for biological interpretation in the context of the Gene Ontology. The original version of GoMiner was oriented toward visualization and interpretation of the results from a single microarray (or other high-throughput experimental platform), using a graphical user interface. Although that version can be used to examine the results from a number of microarrays one at a time, that is a rather tedious task, and original GoMiner includes no apparatus for obtaining a global picture of results from an experiment that consists of multiple microarrays. We wanted to provide a computational resource that automates the analysis of multiple microarrays and then integrates the results across all of them in useful exportable output files and visualizations.
Results
We now introduce a new tool, High-Throughput GoMiner, that has those capabilities and a number of others: It (i) efficiently performs the computationally-intensive task of automated batch processing of an arbitrary number of microarrays, (ii) produces a human-or computer-readable report that rank-orders the multiple microarray results according to the number of significant GO categories, (iii) integrates the multiple microarray results by providing organized, global clustered image map visualizations of the relationships of significant GO categories, (iv) provides a fast form of 'false discovery rate' multiple comparisons calculation, and (v) provides annotations and visualizations for relating transcription factor binding sites to genes and GO categories.
Conclusion
High-Throughput GoMiner achieves the desired goal of providing a computational resource that automates the analysis of multiple microarrays and integrates results across all of the microarrays. For illustration, we show an application of this new tool to the interpretation of altered gene expression patterns in Common Variable Immune Deficiency (CVID). High-Throughput GoMiner will be useful in a wide range of applications, including the study of time-courses, evaluation of multiple drug treatments, comparison of multiple gene knock-outs or knock-downs, and screening of large numbers of chemical derivatives generated from a promising lead compound.
==== Body
Background
The original version of GoMiner [1,2] was oriented toward visualization and interpretation of the results from a single microarray (or other high-throughput experimental platform), using a graphical user interface (GUI). Although the GUI can be used to examine the results from a number of microarrays one at a time, that is a rather tedious task, and there is no apparatus for obtaining a global picture of results from an experiment that consists of multiple microarrays:
• Suppose, for example, that combinatorial chemistry were used to generate a large number of derivatives of a lead compound. If microarrays were used to monitor the efficacy of those derivatives, then it is likely that none, or at most a few, of the microarrays would be interesting. It would be a thankless task to use the GUI to analyze and interpret the large number of uninformative microarrays. It would make much more sense to apply an automated batch procedure to generate a report that highlighted the interesting microarrays and then to examine just those in the GUI.
• As another example, suppose that a series of microarrays were used to generate a time-course. One would want to obtain a high-level, global picture of the relationships of the categories that were significant at different time points – for instance, to differentiate phases of a disease process or to explore the temporal sequence of events consequent to treatment with a drug.
High-Throughput GoMiner performs those tasks. As a tool for investigators with large sets of results, it complements and extends the GUI version's analysis and visualization capabilities. Both the command line and web application interfaces of High-Throughput GoMiner are freely available to all users [3]. To our knowledge, this is the first resource that integrates information and illuminates patterns from multiple microarrays in relationship to the Gene Ontology.
In the original GoMiner article [1,2], we noted that the Fisher's exact p-values require adjustment to account for the multiple comparisons problem. We proposed a resampling approach that would avoid major drawbacks of the Bonferroni correction (see, for example [4]) – the assumption of independence of categories and the likelihood of rejecting too many true positives. To provide a more balanced solution to the problem, we have now implemented a fast 'false discovery rate' (FDR) approach in High-Throughput GoMiner.
Another noteworthy feature of High-Throughput GoMiner is the integration of transcription factor binding site information with GO categorization and gene expression data so that the user can explore regulatory relationships. Although each of these types of information has been used alone previously, our approach of integrating them provides a powerful, novel analysis tool.
In addition to the functionality that it provides, High-Throughput GoMiner also serves as a model for integration of the command line interface of the original GoMiner into other applications. The command-line interface permits platform-independent integration of GoMiner's functionality into any data processing stream without modification of the GoMiner source code.
Overview of High-Throughput GoMiner
Throughout the text of this article, we will focus for concreteness of terminology on use of High-Throughput GoMiner for gene expression microarrays, but the range of application is much broader; it can be used for any high-throughput data set in which genes or proteins are flagged as 'interesting' for whatever reason, either as the result of a real experiment or a 'conceptual' in silico experiment.
The program requires two kinds of input: a list of the total set of genes on the microarray and a set of 'changed-gene' files. Each changed-gene file contains a subset of genes that the user considers interesting in the experiment (for example, genes that are under-or over-expressed). The formats of the files are the same as those required for the original GoMiner. High-Throughput GoMiner creates two types of output. The first is a set of reports and data files integrating the results from all of the microarrays. The second is a set of subdirectories, each of which contains results files for one of the microarrays.
Two especially noteworthy features of the data processing stream are (i) the implementation of a fast, efficient solution to the multiple testing problem (see sections on 'Computational Efficiency of High-Throughput GoMiner' and 'The Multiple Comparisons Problem in High-Throughput GoMiner: Estimating FDRs') and (ii) the integration of results from multiple microarrays.
There are a large number of output files. Some of them focus on the results from a particular microarray; others integrate the results from all of the microarrays. Both types can be used as input to tools such as Excel or CIMminer [5,6] to provide an integrative visualization of the results of one or all microarrays in the study. The companion web site [3] contains a detailed description of the input and output files.
Significant advances relative to the original version of GoMiner
High-Throughput GoMiner provides significant advances relative to the original GoMiner. Those advances are manifested in both scientific value-added and usability.
Scientific value-added
1. Integration across multiple microarray experiments: CIMs
The primary scientific value-added in High-Throughput GoMiner relative to the original GoMiner is integration of the results across multiple related microarrays. That type of integration is particularly useful when the set of changed-gene files represents a time course, such as progression of a disease, response to a drug, or development of an organism. The integration can also be useful in pharmaceutical discovery and development – for example in the parallel testing of many combinatorial chemistry products against cells or organisms when microarrays are used to provide multiplexed assay end-points. The diagnostics in High-Throughput GoMiner can indicate which compounds appear to be related in their activity to which GO categories.
Visual integration of results is achieved by producing CIMminer [6] input files. CIMminer is our program package for computing and displaying clustered image maps (CIMs). We introduced CIMs [5], also called clustered 'heat maps', in the mid-1990's, and they have since become the ubiquitous summary graphic for high-throughput 'postgenomic' data, for example from microarray experiments. We have extended the CIM paradigm to permit visualization of significant GO categories integrated across multiple microarrays.
2. Integration across multiple microarray experiments: category|gene-disease export files
Another form of integration is provided by a set of files each of which contains a matrix whose rows are category|changed gene pairs and whose columns are names of changed-gene files. When the names of changed-gene files are diseases to be compared, we refer to the output as the 'category|gene-disease' set. That set of output files is designed for analysis by the CIMminer program package. Clinical collaborators have found that the resulting CIMs greatly facilitate analysis of genes with altered expression and the interpretation of significant GO categories across related disease phenotypes.
3. CIMs of genes with altered expression versus significant GO categories
High-Throughput GoMiner generates a set of files for producing CIMs of genes with altered expression versus significant GO categories. Those CIMs facilitate determination of the relatedness of significant GO categories as defined by the degree of sharing of genes. The importance of that capability can be demonstrated by two canonical examples:
• Several GO categories can be combined into a single cluster if they contain essentially the same set of changed genes. Combining multiple categories in that manner brings about a simplification since the user can think in terms of a smaller number of clusters of related categories rather than in terms of a larger number of individual categories. That procedure is especially important because the parent-child structure of GO can result in a number of statistically significant categories that may contain nearly redundant sets of changed genes.
• Apparently unrelated GO categories might in fact be connected by containing changed genes in common. Such 'cross talk' can often explain apparently surprising instances in which unexpected GO categories achieve statistical significance.
4. Annotation of genes and GO categories with transcription factor binding site information
High-Throughput GoMiner generates automated annotation of genes and GO categories with transcription factor binding site information. For instance, one output in this set is a file whose columns are significant GO categories and whose rows are transcription factor binding sites for the genes within each category. This enhancement was motivated by the intense interest in inferring genomic regulatory networks from the results of microarray experiments.
Improvements in usability
1. Automated batch processing of an arbitrary number of changed-gene files
The original GoMiner requires a substantial number of manual operations to analyze a changed-gene file. That process is feasible for analysis of one or two files but becomes tedious, time-consuming, and error-prone when more files are to be analyzed. In a high-throughput context, there may be dozens or even hundreds of files to analyze for a study. It would be virtually impossible to use the original GoMiner to analyze that number of files. A human operator would almost inevitably introduce errors in the I/O, analysis, and book-keeping phases. Furthermore, in a large set of files, there may be only one or two that produce interesting results. Manual analysis of a hundred files to identify one or two files of interest would not be cost effective. In contrast, High-Throughput GoMiner automates both the analysis of any number of files and the selection of those files that are likely to be worth follow-up.
2. Automated report generation
High-Throughput GoMiner generates a report that summarizes the results from all microarrays, with the best ones – those with the largest number of significantly enriched categories – presented at the top of a sorted list. The user (or the computer if further downstream processing is to be performed) can focus principally on the top entries in the report.
The complete analysis process is documented and can be reviewed if a question arises later. If the original GoMiner had been used and a question arose about whether file number 57 was processed correctly, or whether the results from file number 57 were recorded correctly, many manual operations would need to be repeated. In contrast, High-Throughput GoMiner would require only a few moments. Accurate summary reports and a well-defined and reproducible directory structure make all of the results immediately available. The output files and directory structure permit user-friendly access to high-level information that characterizes all of the changed-gene files as well as to detailed information about the results for any particular changed-gene file. The output of High-Throughput GoMiner is essentially self-documenting. With the original GoMiner, manual record-keeping is a considerable burden; the automated record-keeping in High-Throughput GoMiner is a major asset.
3. Elimination of relatively slow access to our database server
Another important usability issue encountered by a number of users of the original GoMiner is relatively slow access to our database server. That issue can arise because of the high overhead of performing successive database accesses via the internet. The problem is not fatal, but in practice it can limit the number of changed-gene files that a user is able to process manually in a reasonable amount of time. To overcome that limitation, the original GoMiner web site provides support for the user with IT skills who wants to install a local version of the database. Unfortunately, such in-house IT support may be unavailable to many clinicians and biologist. The web version of High-Throughput GoMiner eliminates that problem because it processes the web-based queries on our server using the database that is local to our server.
Program description: procedures and files
High-Throughput GoMiner provides a choice of command line and web application interfaces. Here, we present a brief summary of both interfaces. The command line interface runs on Unix-based operating systems (including Mac OS X). Complete descriptions appear at the High-Throughput GoMiner website [3].
Both interfaces use the processing model envisioned in our original GoMiner article [1]:
1. A statistical operation identifies a set of changed genes.
2. The set of genes is listed in the GoMiner changed-gene file format.
3. GoMiner processes the total-and changed-gene files using a command line interface.
4. The results are exported from GoMiner and analyzed.
5. The user is notified which changed-gene files are of interest.
High-Throughput GoMiner generates two generic types of output. The first pertains to integrative results for all of the microarrays. The second pertains to each individual microarray.
In the integrative summary report, there are three entries for each microarray, corresponding to underexpressed, overexpressed, and total changed genes. The entries are sorted in descending order according to the potential interest of the result as indicated by the number of categories that satisfy a user-defined FDR. Integrative output files can be used to generate clustered image maps (CIMs) [5,6] showing 'significant categories' versus 'microarrays.' By invoking a program that identifies transcription factor binding sites (R. Stephens, unpublished), High-Throughput GoMiner can also generate CIMs for 'transcription factor binding sites' versus 'genes in a category' or 'transcription factor binding sites' versus 'categories in a microarray'.
The web application version works with any browser. The user uploads a total-gene file and either a single changed-gene file or a zip file containing a set of changed-gene files. The user receives an email containing a hyperlink to a URL from which to download a compressed archive containing the results.
Implementation
High-Throughput GoMiner is based on incorporation of command line GoMiner into a set of C Shell scripts. It is freely available and can be downloaded as a compressed tar file for use in a Unix-based environment. The scripts have been released under the GPL [7] open source license, so users are welcome to edit and extend them. The utilities used in the scripts (join, grep, gawk, sed, and curl) are commonly available in most Unix implementations or can be freely downloaded from sites such as the GNU Project [8]. The program was developed and tested on Mac OS X, Solaris, and Red Hat Linux. In our experience, there are slight variations among operating systems, computers, and versions of the Unix utilities. Porting between platforms might require minor user intervention, such as downloading gawk from the GNU Project rather than using awk.
The web application version of High-Throughput GoMiner (Figure 1) is implemented by using a simple Java servlet as a wrapper around the main scripts in the command-line version. The servlet manages the uploading of files, sets up an individual workspace for each request, and e-mails the user a URL for downloading results. To upload multiple changed-gene files, the user combines them into a single zip file. The web application version eliminates the minor manipulations described above for the command line version. It may take longer to complete requests because they are being executed in a shared environment, but the user interface is much simpler. The web version may be more suitable for the casual user who does not have the Unix background or resources to implement the command line version.
Figure 1 Schematic of stand-alone and web versions of High-Throughput GoMiner architecture and data flow.
Computational efficiency of High-Throughput GoMiner
High-Throughput GoMiner faces two daunting computational tasks: processing an arbitrary number of microarrays and performing many re-sampling instances in order to estimate the FDR. Each microarray typically requires hundreds of instances of re-sampling, and each instance requires the same computations as does analysis of the real data.
To make that computational burden manageable, we have developed a procedure that speeds up the processing time by several orders of magnitude relative to a naïve, brute force approach. At the heart of that procedure is the fact that the (real) total-gene file is used as both the (conceptual) total-and (conceptual) changed-gene files once, and the resulting gene-category export file is generated. The Unix 'join' utility is then applied to that file and to the (real) changed-gene files and re-sampled gene files in lieu of the much more time-consuming original GUI GoMiner process.
The multiple comparisons problem in High-Throughput GoMiner: estimating a false discovery rate (FDR)
Most investigators use microarray results to decide what follow-up studies to do, rather than as definitive evidence. The goal of the statistical analysis is to provide the experimenter with a good list of candidate categories for follow-up. To decide whether or not to follow up a category that appears enriched in changed genes, the experimenter should know the statistical reliability of the apparent enrichment. However, without a multiple comparisons correction, some categories would appear enriched (have a low p-value in the Fisher test) simply by chance. To assess the significance of a particular category, we need to know the distribution of p-values that would occur by random chance. The expected number of false positives should be some manageable percentage (for example, less than 10%) of the categories selected. The percentage of false positives to be tolerated will generally depend on the relative costs of false positives and false negatives in whatever follow-up study is to be done. This way of framing the question leads us to specify the false discovery rate (FDR) for a set of categories, rather than significance level (p-value) for each category [9]. In practice High-Throughput GoMiner reports q-values for individual categories. The q-value (see below) represents the smallest false discovery rate at which that category would be classed as enriched.
For focus we will consider only 'biological_process' categories in the following discussion. Depleted categories are of less interest to most investigators than are enriched ones, so they will be ignored. Within a given category, the enrichment Re is given by
Re = (nf/n)/(Nf/N)
where nf is the number of flagged genes within the particular category (i.e., genes whose expression levels are considered to be changed beyond a given threshold), n is the total number of genes within that same category, Nf is the number of flagged genes on the entire microarray, and N is the total number of genes on the microarray.
The enrichment values are derived from categories of different sizes and do not have a common distribution. To assess the number of false positives, we need measures that are directly comparable. Fisher's exact p-values (for the one-tailed test) meet this need. See the original publication on GoMiner [1] for an extensive discussion of the statistical and conceptual bases for choosing the Fisher's exact test.
One way to address the multiple comparisons problem is to associate an individual measure of reliability for an ordered list of most enriched categories by specifying the FDR for each. Storey et al. [9] call this a q-value. The distinction between a q-value and a Fisher's Exact p-value as the two are used in the context of High-Throughput GoMiner is as follows:. The p-value, uncorrected for multiple comparisons, is a measure of the statistical significance of a single category. The q-value of a category is the FDR of the list of categories whose p-values are equal to or smaller than the p-value of that category.
To estimate the q-value for each category at each level of significance, we use a resampling algorithm. First, we select random samples of Nf genes at each iteration and compute Fisher's exact test p-values for over-representation of the selected genes in all GO biological categories. After T resamplings, the q-value for the k-th most significant category is assessed as follows: We count the number of times that a Fisher's exact p-value less than or equal to the p-value, pk, of the k-th category, is found in the resampled data. Then, we divide by the number of resamplings:
mk = Σi=1,...,T N(p < pk ; i)/T,
where N(p < pk ; i) refers to the number of p-values less than pk on resampling iteration i. Finally,
qk = mk/Nf.
To ask how many permutation samples are needed before the qk values approximately reach their asymptotic values, we performed randomization studies (see 'Stability of Estimates of the False Discovery Rate' in Supplementary Materials [3]; [see Additional file 1]). Those studies indicate that the distribution of q-values for different categories in one resampling is usually not too different from the distribution of q-values for one category during many resamplings. Although T = 5000 is usual for permutation tests on microarray data, we found that, for data sets similar to those reported here for CVID, T = 1000 and T = 100 give essentially identical results. For any finite number of permutation samples, there is the possibility, of course, that one or more of the FDR estimates will be spuriously slightly below or slightly above the selected threshold value, but, as with ordinary p-values, slight deviations from the threshold shouldn't be over-interpreted. A menu in the web interface permits the user to select the appropriate number of resamplings. Only rarely does the ordering of categories by FDR differ from the ordering by Fisher's exact p-value. Thus, even though the FDR computation is only an approximation, it does not appreciably change which categories would have been given priority in the absence of multiple comparisons testing.
Results
Applying High-Throughput GoMiner to gene expression analysis of Clinical Common Variable Immunodeficiency (CVID)
We now illustrate the use of high-throughput GoMiner for interpretation of gene expression microarray data in a medical context, that of CVID. The input [see Additional file 2] and output [see Additional file 3] data are available as supplementary material.
Background information on CVID
CVID is the most common symptomatic primary immunodeficiency disease, manifested by low levels of switched immunoglobulin isotypes (IgG, IgA, IgE) in the serum and by lack of humoral immune response to specific antigens [10]. It is a heterogeneous disease characterized by defects in humoral and cellular immunity [11]. The disease usually occurs in the second or third decade of life, often heralded by recurrent pyogenic infection [12]. CVID is associated with an increased incidence of autoimmune disease [13,14].
High correlation of CVID and cancer
In one study, CVID patients were shown to have an 8-to 13-fold increased incidence of cancer overall, with a 438-fold increase in lymphoma for females [15]. Another, larger study showed an overall increased incidence of cancer of 1.8-fold, with the relative risks of stomach cancer and lymphoma at 10.3-and 12.1-fold, respectively [16]. The risk for non-Hodgkin's lymphoma over the period of 25 years has been estimated to be between 1.4% and 7.0% [17].
Using microarray technology to study global gene expression in CVID
Microarray technology and analysis tools have made it possible to study global gene expression patterns in primary blood cells from CVID patients. To our knowledge this is the first such study, and it can contribute valuable information to what is currently known about the pathology and pathogenesis of the disease.
Experimental methodology
Global gene expression patterns in twenty CVID patients are currently under study in our laboratory. One patient was selected for preliminary analysis and proof of concept for High-Throughput GoMiner. That patient had typical symptoms and laboratory findings, including repeated respiratory infections and low levels of all serum immunoglobulins.
Peripheral blood mononuclear cells (PBMC) from fresh blood were stimulated with CD3 and CD28 for 24 hours and used for RNA extraction. mRNA of the sample were amplified into aRNA and coupled with Cy3 or Cy5 fluorescent dye for microarray hybridization.
A detailed description of materials, methods, and data processing is provided in the Supplementary Materials [3]; [see Additional file 4]. An illustrative summary report (Figure 2) and clustered image map (CIM; Figure 3) are given in the main text (below).
Figure 2 Screen shot of High-Throughput GoMiner results in Excel for GO categories enriched in genes with altered expression. The 30 GO categories with FDR = 0.10 are color-coded red; the other GO categories are color-coded blue.
Figure 3 Clustered image map (CIM) [5,6] showing GO categories versus genes for genes with altered expression in a patient with CVID. Yellow indicates absence of the gene from the GO category. Red and green indicate over-and underexpressed genes, respectively. Clustering was performed with the Pearson correlation metric and average linkage algorithm. Instructions for using CIMminer to generate the CIMs in this paper are given in Supplementary Materials [see Additional file 5].
High-Throughput GoMiner identifies biologically-relevant categories
The CIM (Figure 3) serves as a fingerprint of the patient's molecular phenotype. The 24 genes with altered expression that caused these GO categories to be selected are shown at the bottom, and the categories are listed along the left-hand side. The largest category selected was 'response to external stimuli' (enrichment = 2.6-fold; p = 10-3.9; FDR = not detectable). 'Not detectable' means that the FDR was not distinguishable from zero given the number of randomizations used. See 'Summary Report' in the Supplementary Materials [3] for further details [see Additional file 6]. 'Response to external stimuli' contained 16 changed genes. All but one of those genes (DEFB1) were also in 'immune response' (enrichment = 3.4-fold; p = 10-5.1; FDR = not detectable). The Summary Report shows that, of the 254 genes in this GO category, 15 exhibited significant differences in expression level when the patient was compared with the normal controls. The categories 'cell surface receptor linked signal transduction' (enrichment = 2.7-fold; p = 10-2.1; FDR = 0.09), 'cell adhesion' (enrichment = 3.3-fold; p = 10-2.1; FDR = 0.10), and 'organismal physiological process' (enrichment = 2.8-fold; p = 10-4.3; FDR = not detectable), along with their child categories, contained the other changed genes.
The pathology of CVID in this patient was reflected by the GO categories with low values of FDR. CVID patients have defects in production of specific antibodies, and those defects may be associated with disturbed expression of 'immune/defense response genes' [18-22]. As indicated in Figure 3 and in Summary Report, the 'immune response' (enrichment = 3.4-fold; p = 10-5.1; FDR = not detectable), 'defense response' (enrichment = 3.2-fold; p = 10-4.8; FDR = not detectable), and related GO categories were prominent. The pathogenesis of CVID is not fully understood, but, from other primary immune deficiency diseases, we know that immunodeficiency may result from defects in 'antigen processing' (enrichment = 12.8-fold; p = 10-2.0; FDR = 0.10) [23,24], 'antigen presentation exogenous antigen' (enrichment = 12.8; p = 10-2.0; FDR = 0.10) [25], 'humoral immune responses' (enrichment = 5.4-fold; p = 10-2.7; FDR = 0.05) [26], and/or 'cell-cell signaling' (enrichment = 4.7-fold; p = 10-2.4; FDR = 0.06) [27-29]. Altered expression of genes in those categories may affect B cell activation, differentiation and maturation, and, ultimately, immunoglobulin production in CVID [30,31].
Several forms of signal transduction, not previously associated in the literature with CVID, were also enriched. Signal transduction plays an indispensable role in the immune response, and we suggest that signal transduction also plays an important role in CVID [27-29]. Defective B-or T-cell signaling can cause immunodeficiency, as can defective cytokine production or action [28,29]. Our microarray results show that several genes related to signal transduction (CCL2, CXCL9, EMR2, GPR56, RGS2, TGFBR2, RGFBR3, and TGM2 in the categories 'cell surface receptor linked signal transduction' (enrichment = 2.7-fold; p = 10-2.1; FDR = 0.09) and 'G-protein coupled receptor protein signaling pathway' (enrichment = 5.1-fold; p = 10-3.1; FDR = 0.04) are differentially expressed in the CVID patient. Full details of the roles of these genes in signal transduction, as well as the roles of all genes mentioned in this article, can be conveniently obtained via the hyperlinks to NCBI Entrez [32] in the Gene Category Report (Supplementary Materials [3]; [see Additional file 7]).
High-Throughput GoMiner CIM facilitates grouping of closely-related categories into a single cluster
The 30 significant GO categories can be grouped into a smaller number of clusters, each of which contains several closely related categories. Grouping is desirable because it can remove the parent-child node redundancy inherent in the 'directed acyclic graph' (DAG) structure of GO. Removal of that redundancy effects a modest 'dimensionality reduction' and simplifies interpretation of the results vis a vis the disease phenotype. The CIM (Figure 3) of GO categories versus genes was used to create a tabulation (Table 1) of 7 clusters that result from grouping the 30 significant GO categories.
Table 1 Clusters Of Categories Derived From The CIM (Figure 3)
Cluster Number Cluster Name Category
1 Exogenous Antigen antigen presentation exogenous antigen
antigen processing exogenous antigen via MHC class II
2 Xenobiotic xenobiotic metabolism
response to xenobiotic stimulus
3a Signaling cell surface receptor linked signal transduction
G-protein coupled receptor protein signaling pathway
4 Homeostasis calcium ion homostasis
di-tri-valent inorganic cation homeostasis
cell ion homeostasis
cation homeostasis
metal ion homeostasis
ion homeostasis
5a Response response to chemical substance
response to abiotic stimulus
3b Signaling cell-cell signaling
6 Taxis taxis
chemotaxis
5b Response cellular defense response
5c Response response to stimulus
response to external stimulus
organismal physiological process
immune response
response to biotic stimulus
defense response
5d Response response to wounding
5e Response response to pest pathogen parasite
response to stress
7 Adhesion cell adhesion
5f Response humoral immune response
humoral defense mechansim
The CIM also permits detection of 'cross-talk' between GO categories that might at first appear to be unrelated. A number of examples can be found in Figure 3. For example, 'G-protein coupled receptor protein signaling pathway' and 'cell adhesion' both contain the changed genes GPR56 and CCL2. Detailed analysis of cross-talk can potentially provide an important systems biology interpretation of the particular set of significant GO categories in a disease state.
High-Throughput GoMiner transcription factor binding site CIM can help to detect genomic regulatory networks
Figure 4 indicates the richness of information available for inference of genomic regulatory networks from a CIM for transcription factor binding site vs. GO category. A full-size version in which all the transcription factor binding site names are readable is available in the Supplementary Materials [see Additional file 8]. Among the numerous relationships that can be mined by a systematic analysis of this novel type of CIM is a set of transcription factors (Table 2) that co-regulate the changed genes in the GO category 'G-protein coupled receptor protein signaling pathway' and a large core of 'response' categories.
Figure 4 Clustered image map (CIM) 5,6 showing transcription factor binding sites versus GO categories in a patient with CVID. Red indicate FDR = 0.0, and yellow indicates FDR > 1.0 or a missing value. Clustering was performed with the Pearson correlation metric and average linkage algorithm. The inset is a blow-up of the first few transcription factor binding site names. A full-size version in which all the transcription factor binding site names are readable is available in Supplementary Materials [see Additional file 8]. There are 35 rather than 30 GO categories because this result was computed with a more recent version of the GO Consortium database.
Table 2 Names and Consensus Sequences for Transcription Factors that Co-Regulate the Changed Genes in the GO Category 'G-protein Coupled Receptor Protein Signaling Pathway' and a Large Core of 'Response' Categories (Figure 4 and Supplementary Materials [see Additional file 8])
Name Consensus
AP-1-IL-3 TGAGTCA
AP-1-involucrin-H2 TGCCTCA
ASP-CYP21 CTCTGTGG
AlphaCE2Maf-Bfsp1 TGCTGAC
B2_RS TCCTATCA
CP2-consensus GCNMNANCMAG
CPSI-B1 TCTCCCA
EBF/Olf-1_site_2 TCCCNNRRGRR
EBV-ZRE2 TGAGCAA
FHX-type-A-CS WMARYAAAYA
GAGA_box/CT_element AGAGARRRR
GH-CSE2 AATAAAT
GRE_CS7 WCTGWTCT
GRE_CS8 AGAWCAGW
GT-2B_RS CCAGCTG
GT-IIBa-SV40 ACAGCTG
HNF5-erk1 TATTTGT
HiNF-Ahist AGAAATG
IL4-P0/P1 ATTTTCC
Initiator_CS CTCANTCT
MEF-2-consensus YTWWAAATAR
MEF-2_CS YTAWAAATAR
NF-Y-consensus BVDCCAATVVVVD
PuF_RS GGGTGGG
RadLV-core TGTGGTCA
Runx_CS AACCACA
Six5_CS TCARRTTNC
Sp1-VGF_1 AGGGAGG
TCF-2-alpha_CS SAGGAAGY
TCR-beta-site-6 AATACAA
TRE.1 TGACTCA
c-Myc_RS1 TCTCTTA
c-mos_DS3 GTTTTAA
delta-rpL7 GGAGGCTG
forkhead_CS WAARYAAAYW
p300-consensus GGGAGTG
A detailed discussion of differential expression of individual genes is provided in Supplementary Materials [3]; [see Additional file 9].
Correlation of GO categories and disease phenotype
The GO categories and genes identified by High-Throughput GoMiner are shown in Figure 3 and in Gene Category Report. The categories and genes correlate well with the disease's phenotype. Thus, High-Throughput GoMiner can integrate information from entire gene expression microarray studies into a coherent picture of biological process gene category and disease phenotype at the molecular level. It provides information with which researchers can develop new hypotheses or explore potential therapeutic targets.
Applying High-Throughput GoMiner to gene expression analysis of schistosomiasis
To highlight how High-Throughput GoMiner can be used to integrate time series data, we consider briefly an example from preliminary analysis of gene expression in schistosomiasis (Elnekave et al., in preparation). Schistosomiasis is associated with bladder cancer in third-world countries [33-36]. Figure 5 shows the GO categories with low FDR for overexpressed genes in the form of a 3-D bar graph in Excel. Figure 6 shows a clustered image map (CIM) generated using CIMminer [6] to show the time course.
Figure 5 Time series for GO categories with low FDR for overexpressed genes. The data were obtained from a study of schistosomiasis in a murine model [37-40] over the course of 20 weeks after infection. 3D bar graph visualization in Excel. (Elnekave et al, in preparation).
Figure 6 CIM [5,6] with hierarchically clustered categories (Pearson correlation, average linkage clustering) versus time (Elnekave et al., in preparation).
Discussion
Comparison of High-Throughput GoMiner with related programs
While this work was in progress, a number of tools for GO analysis of microarray data have become available. Included are EASE [41], FatiGO [42], FunSpec [43], GoSurfer [44,45], GO::TermFinder [46,47], Onto-Express [48], and Ontology Traverser [49]. However, none of them have the central integrative features that characterize High-Throughput GoMiner. Of those tools, only GO::TermFinder permits batch processing. Since the others have been reviewed previously and since the main features of High-Throughput GoMiner relative to the original GoMiner package derive from batch processing, we will confine the present discussion to a comparison between GO::TermFinder and High-Throughput GoMiner.
High-Throughput GoMiner appears to offer all of the functionality of GO::TermFinder, as well as additional important features. In an earlier section, entitled 'Significant Advances Relative to the Original Version of GoMiner,' we detailed the scientific value-added and usability features of High-Throughput GoMiner vis a vis the original GoMiner. Those same scientific value-added and usability features distinguish High-Throughput GoMiner from GO::TermFinder. Because the detailed description can be found in the above-mentioned section and technical details can be found in the Output Files page of the High-Throughout GoMiner web site [3], we confine ourselves here to listing the scientific value-added and usability features of High-Throughput GoMiner that are not present in GO::TermFinder:
1. Integration of information across multiple microarrays in a study: CIMs (e.g., Figure 5, 6).
2. Integration across multiple microarrays: category|gene-disease export files.
3. CIMs of genes with altered expression versus significant GO categories (Figure 3).
4. Annotation of genes and GO categories with transcription factor binding site information (Figure 4).
5. Prioritization of microarrays in a study on the basis of the number of categories that are statistically significantly enriched.
There are also a number of usability features that distinguish High-Throughput GoMiner from GO::TermFinder:
6. Greater simplicity in the running of High-Throughput GoMiner. The GO::TermFinder batch processing feature [50] is implemented through the 'analyze.pl' module. The implementation requires that the user supply both an annotation file and an ontology file, obtained from the GO Consortium web site. In contrast, High-Throughput GoMiner accesses the annotation and ontology information through a database that we maintain on our server, so the low-level implementation details are transparent to the user.
7. Greater generalization of annotations in High-Throughput GoMiner. The default mode of High-Throughput GoMiner includes all annotations in the GO Consortium database, with the option to restrict the annotations to any arbitrary combination of annotation sources.
8. Greater control over species selection for High-Throughput GoMiner. The default mode of High-Throughput GoMiner permits restricting the query to any single species (e.g., mouse or human) or to any combination of species represented within the GO Consortium database. That is a functionally important type of flexibility.
9. Recognition of HUGO gene names by High-Throughput GoMiner. The High-Throughput GoMiner database can recognize HUGO names as well as any of the other identifier types provided by the GO Consortium database. The ability to recognize HUGO names is not an inherent feature of the annotation provided by the GO Consortium, so users of GO::TermFinder are not able to access HUGO names in their queries.
10. Platform-independence of the Web Interface Version of High-Throughput GoMiner. The Web Interface Version of High-Throughput GoMiner is platform-independent because it is a web server application. In contrast, a Unix environment is required to use the batch processing capability of GO::TermFinder [51].
11. Output formatting differs between High-Throughput GoMiner and GO::TermFinder. The latter appears to present output in a list format, whereas several modes of output (at the level of either individual microarray or integration of all microarrays) are available with High-Throughput GoMiner.
In summary, of the tools available for using GO to interpret microarray (or analogous) data, only GO::TermFinder and High-Throughput GoMiner offer batch-processing capability. GoMiner, in addition, provides tools and visualizations for integrating information from the batch of microarrays and for relating them to transcription factor binding sites and regulatory networks. Because of additional scientific and usability characteristics of High-Throughput GoMiner, it is particularly well suited to the needs of the molecular biology, genomics, and proteomics communities.
Acknowledgements
The research was supported [in part] by the Intramural Research Program of the N1H, National Cancer Institute, Center for Cancer Research.
Conclusion
High-Throughput GoMiner efficiently performs the computationally challenging task of automated batch-processing of an arbitrary number of microarrays (or other conceptually similar sets of large data sets). To our knowledge, it is the first resource for integration of high-throughput analyses of multiple microarrays. The automatically generated output files permit visualization of time series data in a 3-D bar chart in Excel or as a hierarchically clustered image map (CIM) of the interesting GO categories in relation to expression (or transcription factor binding sites).
As a proof of concept, we used High-Throughput GoMiner to analyze the results of a microarray study of differences in gene expression between a patient with CVID and normal controls. Because the phenotypes of immunodeficiency diseases have been well described, we were able to demonstrate that the GO categories found by High-Throughput GoMiner were those that were expected (e.g., 'immune response'). Since this was the first global gene expression study of CVID, the analyses using High-Throughput GoMiner have provided new information on biological process categories and specific genes in the disease. For example, most of the differentially expressed genes were found in signal transduction categories. Signal transduction had not previously been reported to play a role in CVID. High-Throughput GoMiner thus has the potential to generate new biomedical hypotheses and identify new targets for research.
We have tried to make High-Throughput GoMiner as flexible as possible by providing both command line and web server versions. The command line version provides faster calculation and the potential for highly parallel processing; the web server version provides transparency and ease of use. The next phases of genomics and proteomics will impose increasing demands for flexible, large-scale, automated information processing. We see High-Throughput GoMiner as a key resource for addressing that challenge.
Availability and requirements
Project name: High-Throughput GoMiner; Project home page: ; Operating system(s): web version is platform independent, command line version requires Unix; Programming language: java and Unix C shell; Other requirements: detailed on web site and on documentation packaged in command line version download; License: GNU GPL; Restrictions to use by non-academics: none.
Supplementary Material
Additional File 1
Stability of Estimates of the False Discovery Rate
Click here for file
Additional File 2
Expression Data
Click here for file
Additional File 3
Output Files Generated from High-Throughput GoMiner
Click here for file
Additional File 4
Methodology Description
Click here for file
Additional File 6
Summary Report
Click here for file
Additional File 7
Gene Category Report
Click here for file
Additional File 8
CIM of Transcription Factors versus GO Categories
Click here for file
Additional File 9
Discussion of Results
Click here for file
Additional File 5
Instructions for Generating the CIMs in the Manuscript
Click here for file
Acknowledgements
We thank M.D. Wang, W. Zheng, and G. Wang of Georgia Institute of Technology and Emory University and W.C. Reinhold, S. Lababidi, K.J. Bussey, J. Riss, and J.C. Barrett of the CCR, NCI for their collaboration in developing the original GoMiner package. This project is being supported in part by a contract with SRA International, Inc. funded by the CCR, NCI and by the Cancer Biomedical Informatics Grid (caBIG) project of the NCI Center for Bioinformatics, DCTD, NCI. This research was supported by the Intramural Research Program of the National Cancer Institute (NCI) of the National Institutes of health (NIH).
==== Refs
Zeeberg BR Feng W Wang G Wang MD Fojo AT Sunshine M Narasimhan S Kane DW Reinhold WC Lababidi S GoMiner: a resource for biological interpretation of genomic and proteomic data Genome Biol 2003 4 R28 12702209
GoMiner
High-Throughput GoMiner
Bonferroni
Weinstein J Myers T O'Connor P Friend S Fornace A JrKohn K Fojo T Bates S Rubinstein Anderson N Buolamwini J van Osdol W Monks A Scudiero D Sausville E Zaharevitz D Bunow B Viswanadhan V Johnson G Wittes Paull K An information-intensive approach to the molecular pharmacology of cancer Science 1997 275 343 349 8994024
CIMminer
GPL License
GNU Project
Storey JD Tibshirani R Statistical significance for genome-wide studies Proc Natl Acad Sci 2003 100 9440 9445 12883005
Spickett GP Current perspectives on common variable immunodeficiency (CVID) Clin Exp Allergy 2001 31 536 542 11359420
Ballow M Primary immunodeficiency disorders: antibody deficiency J Allergy Clin Immunol 2002 109 581 591 11941303
Simonte SJ Cunningham-Rundles C Update on primary immunodeficiency: defects of lymphocytes Clin Immunol 2003 109 109 118 14597209
Cunningham-Rundles C Hematologic complications of primary immune deficiencies Blood Rev 2002 16 61 64 11913998
Warnatz K Wehr C Drager R Schmidt S Eibel H Schlesier M Peter HH Expansion of CD19(hi)CD21(lo/neg) B cells in common variable immunodeficiency (CVID) patients with autoimmune cytopenia Immunobiology 2002 206 502 513 12607725
Cunningham-Rundles C Siegal FP Cunningham-Rundles S Lieberman P Incidence of cancer in 98 patients with common varied immunodeficiency J Clin Immunol 1987 7 294 299 3611296
Mellemkjaer L Hammarstrom L Andersen V Yuen J Heilmann C Barington T Bjorkander J Olsen JH Cancer risk among patients with IgA deficiency or common variable immunodeficiency and their relatives: a combined Danish and Swedish study Clin Exp Immunol 2002 130 495 500 12452841
Cunningham-Rundles C Cooper DL Duffy TP Strauchen J Lymphomas of mucosal-associated lymphoid tissue in common variable immunodeficiency Am J Hematol 2002 69 171 178 11891803
Groth C Drager R Warnatz K Wolff-Vorbeck G Schmidt S Eibel H Schlesier M Peter HH Impaired up-regulation of CD70 and CD86 in naive (CD27-) B cells from patients with common variable immunodeficiency (CVID) Clin Exp Immunol 2002 129 133 139 12100033
McQuaid A Tormey VJ Trafford B Webster AD Bofill M Evidence for increased expression of regulatory cytokine receptors interleukin-12R and interleukin-18R in common variable immunodeficiency Clin Exp Immunol 2003 134 321 327 14616794
Di Renzo M Serrano D Zhou Z George I Becker K Cunningham-Rundles C Enhanced T cell apoptosis in common variable immunodeficiency: negative role of the fas/fasligand system and of the Bcl-2 family proteins and possible role of TNF-RS Clin Exp Immunol 2001 125 117 122 11472434
Holm AM Aukrust P Aandahl EM Muller F Tasken K Froland SS Impaired secretion of IL-10 by T cells from patients with common variable immunodeficiency – involvement of protein kinase A type I J Immunol 2003 170 5772 5777 12759461
Farrington M Grosmaire LS Nonoyama S Fischer SH Hollenbaugh D Ledbetter JA Noelle RJ Aruffo A Ochs HD CD40 ligand expression is defective in a subset of patients with common variable immunodeficiency Proc Natl Acad Sci U S A 1994 91 1099 1103 7508119
Schiffer R Baron J Dagtekin G Jahnen-Dechent W Zwadlo-Klarwasser G Differential regulation of the expression of transporters associated with antigen processing, TAP1 and TAP2, by cytokines and lipopolysaccharide in primary human macrophages Inflamm Res 2002 51 403 408 12234057
Faigle W Raposo G Tenza D Pinet V Vogt AB Kropshofer H Fischer A de Saint-Basile G Amigorena S Deficient peptide loading and MHC class II endosomal sorting in a human genetic immunodeficiency disease: the Chediak-Higashi syndrome J Cell Biol 1998 141 1121 1134 9606205
Abdulkadir SA Ono SJ How are class II MHC genes turned on and off? FASEB J 1995 9 1429 1435 7589984
Stead A Douglas JG Broadfoot CJ Kaminski ER Herriot R Humoral immunity and bronchiectasis Clin Exp Immunol 2002 130 325 330 12390323
Conley ME Genetics of primary immunodeficiency diseases Rev Immunogenet 2000 2 231 242 11258420
Wienands J Signal transduction elements of the B cell antigen receptor and their role in immunodeficiencies Immunobiology 2000 202 120 133 10993287
Notarangelo LD Giliani S Mella P Schumacher RF Mazza C Savoldi G Rodriguez-Perez C Badolato R Mazzolari E Porta F Candotti F Ugazio AG Combined immunodeficiencies due to defects in signal transduction: defects of the gammac-JAK3 signaling pathway as a model Immunobiology 2000 202 106 119 10993286
Stagg AJ Funauchi M Knight SC Webster AD Farrant J Failure in antigen responses by T cells from patients with common variable immunodeficiency (CVID) Clin Exp Immunol 1994 96 48 53 8149665
Sneller MC Common variable immunodeficiency Am J Med Sci 2001 321 42 48 11202479
NCBI Entrez
Ji W Naguib RN Ghoneim MA Neural network-based assessment of prognostic markers and outcome prediction in bilharziasis-associated bladder cancer IEEE Transactions on Information Technology in Biomedicine 2003 7 218 24 14518736
Zacharia B Sherman P Atopy, helminths, and cancer Medical Hypotheses 2003 60 1 5 12450762
Fadl-Elmula I Kytola S Leithy ME Abdel-Hameed M Mandahl N Elagib A Ibrahim M Larsson C Heim S Chromosomal aberrations in benign and malignant bilharzia-associated bladder lesions analyzed by comparative genomic hybridization BMC Cancer 2002 2 5 11914143
Samandouras G King A Kellerman AJ Schistosoma haematobium presenting as an intrinsic conus tumour British Journal of Neurosurgery 2002 16 296 300 12201402
Lucey DR Maguire JH Schistosomiasis Infect Dis Clin North Am 1993 7 635 653 8254164
Elliott DE Schistosomiasis. Pathophysiology, diagnosis, and treatment Gastroenterol Clin North Am 1996 25 599 625 8863042
Hoffmann KF Wynn TA Dunne DW Cytokine-mediated host responses during schistosome infections; walking the fine line between immunological control and immunopathology Adv Parasitol 2002 52 265 307 12521263
Pearce EJ MacDonald AS The immunobiology of schistosomiasis Nat Rev Immunol 2002 2 499 511 12094224
Hosack DA Dennis G Sherman BT Lane HC Lempicki RA Identifying biological themes within lists of genes with EASE Genome Biol 2003 4 R70 14519205
Al-Shahrour F Díaz-Uriarte R Dopazo J FatiGO: a web tool for finding significant associations of Gene Ontology terms with groups of genes Bioinformatics 2004 20 578 580 14990455
Robinson MD Grigull J Mohammad N Hughes TR FunSpec: a web-based cluster interpreter for yeast BMC Bioinformatics 2002 3 35 12431279
Zhong S Storch KF Lipan O Kao MC Weitz CJ Wong WH GoSurfer: A graphical interactive tool for comparative analysis of large gene sets in gene ontology Appl Bioinformatics 2004 3 261 264 15702958
GoSurfer
Boyle EI Weng S Gollub J Jin H Botstein D Cherry JM Sherlock G GO::TermFinder – open source software for accessing Gene Ontology information and finding significantly enriched Gene Ontology terms associated with a list of genes Bioinformatics 2004 20 3710 3715 15297299
GO::TermFinder
Khatri P Draghici S Ostermeier GC Krawetz SA Profiling gene expression using onto-express Genomics 2002 79 266 270 11829497
Ontology Traverser
GO::TermFinder batch processing
GO::TermFinder README
|
15998470
|
PMC1190154
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Jul 5; 6:168
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-168
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1741601180810.1186/1471-2105-6-174Research ArticleProtein subcellular localization prediction for Gram-negative bacteria using amino acid subalphabets and a combination of multiple support vector machines Wang Jiren [email protected] Wing-Kin [email protected] Arun [email protected] Kuo-Bin [email protected] Bioinformatics Institute, 30 Biopolis Street, #07-01 Matrix, Singapore 1386712 Department of Computer Science, National University of Singapore, 3 Science Drive 2, Singapore 1175432005 13 7 2005 6 174 174 14 2 2005 13 7 2005 Copyright © 2005 Wang et al; licensee BioMed Central Ltd.2005Wang et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Predicting the subcellular localization of proteins is important for determining the function of proteins. Previous works focused on predicting protein localization in Gram-negative bacteria obtained good results. However, these methods had relatively low accuracies for the localization of extracellular proteins. This paper studies ways to improve the accuracy for predicting extracellular localization in Gram-negative bacteria.
Results
We have developed a system for predicting the subcellular localization of proteins for Gram-negative bacteria based on amino acid subalphabets and a combination of multiple support vector machines. The recall of the extracellular site and overall recall of our predictor reach 86.0% and 89.8%, respectively, in 5-fold cross-validation. To the best of our knowledge, these are the most accurate results for predicting subcellular localization in Gram-negative bacteria.
Conclusion
Clustering 20 amino acids into a few groups by the proposed greedy algorithm provides a new way to extract features from protein sequences to cover more adjacent amino acids and hence reduce the dimensionality of the input vector of protein features. It was observed that a good amino acid grouping leads to an increase in prediction performance. Furthermore, a proper choice of a subset of complementary support vector machines constructed by different features of proteins maximizes the prediction accuracy.
==== Body
Background
Subcellular localization is a key functional attribute of a protein. Since cellular functions are often localized in specific compartments, predicting the subcellular localization of unknown proteins may be used to obtain useful information about their functions and to select proteins for further study. Moreover, studying the subcellular localization of proteins is also helpful in understanding disease mechanisms and for developing novel drugs.
As a result of large-scale genome sequencing efforts in recent years, protein data has accumulated in public data banks at an increasing rate. Analyzing protein data to extract useful knowledge is thus essential for projects like automatic annotation. It is desirable to have an automated and reliable system for predicting subcellular localization of proteins from amino acid sequences.
A number of efforts [1-21] have been made to predict protein subcellular localization. Most of these prediction methods can be classified into two categories: one based on the recognition of protein N-terminal sorting signals and the other based on amino acid compositions [22].
Previous works have been focused on protein localization prediction for Gram-negative bacteria. There are five primary localization sites in Gram-negative bacteria, which are the cytoplasm, the extracellular space, the inner membrane, the outer membrane, and the periplasm. PSORT I [23] is the most widely used tool for predicting multiple localizations for Gram-negative bacteria. It uses biological knowledge represented by "if-then" rules for predicting protein localization sites. Most of these rules were derived from experimental observations. However, the PSORT I does not consider the extracellular space site. Additionally, the overall recall for the data set [24] only attains 60.9%.
Gardy et al. [24] presented PSORT-B to improve the prediction performance of PSORT I. PSORT-B combines information of the amino acid composition, similarity to proteins of known localization, presence of a signal peptide, transmembrane alpha-helices and motifs corresponding to specific localizations for a given protein sequence, through a probabilistic approach. It returns a list of five possible localization sites with associated probability scores. It attains an overall recall of 74.8% for the same data set mentioned above.
Recently, Yu et al. [25] proposed a predictive system called CELLO for Gram-negative bacteria by using support vector machines based on n-peptide compositions. They classified 20 amino acids into four groups (charged, polar, aromatic and nonpolar) to reduce the dimensionality of the input vector. Forty SVM classifiers were used to predict the localization sites. Their overall recall was 88.9%. It was a significant improvement over the previous results of PSORT-B. However, the recall for extracelluar proteins was still relatively low at 78.9%.
This paper studies ways to improve the accuracy for predicting extracellular localization in the Gram-negative bacteria. We explored a new way to extract features from protein sequences for protein localization prediction by clustering 20 amino acids into a few groups using a greedy algorithm. Our method for clustering 20 amino acids considers the factors of both amino acids' physical-chemical properties and their contextual correlations. In contrast, the method presented by Yu et al. classifies the 20 amino acids into 4 groups (charged, polar, aromatic and nonpolar) based on physical-chemical properties of amino acids alone. Instead of simply combining multiple SVMs to give a better prediction, we propose a selection score function and a greedy algorithm to select a subset of SVMs to maximize the prediction accuracy.
Based on the proposed approaches, we have developed a system called P-CLASSIFIER for predicting the subcellular localization of Gram-negative bacteria by using a combination of multiple support vector machines. This has resulted in an improvement in the recall for extracellular proteins from 78.9% in CELLO [25] (currently the best predicting system for Gram-negative bacteria) to 86.0% in P-CLASSIFIER. The overall recall of P-CLASSIFIER reaches 89.8%. To the best of our knowledge, these are the most accurate results for predicting protein subcellular localization in Gram-negative bacteria.
Results
The dataset used in this study is from [24] and was extracted from SWISS-PROT release 40.29 [26]. It contains 1441 proteins of experimentally determined localization, where 1302 proteins are resident at a single localization site and 139 proteins are resident at dual localization sites. Table 1 lists the number of protein sequences from different sites in the data set.
Table 1 Number of protein sequences in different sites
Localization sites No.
cytoplasmic 248
inner membrane 268
periplasmic 244
outmembrane 352
extracellular 190
cytoplasmic / inner membrane 14
membrane / periplasmic 49
outer membrane / extracellular 76
All sites 1441
The prediction performance of our prediction system is estimated from a 5-fold cross-validation where the given training samples are randomly partitioned into 5 mutually exclusive sets of approximately equal size and approximately equal class distribution.
It is observed that there are some protein sequences in the dataset containing character "X". To avoid possible noise from ambiguous information, the protein entries containing "X" in the protein sequence are excluded in the cross-validation training set, but included in the testing set in this work.
Table 2 shows the prediction recall for single localization. The recall is calculated as TPx / (TPx + FNx), where TPx and FNx represent true positives (number of samples correctly classified as X) and false negatives (number of samples classified as not X that are actually X) over the predictive site X.
Table 2 Prediction recall for a single localization.
Localization Recall (TPx/(TPx+FNx))
Cytoplasmic 94.8% (235 / 248)
Extracellular 83.2% (158 / 190)
Innermembrane 88.1% (236 / 268)
Outermembrane 93.2% (328 / 352)
Periplasmic 86.9% (212 / 244)
Overall recall 89.8% (1169/1302)
In the dataset, some proteins occur in two different subcellular localizations. Since we are comparing our combined classifier P-CLASSIFIER with the P-SORTB and CELLO classifiers, we followed their method in evaluating the classifier for proteins resident at dual localization sites, where we consider them as predicted correctly if one of their localization sites is predicted correctly. Table 3 shows the prediction recall for dual localizations.
Table 3 Prediction recall for dual localizations.
Localization Recall (TPx/(TPx+FNx))
Cytoplasmic / innermembrane 92.9% (13/14)
Outermembrane / extracellular 98.9% (75/76)
Periplasmic / innermembrane 75.5% (37/49)
Overall recall 89.9% (125/139)
The Matthews correlation coefficient [27] is used to measure the predictive performance for five predictive sites. The Mattews correlation coefficient (MCC) is defined by:
where TPx, TNx, FPx, and FNx are true positives, true negatives (the number of samples correctly predicted as not X that are actually not X), false positives (the number of samples incorrectly predicted as X that are actually not X), and false negatives of localization site X, respectively. MCC offers a comprehensive and robust measurement for the predictive performance as this measurement considers both under-and over-predictions. The value of MCC equals 1 for a perfect prediction, and 0 for a completely random assignment.
Table 4 lists the performance comparisons among P-CLASSIFIER's (our system), PSORT-B's, and CELLO's [25] systems. As shown in Table 4, the values of MCC of all five sites in our system is greater than or equal to the values in CELLO's system, currently the best predicting system for Gram-negative bacteria. Moreover, we increase the recall for the extracellular site from 78.9% in CELLO to 86.0% in P-CLASSIFIER, a significant improvement for the extracellular site on the previous results. The overall recall of P-CLASSIFIER reaches 89.8%, which is better than previous results. To the best of our knowledge, these are the most accurate results for predicting Gram-negative bacteria localization.
Table 4 Performance comparisons among P-CLASSIFIER's, PSORT-B's, and CELLO's methods.
P-CLASSIFIER CELLO PSORT-B
Localization Recall MCC Recall MCC Recall MCC
Cytoplasmic 94.6% 0.85 90.7% 0.85 69.4% 0.79
Extracellular 86.0% 0.89 78.9% 0.82 70.0% 0.79
Innermembrane 87.1% 0.92 88.4% 0.92 78.7% 0.85
Outermembrane 93.6% 0.90 94.6% 0.90 90.3% 0.93
Periplasmic 85.9% 0.81 86.9% 0.80 57.6% 0.69
Overall recall 89.8% - 88.9% - 74.8% -
Discussion
To computationally analyse protein data, the representation of protein sequences is an important issue. A good input representation makes it easier for the SVM to identify underlying regularities and therefore is crucial to the success of SVM learning.
In this paper, we encode protein sequences by using the patterns of one amino acid, two adjacent amino acids, three adjacent amino acids, and four adjacent amino acids.
As there are 8000 and 160000 different patterns for the three and four adjacent amino acids cases, clustering 20 amino acids into several groups provides a way to reduce the number of unique patterns since it is difficult to train the SVM with very large number of features such as 160000 for all possible patterns of four adjacent amino acids. Since amino acids in proteins do not contribute to the function of proteins independently and functional patterns in proteins are embedded as sequence correlations, amino acids may not be grouped based on their physical-chemical properties alone [28]. For the prediction task, a good amino acid grouping leads to an improvement in prediction performance.
It is observed that the prediction results from SVMs constructed by different lengths of adjacent amino acid patterns, e.g. the patterns of a single amino acid and amino acid pairs, are complementary. That is, there are some cases where the prediction made by the SVM constructed by patterns of some particular length is correct while the prediction made by the SVM constructed by patterns of another length is incorrect, and vice versa. Therefore, combining complementary results provides a way to improve the prediction accuracy. However, combining all complementary results together may not be a good choice. Therefore, we propose to choose a subset of complementary support vector machines properly that will maximize the prediction accuracy.
After analyzing the predictive results, it is observed that there are some protein sequences that cannot be predicted correctly by any SVM in the combined classifier. It means that these protein sequences cannot be correctly classified by their composition. This is the reason why the recall of some predictive sites in Gram-negative bacteria cannot be further improved.
Since we are comparing our combined classifier P-CLASSIFIER with the P-SORTB and CELLO classifiers, we use the same data set as theirs. We did not check the sequence redundancy in the dataset. As the level of sequence redundancy normally strongly affects prediction accuracy, removing those protein sequences which have high sequence identity (e.g. more than 40%) with each other in the dataset can avoid redundancy and bias.
Instead of giving full credit for dual-localized proteins if either of the sites is predicted correctly, we also evaluate the prediction performance by counting "half" correct when only one of the sites of dual-localized proteins is predicted correctly. Table 5 shows their prediction recalls. The full credit for dual-localized proteins is only given when two possible localization sites with the top two associated probability scores match to actual dual localizations of the protein. The corresponding overall recall for predicting dual localizations only reaches 67.3%. To properly deal with subcellular localizations for proteins resident in several different sites is a challenging problem. The paper [5] addressed the problem of subcellular localizations for proteins resident in several different sites.
Table 5 Prediction recall for dual localizations when "half" predictions are only counted as half correct.
Localization Recall (TPx/(TPx+FNx))
Cytoplasmic / innermembrane 75.0% (10.5/14)
Outermembrane / extracellular 84.2% (64/76)
Periplasmic / innermembrane 38.8% (19/49)
Overall recall 67.3% (93.5/139)
There are three methods used for cross-validation test: the independent dataset test, n-fold cross-validation test, and the leave one out cross-validation test. Among these methods, the leave one out cross-validation test is the most rigorous and objective [29,42]. However, the leave one out cross validation test is very expensive computationally and is often impractical for large datasets. The n-fold cross-validation test provides a bias-free estimation of the accuracy [30] at much reduced computational cost and is considered as an acceptable test for evaluating predictive performance of an algorithm [31] for large datasets.
Conclusion
This paper introduces a protein subcellular localization prediction method using amino acid subalphabets and a combination of multiple support vector machines.
The main contributions of our work include: (1) A new way to extract features from protein sequences by clustering 20 amino acids into a few groups using the proposed greedy algorithm to reduce the input dimensionality of support vector machines. Our method for clustering 20 amino acids considers not only the factor of the amino acids' physical-chemical properties but also the factor of their contextual correlations. (2) A selection score function and a greedy algorithm are proposed to select a subset of candidate support vector machines to maximize the cross-validation accuracy instead of simply combining multiple support vector machines to give better prediction. (3) A web-based system has been developed for predicting protein subcellular localization of Gram-negative bacteria. It allows people to submit multiple Gram-negative bacteria protein sequences to perform protein subcellular localization prediction. It is available at [43].
Clustering 20 amino acids into a few groups by our proposed greedy algorithm provides a new way to extract features to cover more adjacent amino acids from protein sequences and reduce the dimensionality of these features. Since amino acids in proteins do not contribute to the function of proteins independently, it may not be a good idea to group amino acids based on their physical-chemical properties alone. For the prediction task, a good amino acid grouping leads to an increase in prediction performance. Furthermore, properly choosing a subset of complementary support vector machines constructed by different features of proteins maximizes the prediction accuracy.
Methods
Support vector machines
Support Vector Machines (SVMs) have been widely used in the analysis of biological data [32-34]. SVM is a relatively new family of learning methods and has some theoretical support from statistical learning theory [35,36]. SVM non-linearly maps the input space into a high dimensional feature space, and seeks a hyperplane in this space that separates the positive samples from the negative ones with the largest possible margin and optimizes the trade-off between good classification and large margin. Instead of explicitly mapping the objects to the high dimensional feature space, SVM usually works implicitly in the feature space by only computing the corresponding kernel between any two objects.
Several parameters need to be set during the SVM training phase. These parameters include the regularization parameter, which controls the trade-off between good classification and large margin, the kernel type, and the kernel parameters. These parameters are tuned based on the criteria of cross-validation accuracy. The radial basis function (RBF) kernel is used for all our experiments and the software BSVM [44], a multi-class SVM [37], is used in this work.
Protein features
The amino acid compositions in the full or partial sequences are considered as global features, which represent the overall similarity among multiple protein sequences. In this paper, the global features are used as the input of the SVMs to predict protein subcellular localization.
a. W-gram protein encoding
Two types of features are considered in our work: W-gram and gapped 2-gram. A W-gram is defined as patterns of W (W ≥ 1) consecutive amino acid residues without any gaps and a gapped 2-gram is defined as two amino acid residues with some specified number of gaps in a protein sequence. Here, a gapped 2-gram is also referred to as a 2-gram. The main purpose of introducing the gapped encoding features for 2-gram is to increase the number of 2-gram feature candidates.
For each protein sequence P and each W-gram (or feature) F, let N(P, F) be the number of occurrences of F in the protein sequence P. Further, let T(P, W) be the total number of possible W-grams in P, length(P) be the length of P, and G(F) be the specified number of gaps. We have T(P, W) = length(P) - W + 1 - G(F), where G(F) = 0 if W ≠ 2 and G(F) ≥ 0 if W = 2. The feature value U(P, F) with respect to the feature F and the sequence P is defined as N(P, F) / T(P, W). For example, suppose P = "LAEVLAAA" and F = "LA" (without any gaps), then the feature value U(P, F) is 2 / (8 - 2 + 1 - 0) = 0.2857, where F = "LA", N(P, F) = 2, length(P) = 8, W = 2, G(F) = 0, and T(P, W) = 7. Intuitively, U(P, F) measures the proportional occurrences of F among all possible W-grams in P. This measurement is length independent.
In the W-gram protein encoding method, the total number of different possible features is 20w.
b. Amino acid subalphabets
It is difficult to train the SVM with very large number of features such as 8000 for 3-gram. To reduce dimensionality, one way is to classify the 20 amino acids into small number of groups based on their physical-chemical properties. All members in the same group can be represented by one symbol. The merged amino acid alphabet has fewer than 20 symbols and is called the amino acid subalphabet, which can be used to re-encode the original protein sequences. The re-encoded protein sequences have fewer features. For example, if the number of symbols in an alphabet is reduced from 20 to 6, the number of 3-gram features is reduced from 4000 (20 × 20 × 20) to 216 (6 × 6 × 6). Reducing the number of features to a manageable size for SVMs can help to improve the predictive performance.
This paper suggests optimizing the grouping by using the proposed greedy algorithm, which considers the factors of both the amino acids' physical-chemical properties and their contextual correlations, instead of using the grouping based on their physical-chemical properties alone. Note that there are an exponential number of ways to group the 20 amino acids. For example, there are 580606446 and 45232115901 ways to divide 20 amino acids into 3 and 4 groups, respectively. The number of subalphabets with m groups (1 ≤ m ≤ 20) for the protein alphabet size of 20, N(m) can be calculated by the formula [28] below.
We learn the local optimal grouping based on a greedy algorithm using the SVM classification algorithm to evaluate the fitness of each candidate subalphabet, where the criteria for evaluation is the 5-fold cross-validation accuracy.
c. Search for amino acid subalphabets
This section presents our greedy algorithm for finding a good grouping for the amino acids. Given a particular subalphabet encoding schema S, supposing Ng and Tc are the predefined number of groups and threshold of cross-validation accuracy, respectively. Further, we assume the parameters of a SVM to evaluate the fitness of a candidate subalphabet are given. These SVM parameters can be set either by the values suggested by the SVM software or by the tuning result of the SVM, which is constructed from features re-encoded by grouping 20 amino acids based on their physical-chemical properties, according to the criteria of cross-validation accuracy. For a particular subalphabet encoding schema S, let the grouping score h(S) be the cross-validation accuracy when prediction is done by a SVM using W-gram and the subalphabet scheme S. h(S) can be used to measure the goodness of the grouping S.
Table 6 shows an example of clustering 20 amino acids into 4 groups for the 4-gram protein encoding method using the proposed greedy algorithm. The initial node with 4-group assignment is set to {(A, G, I, L, M, P, V), (C, N, Q, S, T), (D, E, K, H, R), (F, W, Y)}, which is based on the physical-chemical property of amino acids. The process for searching for an amino acid subalphabet is iterated until it reaches a local maximal grouping score at 79.0285%, where the final four groups are {(I, L, M, V), (N, S, T), (C, D, E, H, K, Q, R, Y), (A, F, G, P, W)}. Note that some group members in the classified result have the same physical-chemical property of amino acids. For example, the amino acids A, F, G and W in the fourth group (A, F, G, P, W) are all hydrophobic. In particular, the amino acids F and W are aromatic while amino acids A and G are tiny. Further, the hydrophilicity scale indices of A, G, P, and W have approximately the same values in the amino acid index database [38], which suggests that the hydrophilicity of amino acids may be an important property in classifying the 20 amino acids.
Table 6 An example of clustering 20 amino acids into 4 groups.
Searching states Cross-validation accuracy Actions
(A, G, I, L, M, P, V)
(C, N, Q, S, T)
(D, E, H, K, R)
(F, W, Y) 71.2413% Move 'G' from group 1 to group 4
(A, I, L, M, P, V)
(C, N, Q, S, T)
(D, E, H, K, R)
(F, G, W, Y) 74.0941% Move 'A' from group 1 to group 4
(I, L, M, P, V)
(C, N, Q, S, T)
(D, E, H, K, R)
(A, F, G, W, Y) 75.9445% Move 'P' from group 1 to group 4
(I, L, M, V)
(C, N, Q, S, T)
(D, E, H, K, R)
(A, F, G, P, W, Y) 77.5636% Move 'C' from group 2 to group 3
(I, L, M, V)
(N, Q, S, T)
(C, D, E, H, K, R)
(A, F, G, P, W, Y) 78.4888% Move 'Q' from group 2 to group 3
(I, L, M, V)
(N, S, T)
(C, D, E, H, K, Q, R)
(A, F, G, P, W, Y) 78.9514% Move 'Y' from group 4 to group 3
(I, L, M, V)
(N, S, T)
(C, D, E, H, K, Q, R, Y)
(A, F, G, P, W) 79.0285% Reach local maximal grouping score and stop.
The proposed greedy algorithm to search for amino acid subalphabets is described in Table 7. The greedy local search [39] has been used for learning the subalphabets. In the search tree [39], every node represents an amino acid subalphabet encoding schema. The child nodes of a node are subalphabets encoding schemata, which are generated by moving every group member to each other group if the number of members in this group is greater than one.
Table 7 Algorithm for amino acid subalphabets searching
1 current_node ← the initial group assignment by dividing the 20 amino acids into Ng groups.
2 REPEAT
3 best_node ← current_node
4 REPEAT
5 current_node ← best_node
6 generate all child nodes of the current node in the search tree.
7 best_node ← the child node with the highest h-value among all child nodes of the current node.
8 UNTIL h(best_node) <h(current_node)
9 IF h(current_node) < Tc THEN
10 current_node ← randomly re-generate initial group assignment
11 ENDIF
12 UNTIL h(current_node) ≥ Tc
This algorithm is composed of the following four steps. First, the 20 amino acids are initially divided into Ng groups either randomly with approximately the same size or based on some physical-chemical properties of the 20 amino acids. Amino acids in the same group are denoted by one symbol in a subalphabet. Suppose the current subalphabet encoding schema is represented by current node, its grouping score is calculated where the grouping score is the cross-validation accuracy when prediction is done by a SVM using W-gram and this subalphabet scheme.
Second, all child nodes of the current node are generated. If there is only one member in some group, this member cannot move to any other group. Otherwise, the total number of groups will be less than Ng. There are at most 20 × (Ng - 1) possible child nodes in the searching space since there are 20 amino acids and each amino acid can only move to at most (Ng - 1) other groups. If the highest grouping score among the child nodes is greater than the grouping score of the current node, this child node will become the current node.
Third, the above process for searching the child node with the highest grouping score among all child nodes will be repeated until the grouping scores of all child nodes are less than the grouping score of the current node.
Fourth, if the grouping score in the final current node is greater than Tc, the Ng groups in the current node will become the accepted merged subalphabets. Otherwise, we randomly re-generate the current node and repeat the Steps 2 to 4 above.
The training sequences are divided into two parts: One part is used for choosing the subalphabet while the other is used for evaluating the performance of a subalphabet.
The greedy algorithm is applied to reduce the number of W-gram features. In particular, for 3-gram, we classify the 20 amino acids into 6, 7, and 8 groups. For 4-gram, we classify the 20 amino acids into 4 groups. The number of features is mW, where m is the number of groups and W is the number of protein peptides in W-gram encoding methods. For example, the number of features is 6 × 6 × 6 = 216 for 6 groups in 3-gram encoding method.
Multiple SVMs
Due to the nature of the multi-class classification, it may not be easy to obtain a single SVM that can return high accuracies for the subcellular localization prediction. Therefore, multiple SVMs are trained from different features and their results are combined using voting.
Currently most of the existing protein subcellular localization prediction systems using SVMs only use the features generated from 1-gram or 2-gram protein encoding methods. For example, the extracted features of amino acid compositions [2] and features of amino acid pair and gapped amino acid pair compositions [40] can be considered as the features generated from the 1-gram and 2-gram encoding methods, respectively.
As many functional patterns in proteins are embedded as sequence correlations, it is expected that more information will be included by combining classifiers constructed from features generated by 1-gram, 2-gram, 3-gram, and 4-gram protein encoding methods, instead of just using the classifiers constructed from 1-gram and 2-gram encoding methods since more adjacent amino acid residues will be considered.
In this paper, the following four types of features are extracted from protein sequences. The first type is the 1-gram encoding feature, which includes amino acid compositions and the partitioned amino acid compositions, where the protein sequence is partitioned into P parts with approximately the same length. The total number of these features is 20 × P. In this work, P is set from 2 to 6. The second one is 2-gram encoding feature, which includes amino acid pair and gapped amino acid pair compositions, where the number of features is 400 (20 × 20) and the number of gaps is set from 1 to 2. The purpose of introducing the gapped encoding features only for 2-gram is to increase the number of 2-gram feature candidates. The third one is the 3-gram encoding feature, where the 20 amino acids are divided into 6, 7, and 8 groups whose numbers of features are 216 (6 × 6 × 6), 343 (7 × 7 × 7), and 512 (8 × 8 × 8), respectively. The last one is the 4-gram encoding method, where the 20 amino acids are divided into 4 groups, whose number of features is 256 (4 × 4 × 4 × 4).
Feature selection
We apply the wrapper approach [41] in the backward elimination version to select the feature subset for our SVM classifiers and use 5-fold cross-validation accuracy as the criteria for evaluation.
Let SVMa and SVMb be the SVM classifiers using all features and features selected by the wrapper approach, respectively. Although the prediction accuracy of SVMb is improved, the prediction results from SVMa and SVMb are different. There are some cases where the prediction made by SVMa is correct while the prediction made by SVMb is not correct, and vice versa. Therefore, both SVMa and SVMb can be considered as candidates to build the final combined classifier.
SVM subset selection
Different SVMs give different predictions. One way to combine their predictions is by voting. That is, each protein sequence is assigned to a class with the most votes. For cases where two or more classes get the most votes, we assign these cases to the predictive results by one of the SVMs, which gets the most number of correct predictions for all these cases.
Suppose S is a set of protein sequences, N is the number of candidate SVMs, M = {SVM1, SVM2, ..., SVMN} is the set of candidate SVMs defined previously, V1(S, M) is the number of correct predictions classified by M with only one class corresponding to the most vote, and V2(S, M) is the number of the correct predictions by the assigned SVM when two or more classes correspond to the most vote. The selection score function V(S, M) is defined as V1(S, M) + V2(S, M) and is used to select a subset of all candidate SVMs to form a combined classifier, which maximizes the cross-validation accuracy. The proposed greedy algorithm to select a subset of M is described in Table 8.
Table 8 Algorithm for SVM subset selection
1 Let M = {SVM1, SVM2, ..., SVMN} be the set of candidate SVMs
2 Let Scoremax = V(S, M) and Setmax = M
3 FOR i = N-1 to 1
4 Vmax = max{V(S, M - {SVMr}) | SVMr ∈ M, 1 ≤ r ≤ N }
5 IF V(S, M - {SVMj}) == Vmax (1 ≤ j ≤ N) THEN
6 M = M - {SVMj}
7 ENDIF
8 IF Vmax ≥ Scoremax THEN
9 Scoremax = Vmax
10 Setmax = M
11 ENDIF
12 END FOR
This greedy algorithm consists of the following two steps. First, set M = {SVM1, SVM2, ..., SVMN}, Scoremax = V(S, M), Setmax = M, and i = N - 1. Second, for every member SVMr ∈ M (1 ≤ r ≤ N), remove SVMr from M and calculate the value of its corresponding selection score function V(S, M - {SVMr}) (1 ≤ r ≤ N). Suppose for some SVMj (1 ≤ j ≤ N), V(S, M - {SVMj) is equal to Vmax, the maximal value of all V(S, M - {SVMr) (1 ≤ r ≤ N), then update the following: M = M - {SVMj, Scoremax = Vmax, Setmax = M, and i = i - 1.The process for removing some SVMp (1 ≤ p ≤ N) will continue until i = 1, that is, only one SVM is left. Then Setmax is selected to be the combined classifier.
We can use the prediction results of four-fifth training protein sequences to select a subset of SVMs and use the prediction results of the rest of one-fifth training protein sequences to evaluate the performance of the result of the SVM subset selection.
In this work, 15 SVMs are selected and combined to form the final classifier. Table 9 shows the encoding methods of input vectors in the fifteen selected SVMs. Rows 12, 13, and 14 represent 3 different merged subalphabets, which are {(A, F, G, P, W), (C, D, E, H, K, Q, R, Y), (N, S, T), (I, L, M, V)}, {(A, C, M, P, V), (F, I, L, W), (D, E, H, Q, R), (G, K, N, S, T, Y)}, and {(A, G, P, Q, Y), (C, D, E, H, K, M, R), (N, S, T), (F, I, L, M, V)}, respectively. Rows 4, 7 and 15 represent the same encoding method as the rows 3, 6 and 14 but with feature selection.
Table 9 The encoding methods of input vectors in the fifteen selected SVMs.
No. Encoding methods of input vectors
1 1-gram with 2 partitioned parts
2 1-gram with 3 partitioned parts
3 1-gram with 4 partitioned parts
4 1-gram with 4 partitioned parts (apply feature selection to No. 3)
5 1-gram with 6 partitioned parts
6 2-gram without any gaps
7 2-gram without any gaps (apply feature selection to No. 6)
8 2-gram with one gap
9 3-gram with 6 merged groups
10 3-gram with 7 merged groups
11 3-gram with 8 merged groups
12 4-gram with 4 merged groups
13 4-gram with 4 merged groups
14 4-gram with 4 merged groups
15 4-gram with 4 merged groups (apply feature selection to No. 14)
We have conducted some experiments on constructing SVMs by using 5-gram encoding method. Preliminary experimental results show that the cross-validation accuracies predicted by SVM constructed by 3-gram, 4-gram, and 5-gram encoding methods are not satisfactory when the number of groups is less than 6, 4, and 4, respectively. When we increase the number of groups to 4 for 5-gram, the time required to train the corresponding SVM and calculate the 5-fold cross validation accuracy is relatively slow as the number of features reaches 1024 (4 × 4 × 4 × 4 × 4). Therefore, only 1-gram, 2-gram, 3-gram, and 4-gram encoding methods are considered in this paper. Furthermore, the 20 amino acids are classified into 6, 7, and 8 groups for 3-gram and 4 groups for 4-gram encoding methods, respectively.
Since there are too many zero elements in the encoding results, 2-gram, 3-gram, and 4-gram protein's encoding methods are not applied to those cases where the protein sequences are partitioned into P (P > 1) parts with approximately same length.
Authors' contributions
JW developed the methods, built the system and drafted the manuscript. WS, AK and KL participated in system design, provided valuable comments, and helped to draft the manuscript.
Acknowledgements
The authors would like to thank the anonymous reviewers whose comments have helped us improve the manuscript.
==== Refs
Emanuelsson O Nielsen H Brunak S von Heijne G Predicting subcellular localization of proteins based on their N-terminal amino acid sequence J Mol Biol 2000 300 1005 1016 10891285 10.1006/jmbi.2000.3903
Hua S Sun Z Support vector machine approach for protein subcellular localization prediction Bioinformatics 2001 17 721 728 11524373 10.1093/bioinformatics/17.8.721
Horton P Nakai K Better prediction of protein cellular localization sites with the k nearest neighbors classifier Proc Int Conf Intell Syst Mol Biol 1997 5 147 152 9322029
Nakashima H Nishikawa K Discrimination of Intracellular and Extracellular Proteins Using Amino Acid Composition and Residue-pair Frequencies J Mol Biol 1994 238 54 61 8145256 10.1006/jmbi.1994.1267
Cai YD Chou KC Predicting 22 protein localizations in budding yeast Biochem Biophys Res Commun 2004 323 425 428 15369769 10.1016/j.bbrc.2004.08.113
Chou KC Prediction of protein cellular attributes using pseudo-amino acid composition PROTEINS: Proteins 2001 43 246 255 (Erratum: ibid., 2001, 44: 60) 10.1002/prot.1035
Chou KC Cai YD A new hybrid approach to predict subcellular localization of proteins by incorporating Gene ontology Biochem Biophys Res Commun 2003 311 743 747 14623335 10.1016/j.bbrc.2003.10.062
Chou KC Cai YD Prediction and classification of protein subcellular localization: sequence-order effect and pseudo amino acid composition Journal of Cellular Biochemistry 2003 90 1250 1260 (Addendum, ibid. 2004, 91(5): 1085) 14635197 10.1002/jcb.10719
Chou KC Cai YD Predicting subcellular localization of proteins by hybridizing functional domain composition and pseudo-amino acid composition Journal of Cellular Biochemistry 2004 91 1197 1203 15048874 10.1002/jcb.10790
Chou KC Cai YD Prediction of protein subcellular locations by GO-FunD-PseAA predictor Bioch Biophys Res Commun 2004 320 1236 1239 10.1016/j.bbrc.2004.06.073
Feng ZP Prediction of the subcellular location of prokaryotic proteins based on a new representation of the amino acid composition Biopolymers 2001 58 491 499 11241220 10.1002/1097-0282(20010415)58:5<491::AID-BIP1024>3.0.CO;2-I
Feng ZP Zhang CT Prediction of membrane protein types based on the hydrophobic index of amino acids Journal of Protein Chemistry 2000 19 269 275 11043931 10.1023/A:1007091128394
Feng ZP Zhang CT Prediction of the subcellular location of prokaryotic proteins based on the hydrophobicity index of amino acids Int J Biol Macromol 2001 28 255 261 11251233 10.1016/S0141-8130(01)00121-0
Pan YX Zhang ZZ Guo ZM Feng GY Huang ZD He L Application of pseudo amino acid composition for predicting protein subcellular location: stochastic signal processing approach Journal of Protein Chemistry 2003 22 395 402 13678304 10.1023/A:1025350409648
Wang M Yang J Liu GP Xu ZJ Chou KC Weighted-support vector machines for predicting membrane protein types based on pseudo amino acid composition Protein Eng Des Sel 2004 17 509 516 15314209 10.1093/protein/gzh061
Wang M Yang J Xu ZJ Chou KC SLLE for predicting membrane protein types J Theor Biol 2005 232 7 15 15498588 10.1016/j.jtbi.2004.07.023
Xiao X Shao S Ding Y Huang Z Huang Y Chou KC Using complexity measure factor to predict protein subcellular location Amino Acids 2005 28 57 61 15611847 10.1007/s00726-004-0148-7
Yuan Z Prediction of protein subcellular locations using Markov chain models FEBS Letters 1999 451 23 26 10356977 10.1016/S0014-5793(99)00506-2
Zhou GP An intriguing controversy over protein structural class prediction Journal of Protein Chemistry 1998 17 729 738 9988519 10.1023/A:1020713915365
Zhou GP Assa-Munt N Some insights into protein structural class prediction 2001 44 57 59 11354006 10.1002/prot.1071
Zhou GP Doctor K Subcellular location prediction of apoptosis proteins Proteins 2003 50 44 48 12471598 10.1002/prot.10251
Nakai K Protein sorting signals and prediction of subcellular localization Adv Protein Chem 2000 54 277 344 10829231
Nakai K Kanehisa M Expert system for predicting protein localization sites in Gram-negative bacteria Proteins 1991 11 95 110 1946347 10.1002/prot.340110203
Jennifer GardyL Cory Spencer Ke Wang Martin Ester Gabor TusnadyE Istvan Simon Sujun Hua Katalin deFays Christophe Lambert Kenta Nakai Fiona BrinkmanSL PSORT-B: improving protein subcellular localization prediction for Gram-negative bacteria Nucleic Acids Research 2003 31 3613 17 12824378 10.1093/nar/gkg602
Yu C-S Lin C-J Hwang J-K Predicting subcellular localization of proteins for Gram-negative bacteria by support vector machines based on n-peptide compositions Protein Science 2004 13 1402 1406 15096640 10.1110/ps.03479604
Bairoch A Apweiler R The SWISS-PROT protein sequence database and its supplement TrEMBL in 2000 Nucleic Acids Research 2000 28 45 48 10592178 10.1093/nar/28.1.45
Matthews BW Comparison of the predicted and observed secondary structure of T4 phage lysozyme Biochim Biophys Acta 1975 405 442 451 1180967
Andersen CAF Brunak S Representation of protein-sequence information by amino acid subalphabets AI Magazine 2004 25 97 104
Mardia KV Kent JT Bibby JM Multivariate Analysis 1979 London: Academic Press 322 381
Stone M Cross-validatory choice and assessment of statistical predictions Journal of the Royal Statistical Society 1974 36 111 147
Kohavi R Wrappers for performance enhancement and oblivious decision graphs PhD thesis 1995 Stanford University
Brown MPS Grundy WN Lin D Cristianini N Sugnet CW Furey TS Ares M Haussler D Knowledge-based analysis of microarray gene expression data by using support vector machines PNAS 2000 97 262 267 10618406 10.1073/pnas.97.1.262
Lee Y Lee C-K Classification of multiple cancer types by multicategory support vector machines using gene expression data Bioinformatics 2003 19 1132 1139 12801874 10.1093/bioinformatics/btg102
Ward JJ McGuffin LJ Buxton BF Jones DT Secondary structure prediction with support vector machines Bioinformatics 2003 19 1650 1655 12967961 10.1093/bioinformatics/btg223
Vapnik V The nature of statistical learning theory 1995 Springer-Verlag, New York
Vapnik V Statistical learning theory 1998 John-Wiley, New York
Hsu C-W Lin C-J A simple decomposition method for support vector machines Machine Learning 2002 46 291 314 10.1023/A:1012427100071
Kawashima S Kanehisa M AAindex: amino acid index database Nucleic Acids Res 2000 28 374 10592278 10.1093/nar/28.1.374
Russel SJ Norvig P Artificial intelligence: a modern approach Prentice Hall 2003
Park K-J Kanehisa M Prediction of protein subcellular locations by support vector machines using compositions of amino acids and amino acid pairs Bioinformatics 2003 19 1656 1663 12967962 10.1093/bioinformatics/btg222
Kohavi R John GH Wrappers for feature subset selection Artificial intelligence 1997 97 273 324 10.1016/S0004-3702(97)00043-X
Chou KC Zhang CT Review: Prediction of protein structural classes Crit Rev Biochem Mol Biol 1995 30 275 349 7587280
Protein subcellular localization prediction for Gram-negative bacteria
BSVM
|
16011808
|
PMC1190155
|
CC BY
|
2021-01-04 16:27:23
|
no
|
BMC Bioinformatics. 2005 Jul 13; 6:174
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-174
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1811602949110.1186/1471-2105-6-181Methodology ArticleVector analysis as a fast and easy method to compare gene expression responses between different experimental backgrounds Breitling Rainer [email protected] Patrick [email protected] Anna [email protected] Molecular Plant Science Group, Institute of Biomedical and Life Sciences, University of Glasgow, Glasgow G12 8QQ, UK2 Bioinformatics Research Centre, Department of Computing Science, University of Glasgow, Glasgow G12 8QQ, UK2005 19 7 2005 6 181 181 4 4 2005 19 7 2005 Copyright © 2005 Breitling et al; licensee BioMed Central Ltd.2005Breitling et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Gene expression studies increasingly compare expression responses between different experimental backgrounds (genetic, physiological, or phylogenetic). By focusing on dynamic responses rather than a direct comparison of static expression levels, this type of study allows a finer dissection of primary and secondary regulatory effects in the various backgrounds. Usually, results of such experiments are presented in the form of Venn diagrams, which are intuitive and visually appealing, but lack a statistical foundation.
Results
Here we introduce Vector Analysis (VA) as a simple, yet principled, approach to comparing expression responses in different experimental backgrounds. VA enables the automatic assignment of genes to response prototypes and provides statistical significance estimates to eliminate spurious response patterns. The application of VA to a real dataset, comparing nutrient starvation responses in wild type and mutant Arabidopsis plants, reveals that consistent patterns of expression behavior are present in the data and are reliably detected by the algorithm.
Conclusion
Vector analysis is a flexible, easy-to-use technique to compare gene expression patterns in different experimental backgrounds. It compares favorably with the classical Venn diagram approach and can be implemented manually using spreadsheets, such as Excel, or automatically by using the supplied software.
==== Body
Background
Large-scale gene expression measurements by microarray technology are used to compare mRNA levels in different experimental or biological conditions [1]. However, in an increasing number of cases, it seems far more relevant to compare differences in expression responses, rather than static expression levels. Perhaps the most common situation involves the comparison between a wild type and a mutant organism. Here, the mRNA profile in any condition will differ between the two genetic backgrounds, but these differences will be a complex combination of the primary effect of the mutation and secondary effects of various kinds. E.g., the mutant may show growth defects, disease reactions, or compensating adjustments in its physiology. All of these make a direct comparison between the expression profiles problematic. In contrast, comparing how organisms of each genetic background respond to a common relevant stimulus can reveal regulatory mechanisms that are lost or gained by the mutation as well as shared or 'disregulated' responses. Of course, the same approach is useful for other studies comparing gene expression in distinct types of background, e.g. between cell lines, tissues, or even organisms. In each case, comparing dynamic responses can provide more biological insight than a static direct comparison of expression profiles.
Despite the importance of comparing expression responses in diverse backgrounds, accessible statistical techniques for this common analytical task are sorely lacking. Usually, genes that are differentially expressed in either background are first identified independently and then compared in the form of Venn diagrams that depict the overlap between the two sets of genes (see [2-5] for examples, and [6,7] for a mathematical introduction to Venn diagrams). This approach is very attractive because of its simplicity and immediate visualization. It is implemented in many commercial microarray analysis packages (e.g. Genespring) and has also been used as an alternative to clustering techniques to identify similarities between experimental results (Venn mapping, [8]) and to visualize general relationships among the functional annotations associated with lists of differentially expressed genes [9]. Venn diagrams, however, have a number of limitations, most importantly the arbitrariness of the initial definition of changed genes. In particular, the content of the intersection of the two gene sets ("shared responses") depends critically on the selection threshold used in the initial definition of differentially expressed genes. Another disadvantage is that differential responses in the two backgrounds are not further characterized, e.g. it is not obvious whether the difference of a gene's response between the two backgrounds is due to the "regulated/non-regulated" or "up-regulated/down-regulated" effect. More sophisticated statistical techniques have been used to approach this issue (e.g. ANOVA [10], Principle Component Analysis [11], Singular Value Decomposition [12], Linear Factor Models [13], or Integrative Correlation Analysis [14]). Each of these successfully addresses certain aspects of the problem, by reducing the dimensionality of the data or identifying consistent patterns of behavior across conditions. However, they all lack the intuitive appeal and simplicity of Venn diagram visualization. Here we present a simple alternative to Venn diagrams that is based on similar concepts but provides more flexibility and an added degree of objectivity of the results.
Results and discussion
The main underlying principle of our method (Vector Analysis, VA) is the idea that expression changes in two backgrounds can be represented by a vector in a Cartesian plane (Fig. 1A). Various sectors of the plane will correspond to various prototypical behaviors of genes: genes that respond the same in both backgrounds, genes that react in opposite directions, or genes that are changed only in one of the backgrounds (Fig. 1B). Like Venn diagrams, VA is not a method to detect differentially expressed genes, but rather a technique that arranges response patterns in an informative way for further study.
Figure 1 Principle of vector analysis. (A) The change in expression of a gene in the two experimental backgrounds is represented by a vector. The two axes correspond to the log-fold changes in the two backgrounds. E.g., Gene 1 is strongly up-regulated in both backgrounds, while Gene 2 is specifically down-regulated in background A, but has lost this response in background B. (B) The plane can be systematically subdivided into sectors corresponding to the main behavior types that are possible. In the centre, genes show very little response in either background (white). Other genes respond about the same in both backgrounds (blue sector), are specifically changed in only one background (yellow), or are regulated in opposite directions in background A and B (red).
If there are replicate experiments, as is generally the case in microarray studies, we calculate the representative "average" vector vREP by (1) determining the individual vectors v[i], where the vector v[i] represents the comparison of the i-th pair of experiments (if there are N replicates in background A and M replicates in background B, there will be n = N × M pairs); (2) calculating the average length of these vectors, , where |v[i]| denotes the length of the vector v[i]; (3) calculating the sum of the unit vectors pointing in the same direction as the individual pairwise vectors, ; and finally (4) determining the representative vector by combining the length (l) and direction information (vSUM), .
The length of the vector (l) indicates the average strength of the response and can be used to filter out genes that show little response in either background. The direction of the vector describes which prototypical behavior comes closest to the behavior of this particular gene. To decide on the assignment of a particular gene to a response prototype, one can calculate the angle between the representative vector and the various possible prototype vectors (e.g., or ) as cosα = vREP·vPrototype/(|vREP||vprototype|), 0 ≤ α < 180°, where vREP·vprototype is the scalar product of the two vectors and |vREP| ≠ 0. The gene is then assigned to the prototype closest to it (minimal α).
The length of the sum vector (|vSUM|) indicates the level of consistency with which the gene shows the assigned behavior type (Fig. 2). If in the individual pairwise comparisons the vectors point in widely varying directions, they will cancel out and the sum vector will be relatively short (the most probable length will approach 0 as the number of replicates increases to infinity). If, however, the behavior is fully consistent, the length of the vector will be maximal.
Figure 2 Principle of determining the consistency of the observed behavior pattern. The gene in panel (A) shows highly consistent regulation among the various pairwise comparisons of replicates. Hence the corresponding unit vectors add to a long sum vector. The gene in panel (B) is noisier and slightly inconsistent in its response pattern among replicates. Its vectors add to a shorter sum vector.
It is clear that the vector approach generalizes to multi-dimensional cases, i.e. to comparisons between more than two backgrounds. However, the number of possible prototype behaviors increases rapidly, as N = 3k - 1, where k is the number of dimensions.
By randomly sampling from the measured expression values and calculating the sum vector lengths for these random data (which should not show consistent behavior) one can estimate the null distribution of the sum vector length. This is done by randomly assigning the original expression values within each replicate to other genes. All consistency between replicates and, thus, between experimental backgrounds should then be lost and the resulting |vSUM| values will be those that are expected if no consistency is present. This can be used to assign a p-value to the assignment of genes to behavior prototypes (consistency p-value). This value, calculated by the procedure described above, will be a non-parametric estimate of the real p-value, and the exact value will vary slightly in each run of the method, unless the same random sampling is used each time.
Additional file 6 shows the results of vector analysis applied to a simulated dataset, where the response type of each gene is known [see Additional file 6]. Three replicates for each experimental background were created by drawing random expression values from normal distributions with variance 1 and a mean of 0, -2, and 2 for unchanged, down-regulated and up-regulated genes, respectively. In this small illustrative example, 87.5% of regulated genes are assigned the correct response type. The remaining genes are assigned one of the neighboring types. Genes that are unchanged in both conditions are also assigned to the closest response prototype, but none of these achieves a significant consistency p-value. Of course, in a real-world application unchanged genes would usually be filtered before applying vector analysis, because otherwise they will be assigned arbitrary angular and location values that add noise to the results. If VA is applied to genes that are not changed at all, it will always assign these genes to "incorrect" response classes, and even when the consistency p-value of VA is used, some of these genes will reach significance simply due to multiple testing. Therefore, VA is usually applied only to genes that are significantly changed in at least one experimental background, based on any of the standard methods for the detection of differentially expressed genes. However, the filtering does not have to be very strict and the results of VA may still yield interesting trends for borderline cases, as shown in the example below.
Table 1 and Fig. 3 show the results of an application of vector analysis to a real experimental dataset. The data used are a subset of a larger study examining the response of wild-type and mutant Arabidopsis thaliana plants to potassium starvation (Armengaud et al., unpublished data). The mutant plants (coi1) lack a critical component of the jasmonate signaling pathway [15], which was shown to be central for the response of plants to potassium starvation [16]. Seedling plants were grown on potassium-free agar plates for two weeks and then re-supplied with either potassium-containing or fresh deficient medium. Labeled cDNA from both conditions was prepared and analyzed on two-color whole-genome microarrays. All data were normalized by quantile normalization and log-fold changes calculated for two replicate measurements in each genetic background. A total of 1000 genes are considered in this example, which is also available as a supplementary material for further analysis.
Table 1 Number of genes showing the various types of prototypic behavior in two genetic backgrounds of Arabidopsis plants as identified by vector analysis.
Mutant specific up 162 Background-specific changes
Mutant specific down 189
WT specific up 137
WT specific down 122
WT and Mutant up 133 Same-direction changes
WT and Mutant down 131
Mutant up, WT down 54 Opposite changes
Mutant down, WT up 72
Figure 3 Vector analysis of gene expression responses in two genetic backgrounds in Arabidopsis plants. Each dot corresponds to a single sum vector from four pairwise comparisons (two replicates per background). Genes towards the periphery of the circle show the most consistent behavior among replicates. Two behavior prototypes are highlighted (corresponding to Gene 1 and 2 in Fig. 1), mutant-specific down-regulation (purple) and WT/Mutant-consistent up-regulation (orange). It can be seen that inconsistent genes (close to the center) are generally showing background-specific responses, i.e. they are enriched along the axes of the plot. Their behavior is most likely the result of spurious noise in a single replicate.
One of the properties of this dataset is that very few genes show a strong expression response in any background. Only one out of 1000 genes has an l-value larger than 1 (roughly corresponding to a two-fold expression change), and only 35 genes have l-values larger than 0.5. Thus, a Venn analysis based on significantly changed genes would be all but impossible. The vector analysis, in contrast, identifies 32 genes with consistency p-values smaller than 0.01 (expected 10) and 258 genes with p-values smaller than 0.1 (expected 100). It thus reveals the presence of consistent response patterns even among genes with very slight absolute expression changes.
Among the 19 most significant genes, with p-values < 0.1 and vector lengths > 0.5, more than half (10 out of 19) are up-regulated in both mutant and wild-type (Fig. 4). The remaining 9 genes show various background-specific responses. None of them shows an "opposite" response pattern, an observation that is highly significant (p = 0.0042). This is in agreement with the known biology of the coi1 mutant, which will lose certain regulatory mechanisms that are important in nutrient starvation, but will not to reverse existing pathways. It is also in agreement with the overall correlation between the average expression pattern in the two backgrounds (Spearman's rank correlation rs = 0.310; p < 0.001). Importantly, the same pattern is already evident in the complete dataset (Tab. 1), where genes assigned the "opposite" prototypes are clearly depleted. The presence of a detectable signal is also confirmed by the distribution of sum vector lengths in the real data compared to randomly sampled data (Fig. 5). This indicates that even for very noisy data vector analysis is able to make meaningful assignments to response patterns.
Figure 4 Expression change profile of the top 19 genes detected by vector analysis of starvation responses in wild type and mutant Arabidopsis plants. The genes have been filtered for vector lengths larger than 0.5 and p-values smaller than 0.1.
Figure 5 Sum vector length (|vSUM|) distribution for the Arabidopsis experiment and randomly permuted datasets. The real data (red) are enriched for longer sum vectors compared to random data (blue), indicating the presence of consistent response patterns.
Using the two parameters of the method (vector length = overall response intensity, and p-value = response pattern consistency) allows the flexible dissection of the observed expression in the two experimental backgrounds. At the same time it is possible to assign the most likely response pattern even to genes that show little absolute expression change.
In contrast to Venn diagrams, which can only be used to compare genes that are reliably identified as responsive, vector analysis assigns all genes to behavioral categories. Also note that these categories are not fixed, but can be adjusted as appropriate for any experiment, by simply changing the boundaries of the sectors. Also, genes can be sorted by their angular distance from any reference gene (or reference behavior), to generate lists that are sorted by closeness of genes to a particular response pattern.
Conclusion
Vector analysis provides a flexible, easy-to-use, and intuitive approach to the comparison of gene expression patterns in different experimental backgrounds. While it does not supply the detailed statistical insights available by alternative classical statistics approaches such as ANOVA, it excels in terms of simplicity and straight-forward interpretation. In this respect vector analysis compares favorably with the Venn diagram technique which is currently in wide-spread use for this common and ubiquitous task, but lacks the flexibility of vector analysis, in particular for noisy data.
Methods
For small datasets with few replicates, vector analysis is straightforward enough to be carried out manually, e.g. in Excel or OpenOffice spreadsheets. It uses only the most basic vector algebra. The Excel file in the supplementary material [see Additional file 1] demonstrates how l, vSUM, and vREP are calculated and used to automatically assign genes to the various response prototypes. A second sheet in the same file is used to randomly permute the experimental measurements by sorting them along a vector of random numbers, so that within each replicate (column) the original expression values are randomly assigned to new genes and all consistencies between columns are lost. The vector lengths calculated from these random data are then used in a third sheet to estimate the p-values associated with the observed response patterns (for details of the procedure [see Additional file 2]). For larger numbers of replicates, the manual procedure becomes quite tedious and a Perl script [see Additional file 3] is provided that performs vector analysis and p-value estimation automatically, taking a tab-delimited text file of log-fold changes in all replicates [see Additional file 4] as its input. The obtained results [see Additional file 5] can then be sorted, filtered and explored in various ways to dissect the details of comparative expression behavior.
Authors' contributions
RB devised and implemented the Vector Analysis method and drafted the manuscript. PA provided the experimental data and helped with the biological interpretation of the results. AA supervised the project. All authors read and approved the final manuscript.
Supplementary Material
Additional File 6
Output generated by vector analysis script on a set of simulated expression data with known response type for each gene.
Click here for file
Additional File 1
Excel file demonstrating the manual performance of vector analysis
Click here for file
Additional File 2
Word document describing the implementation of Vector Analysis in Excel and presenting the details of the equations used.
Click here for file
Additional File 3
Perl script performing vector analysis
Click here for file
Additional File 4
Tab-delimited text file as input file for vector analysis script
Click here for file
Additional File 5
Output generated by vector analysis script
Click here for file
Acknowledgements
This work was supported by BBSRC grants 17/G17989 and 17/P17237 to AA.
==== Refs
Lockhart DJ Winzeler EA Genomics, gene expression and DNA arrays Nature 2000 405 827 836 10866209 10.1038/35015701
Oono Y Seki M Nanjo T Narusaka M Fujita M Satoh R Satou M Sakurai T Ishida J Akiyama K Iida K Maruyama K Satoh S Yamaguchi-Shinozaki K Shinozaki K Monitoring expression profiles of Arabidopsis gene expression during rehydration process after dehydration using ca 7000 full-length cDNA microarray Plant J 2003 34 868 887 12795706 10.1046/j.1365-313X.2003.01774.x
Mariadason JM Corner GA Augenlicht LH Genetic reprogramming in pathways of colonic cell maturation induced by short chain fatty acids: comparison with trichostatin A, sulindac, and curcumin and implications for chemoprevention of colon cancer Cancer Res 2000 60 4561 4572 10969808
Irizarry RA Warren D Spencer F Kim IF Biswal S Frank BC Gabrielson E Garcia JG Geoghegan J Germino G Griffin C Hilmer SC Hoffman E Jedlicka AE Kawasaki E Martinez-Murillo F Morsberger L Lee H Petersen D Quackenbush J Scott A Wilson M Yang Y Ye SQ Yu W Multiple-laboratory comparison of microarray platforms Nat Methods 2005 2 345 350 15846361 10.1038/nmeth756
Ramalho-Santos M Yoon S Matsuzaki Y Mulligan RC Melton DA "Stemness": transcriptional profiling of embryonic and adult stem cells Science 2002 298 597 600 12228720 10.1126/science.1072530
Grünbaum B Venn Diagrams II Geombinatorics 1992 II 25 32
Grünbaum B Venn Diagrams I Geombinatorics 1992 I 5 12
Smid M Dorssers LC Jenster G Venn Mapping: clustering of heterologous microarray data based on the number of co-occurring differentially expressed genes Bioinformatics 2003 19 2065 2071 14594711 10.1093/bioinformatics/btg282
Kestler HA Muller A Gress TM Buchholz M Generalized Venn diagrams: a new method of visualizing complex genetic set relations Bioinformatics 2005 21 1592 1595 15572472 10.1093/bioinformatics/bti169
Pavlidis P Noble WS Analysis of strain and regional variation in gene expression in mouse brain Genome Biol 2001 2 RESEARCH0042 11597334 10.1186/gb-2001-2-10-research0042
Raychaudhuri S Stuart JM Altman RB Principal components analysis to summarize microarray experiments: application to sporulation time series Pac Symp Biocomput 2000 455 466 10902193
Alter O Brown PO Botstein D Generalized singular value decomposition for comparative analysis of genome-scale expression data sets of two different organisms Proc Natl Acad Sci U S A 2003 100 3351 3356 12631705 10.1073/pnas.0530258100
Girolami M Breitling R Biologically valid linear factor models of gene expression Bioinformatics 2004 20 3021 3033 15201181 10.1093/bioinformatics/bth354
Parmigiani G Garrett-Mayer ES Anbazhagan R Gabrielson E A cross-study comparison of gene expression studies for the molecular classification of lung cancer Clin Cancer Res 2004 10 2922 2927 15131026
Feng S Ma L Wang X Xie D Dinesh-Kumar SP Wei N Deng XW The COP9 signalosome interacts physically with SCF COI1 and modulates jasmonate responses Plant Cell 2003 15 1083 1094 12724535 10.1105/tpc.010207
Armengaud P Breitling R Amtmann A The potassium-dependent transcriptome of Arabidopsis reveals a prominent role of jasmonic acid in nutrient signaling Plant Physiol 2004 136 2556 2576 15347784 10.1104/pp.104.046482
|
16029491
|
PMC1190156
|
CC BY
|
2021-01-04 16:27:26
|
no
|
BMC Bioinformatics. 2005 Jul 19; 6:181
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-181
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1891604280010.1186/1471-2105-6-189SoftwareGObar: A Gene Ontology based analysis and visualization tool for gene sets Lee Jason SM [email protected] Gurpreet [email protected] Ravi [email protected] Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA2005 25 7 2005 6 189 189 18 4 2005 25 7 2005 Copyright © 2005 Lee et al; licensee BioMed Central Ltd.2005Lee et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Microarray experiments, as well as other genomic analyses, often result in large gene sets containing up to several hundred genes. The biological significance of such sets of genes is, usually, not readily apparent.
Identification of the functions of the genes in the set can help highlight features of interest. The Gene Ontology Consortium [1] has annotated genes in several model organisms using a controlled vocabulary of terms and placed the terms on a Gene Ontology (GO), which comprises three disjoint hierarchies for Molecular functions, Biological processes and Cellular locations. The annotations can be used to identify functions that are enriched in the set, but this analysis can be misleading since the underlying distribution of genes among various functions is not uniform. For example, a large number of genes in a set might be kinases just because the genome contains many kinases.
Results
We use the Gene Ontology hierarchy and the annotations to pick significant functions and pathways by comparing the distribution of functions in a given gene list against the distribution of all the genes in the genome, using the hypergeometric distribution to assign probabilities. GObar is a web-based visualizer that implements this algorithm.
The public website for GObar [2] can analyse gene lists from the yeast (S. cervisiae), fly (D. Melanogaster), mouse (M. musculus) and human (H. sapiens) genomes. It also allows visualization of the GO tree, as well as placement of a single gene on the GO hierarchy. We analyse a gene list from a genomic study of pre-mRNA splicing to demonstrate the utility of GObar.
Conclusion
GObar is freely available as a web-based tool at [2] and can help analyze and visualize gene lists from genomic analyses.
==== Body
Background
Large scale genomic studies, especially expression microarrays, routinely identify large genes sets (sometimes a few hundred or more) of interest. Researchers are faced with the problem of identifying interesting features in such datasets. A listing of gene annotations (e.g. functions, process) can help identify interesting features, but this is impractical with large sets, due to the labor involved and the difficulty in picking statistically significant trends from large datasets. Thus, a user-friendly method is required for the routine analysis of such datasets.
The Gene Ontology consortium [1] curates the annotations of genes of several model organisms using a controlled vocabulary of GO terms. It also places the GO terms on a hierarchy called the Gene Ontology(GO). There are separate hierarchies for Molecular Functions, Cellular Components and Biological Processes. The terms on the hierarchy share a parent-child relationship in which a child term is either a specific instance or a part of its parent term. The terms get more specific the lower they are on the hierarchy [3]. Each node(GO term) on the hierarchy can have multiple parents and children.
A small portion of the GO molecular function heirarchy around the nucleic-acid binding term is shown in Figure 1. This hierarchy can be generated using the GO term browser on the GObar website [2]. The GO term, GO:0003676, which corresponds to nucleic-acid binding, has two children, RNA binding(GO:0003723) and DNA binding(GO:0003677). RNA binding itself has many children, including double-stranded RNA binding(GO:0003725) and single-stranded RNA binding(GO:0003727). DNA binding has double-stranded DNA binding(GO:0003690) and single-stranded DNA binding(GO:0003697) as children.
Figure 1 A small section of the GO tree. A schematic of a small section of the molecular function branch of the GO tree around the nucleic-acid binding term. The number of D. Melanogaster genes at each node is also given, as are the GO ids and the definitions of the terms at each node. The GOTermBrowser link at the GObar website [2] allows searching for GO terms using keywords and regular expressions (such as *NA*binding) and can also draw relationship diagrams as interactive images.
A single gene can appear in several GO terms. For example, Dicer-1 (FBgn0039016 in D. Melanogaster) has several molecular functions, such as double-stranded RNA binding (GO:0003725) and bidentate ribonuclease III activity (GO:0016443) which are unrelated to each other. Figure 2 shows the placement of Dicer-1 on the GO tree.
Figure 2 Placement of D. Melanogaster Dicer-1 on the GO tree. This is generated by entering FBgn0039016 (Dicer-1 in D. Melanogaster) on the Gobar website, and turning off pruning of the tree in step 4. The leaves (nodes with no children, here shown as green ovals) are the terms associated with Dicer-1. In GObar, green ovals signify nodes that contain genes from the uploaded list, while red nodes do not contain any genes from the uploaded list. In this case they correspond to amino-acid binding (GO:0005515), bidentate ribonuclease III activity (GO:0003725) and double-stranded RNA binding (GO:0016443). The numbers on the path, which signify deviation from the expected values, are used for pruning and highlighting highly interesting nodes, but are not important when pruning has been turned off.
In Figure 1, each node shows the number of genes annotated in the D. Melanogaster genome associated with the term. There are a total of approximately 14,000 genes in the genome. If a hypothetical microarray experiment in D. Melanogaster picks out 100 significant genes, of which 10 are double-stranded RNA (dsRNA) binding genes, then it is intuitively obvious that dsRNA binding is affected by the experiment and pathways using this function might be responding to the experiment. GObar uses the hypergeometric distribution (explained below) to quantify this intuition.
Implementation
The basic idea of the algorithm is to compare the distribution of the genes from a set on the GO tree against the distribution of all the genes of the genome on the GO tree, identify and highlight branches that are improbably enriched by chance alone, and render an image of the GO tree that will allow user interactions to further explore the data.
The GO tree is first populated with all the genes in the genome, which involves placing genes at all the nodes that describe the gene. This operation is carried out only once, and is redone each time the genomic data gets updated. At each node two sets of counts are maintained, the contribution of genes from nodes that are children of the current node (distributed count) and the contribution from the genes at the current node (bare count). The total count at each node (bare count + distributed count), is divided equally amongst the distributed counts of each parent (please note that each node can have multiple parents). To calculate the counts at each node the leaves (nodes with no children) are first considered, and the counts are progressively transmitted up in a level-by-level manner, until the root is reached.
For the analysis the GO tree is populated with the list of genes to be analysed. Once again a set of distributed and bare counts is calculated at each node for this list. The distribution of these counts is compared to the genomic set and significance is assigned to the deviations from the expected counts, using the hypergeometric distribution, which is now described. If there is a collection of N objects of three types, A (count = a), B (count = b), and C (count = c) so that, N = a + b + c, then, a random selection of n objects from these N objects will contain α A objects, β B objects and γ C objects (where α + β + γ = n) with a probability given by the hypergeometric distribution
where,
This equation can be generalized to arbitrary collections of objects. Using this probability distribution, highly improbable deviations from the expected numbers are highlighted under the assumption that they are likely to have a biological significance.
Methods
GO data can be downloaded from the Gene Ontology website [1]. The data contains two sets of information that are used, the parent-child relationships for each node and the definitions of each node or term. The data collection and analysis techniques are described in this section.
Data acquisition
The downloaded GO data is used to populate one table with GO IDs and the ID definitions, and another table with a description of relationships between the GO IDs, which can use terms such as is_a or part_of to define the relationships.
In order to associate GO terms with gene IDs (accession), the files gene2go and gene2accession were retrieved from Entrez Gene [4] for the human and mouse genomes. A similar dataset for D. Melanogaster is acquired from Flybase [5]. Each gene can have multiple GO annotations, so this is a many-to-many association table.
A table, whose columns are shown in Table 1, is used to maintain node information, and to carry out statistical analysis. At each step of the methods listed below, one of the columns gets filled up. The columns in table 1 are filled in the following order,
Table 1 Columns in the GO statistics database table.
GO node statistics
GOid DC BC level num of trails up
• Level (depth in a tree): A recursive depth-first search in a bottom-up fashion is carried out to determine the level of GO terms associated with the experiment, as explained in Figure 5.
Figure 5 Result of a GObar analysis of human genes with AT-AC-U12 type splice sites. The result of a GObar analysis is an SVG (scalable vector graphics) image, with a red path signifying branches that are disproportionately over-represented in the gene list, as compared to the distribution of all the genes from the organism. Placing the mouse over a GO term pops-up a window in the figure, with information on the GO term and links to download data. The tool also allows searching for terms, as well as zooming in and out of the image. Table 2, which is a section of a table that appears in a pop-up window at the website at the end of the calculation, shows GO terms that are significantly enriched in the dataset. The numbers on each path depict the deviation over the expected count, this calculation is described in the text.
• Number of trails up: This is obtained from the table of GO ID relationships by counting the number of parents for a node.
The following two items are calculated once in the beginning for all the genes in the genome and for each analysis of a gene-list.
• BC: Bare count (BC) is a number of genes associated with each GO term (node).
• DC: Starting from the lowest node(s) in the tree (determined by the Level column), the total count, BC + DC is propagated to the node's immediate parent. If a node has more than one parent then total count is divided by the number of trails up, which is the same as the number of parents.
Populating the tree with the reference dataset
We have populated the GO tree with datasets from Entrez Gene for human and mouse data [4], Flybase [5] for fly data and SGD [6] for yeast data. In the case of Entrez Gene, two sets of maps exist, a gene id to GO map and a gene to gene id map. At the end of this process each GO node gets a list of genes. The term bare counts denotes the counts of genes at each node. The genes on children nodes also contribute to the counts on any given node, which are tracked separately and called distributed counts. Thus, the distributed count of a node is the sum of contributions of the nodes below it in the gene ontology hierarchy. Each node contributes the sum of its bare count and its distributed count equally to the distributed counts of each of its parents. This process can be recursively applied, starting from the lowest levels (or greatest depths) of the tree and working the way up the tree.
If the accounting of distributed counts is to be done properly, defining the depth of each node in the tree is important. The rule for assigning depth to each node is that, if a node gets multiple levels, then the highest depth is always assigned to it. This can be done by picking the leaves of the tree (nodes with no children) and travelling recursively all the way up to the root (node with no parents). For each path to the top, depth is assigned to each node based on the number of steps to the node from the root. If a node already has a depth assigned to it, then the depth is replaced with the current depth only if it is bigger. This is explained in Figure 5. Once the leaves have been exhausted, all the nodes in the tree will have depths assigned to them.
In order to calculate the distributed counts for each node, the list of nodes is ordered based on their depths. Starting from nodes with the highest depths the counts are propagated up, as described above, summing up the bare count and distributed count and partitioning the sum equally amongst all the parents. After exhausting the list of nodes, all the nodes should have a bare count and a distributed count assigned to them.
Populating the tree with the experimental dataset
In order to calculate probabilities for a given experimental dataset, we need to first populate the GO tree with the experimental dataset. A procedure identical to the one used in the previous section is implemented, resulting in a GO tree with just the dataset of interest on it.
Calculating the probabilities
Let BCi, DCi be the bare and distributed counts respectively at node i for the genomic dataset and let bci, dci be the bare and distributed counts respectively at node i for the experimental dataset. Then, for the Node 0 in Figure 8.
DC0 = (BC1 + DC1) + (BC2 + DC2) + (BC3 + DC3)/2 (3)
dc0 = (bc1 + dc1) + (bc2 + dc2) + (bc3 + dc3)/2 (4)
The following are defined for ease of notation:
N1 = (BC1 + DC1) (5)
N2 = (BC2 + DC2) (6)
N3 = (BC3 + DC3)/2 (7)
N0 = DC0 = N1 + N2 + N3 (8)
n1 = (bc1 + dc1) (9)
n2 = (bc2 + dc2) (10)
n3 = (bc3 + dc3)/2 (11)
n0 = dc0 = n1 + n2 + n3 (12)
(13)
Then, the probability that a dataset is a random selection from the genes in the genome is given by the Hypergeometric formula (explained above)
where
The expected value for n1 is given by
We define PD as the deviation of the counts on a node i from its expected number and is given by
We use P and PD to prune the trees, as described in the next section.
Pruning the tree
Listing all the nodes of the GO tree for a given dataset is not very informative, especially if only a few nodes are populated or if a large number of GO terms are populated by a small number of genes. This also defeats the purpose of helping users narrow down the GO terms of interest.
A node can only be pruned if every node under it also satisfies the pruning condition. The tree is pruned using the following rules to make the viewing manageable,
1. if n0 <nc, stop traversing the tree, that is, do not show anything below such a node. The population cutoff, nc can be set by the level of details option on the GObar webpage at step 4, shown in figure 3. This determines how low the population of genes in a node can go before it gets pruned. Less Detailed corresponds to a minimum of 6 genes, Detailed corresponds to a minimum of 3 genes and Very Detailed shows every node.
Figure 3 The front page of the GObar website. Selections are made for each step, and the list of genes is entered in the final step before launching the program. The pruning of the tree is controlled in step 4. A node can only be pruned if every node under it also satisfies the pruning condition. The pruning options are explained in the subsection pruning the tree. Very strict pruning might cause useful results to be thrown away, but can also highlight the best information in the dataset. In contrast, using a low stringency at this step, or no pruning, can cause too much information to be presented. Step 5 allows the nodes to be either annotated with GO ids (preferred for large trees) or definitions of GO terms. The GOTermBrowser link at the top of the page allows search for GOids using keywords or regular expressions such as "*NA*binding" to explore the GO tree neighborhood of the search term.
2. Prune nodes that have P > Pth. The threshold Pth is arbitrarily set at 0.1.
3. if ni deviates significantly up from <ni>, then the path is hightlighted using red color. PD can be set using the deviation stringency option in step 4 on the webpage, shown in figure 3. Less strict corresponds to a deviation cutoff value of 0.2, Strict corresponds to a cutoff of 0.5 and very strict corresponds to a cutoff value of 0.8.
The pruning is done starting with leaves (nodes with no children) on the tree, and stops when it reaches a node that should not be pruned according to the rules above. Fine variations of the pruning conditions are not allowed as these do not offer useful biological information and make the tool difficult to use.
Visualizing the tree and user-interaction
Graphviz [7] is used to create the layout of the GO tree, and scalable vector graphics (SVG) is used to make it interactive. An example of the visualization is shown in Figure 2. Javascript is used to animate the SVG rendering of the GO tree. When the cursor is over a node, a window pops up with information on the GO term, genes that belong to the node and links to other resources. This window can be locked in place with a mouse click, allowing further exploration of the gene.
Data download
The tool also allows the downloading of all the genes in the GO tree below any node. The downloaded list is in the form of a comma separated valued (csv) file, which contains the gene, the GO terms for each gene and a short description.
In the downloaded list, the uploaded genes are highlighted, since the list will also contain genes that belong to the nodes but are not in the uploaded list.
Use of the tool
Gobar is accessible at our website [2] and can analyze gene lists from the yeast (S. cerevisiae), fly (D. melanogaster), mouse (M. musculus) and human (H. sapiens) genomes. The front page of the website is shown in Figure 4. The list of genes to be studied can be entered into GObar by either uploading a file containing the list, or by entering the names in the text-area provided on the webpage. Each gene on the list can be annotated with user definitions, by using a colon to separate the gene name from the annotation (for example: FBgn0034246:Dicer2, FBgn0039016:Dicer1). The website offers several options to limit what is rendered in the result page, but using the default settings is recommended for the initial exploration. After the section of the GO tree relevant to the uploaded dataset is drawn, the website allows the user to limit the view to specific branches of the GO tree (either Molecular functions, Cellular components or Biological processes). Nodes containing genes from the uploaded list are rendered in a green oval. A node is red if it does not contain any genes from the uploaded list but one of its children node has genes from the list. A pop-up window also shows the GO terms and their levels in the GO heirarchy, that have been highlighted by the analysis. Taking the mouse over a GO term pops-up an information window that can by locked in place by clicking on the left mouse-button. There are links in this information window that can be used to download all the genes on and below the node on the GO tree, with the input genes highlighted in red. If user annotations are given for each gene, then they appear with the gene name in the data downloads. One of the links also allows listing all the GO terms below the node of interest on the tree. The genes in the data download that are not from the uploaded list might also be worthy of further study, especially if many of the other genes in the pathway or GO term are highlighted in the experiment.
Figure 4 GO tree depth calculation. A section of the GO tree is depicted here. The directed acyclic nature of GO is shown by the red and black trails leading to the same node. The depth of a node is its distance from the root. Thus the node for GO:0003700 at the bottom has different depths on the tree, depending on the path traversed to get to it from the root. We use the higher number (6, the greater depth) as its depth for our calculations, which are described in the text.
The numbers on the lines between the nodes are the deviation from expected counts (whose calculation is described above). A red line implies that the child node connected to the path in the GO hierarchy is over-represented in the uploaded dataset and the corresponding GO term might be biologically significant.
If a single gene (or a few genes) needs to be placed on the GO tree, then the tree should not be pruned, which is achieved by selecting the "No" option in step 4 on the front page. Pruning is the removal of branches of the GO tree which do not carry much information (either a small number of genes or results that can be explained by random selection). The result of placing Dicer-1 (FBgn0039016) in D. Melanogaster is shown in Figure 2.
Results
We will apply GObar to the analysis of results from a genomic study of splicing [8]. Splicing is the excision of introns from the pre-mRNA after transcription [9,10]. Broadly, there are two classes of splicing machinery (spliceosomes), U2-dependent and U12-dependent, named on the basis of the snRNPs involved in the splicing reaction. Each of these spliceosomes is involved in the excision of two sub-classes of introns, defined by the consensus sequences of the 5' and 3' ends of the intron, GT-AG-U2, GC-AG-U2, GT-AG-U12 and AT-AC-U12 type splice sites. The number of U12-dependent splice sites in the genome is dwarfed by the number of U2-dependent splice sites, numbering less than 2% of the total [8]. The snRNPs comprising the U12-dependent splicing machinery are also relatively less abundant. The D. Melanogaster, mouse and human genomes have U12-dependent splicing, while C. elegans seems to have lost it. This conservation leads to speculation regarding the reason for the persistence of the U12-dependent splice sites over evolutionary time scales.
If the U12-dependent splice sites persist for some biological reason, then it seems reasonable to assume that only genes with roles in certain functions should contain these splice sites. There has been some speculation, but no rigorous assessment, regarding the functional bias of these genes [9,10]. A comprehensive annotated collection of splice sites for the worm (C. elegans), fly (D. melanogaster), mouse (M. musculus) and human (H. sapiens) genomes has been generated by classifying the known sites with the help of human curation [8]. The list of genes with AT-AC-U12 type splice sites was analyzed using GObar to identify functional themes that might be highlighted. Figure 5 shows a part of the result of GObar analysis of the human genes with AT-AC-U12 type splice sites. Table 2 shows the GO terms that are highlighted by the analysis of the human set.
Table 2 Significant GO terms in the human AT-AC-U12 set, which accompanies the analysis shown in Figure 5. We show only the more specific terms in the table that appears as a pop-up webpage along with the GO tree shown in Figure 5. The level is the depth from the root, and its calculation is described in Figure 4.
Cellular component
GOid Function Type Level
GO:0030176 integral to endoplasmic reticulum membrane component 7
GO:0001518 voltage-gated sodium channel complex component 6
GO:0005891 voltage-gated calcium channel complex component 6
GO:0008023 transcription elongation factor complex component 6
GO:0005667 transcription factor complex component 6
Molecular Function
GO:0008332 low voltage-gated calcium channel activity function 9
GO:0005248 voltage-gated sodium channel activity function 8
GO:0005245 voltage-gated calcium channel activity function 8
GO:0005262 calcium channel activity function 7
GO:0005272 sodium channel activity function 7
Biological Process
GO:0006814 sodium ion transport process 9
GO:0006816 calcium ion transport process 9
GO:0030029 actin filament-based process process 9
In the molecular function branch, analysis of the mouse set highlighted GTPase regulator activity, cation channel activity and calcium channel activity. The human set also highlighted sodium channel activity and voltage-gated calcium channel activity. In the biological process branch, the analysis highlighted intracellular signalling cascade and actin-filament based process. The Drosophila set was too small (7 genes), to give any detailed statistics, but GObar did highlight transporter activity, which is a parent of the cation channel activity. The highlights that are present in both human and mouse genomes could point to reasons for the persistence of AT-AC U12-dependent splice sites over evolutionary time-scales. We believe this might have some basis in the biological control of the rates of splicing reactions of these genes but reaching a firm conclusion requires an investigation that is beyond the scope of this study.
Discussion
Our proposed method of analysis is mathematically robust and allows visualization and identification of pathways. We can identify sub-groups of genes that cannot be explained by chance alone. This in turn can identify pathways that are of interest in the process under study. Identification of the pathways then allows study of other genes in the pathway that are not picked up in the experiment, allowing for a clearer understanding of subtle effects and quantifying the errors in the experiment.
The conclusions we reach using our method of analysis does depend on the accuracy of the gene annotations. Thus, if the role of a gene in a pathway were unknown, or if a small set of genes could have a strong phenotypic effect (without triggering major changes in the mRNA levels of a large number of genes, which is the only quantity measured in a microarray experiment), then GObar will mislead the investigator. Such phenomena are, in general, more difficult to study using microarrays and require supplementary biological assays to uncover the underlying mechanisms.
Using the GO tree allows us to ameliorate some of the problems inherent in gene annotations, so that genes below the term of interest are also counted as part of the function being studied. Our method involves a one-time analysis of the whole genome dataset, which then allows us to decide, in a straightforward manner, the significance of any number of datasets and allows easy navigation and analysis of the data. The convenience and robustness of the method are the novel contributions here.
Another rigorous approach identifies biological themes from gene lists using GO, by calculating the over-representation of categories in the experimental list relative to a background list (all genes on the chip or all genes in the genome). A problem with this is the underlying assumption that the GO annotations at each node are accurate. In our approach, in order to allow for the fact that genes may be placed at different depths due to human biases, we let every gene below a particular node contribute to the counts on that node, but done in a way that prevents multiple counting. This also allows genes with more specific functional annotations to contribute to the more general annotation. Thus, our approach also differs from this one in the way we define hits, by allowing genes that are lower down in the tree to be a part of the node under consideration.
We offer a detailed comparison with GOstat [11] available at its website [12]. GOstat returns lists of significantly enriched GO terms and genes in those GO terms. In order to see the meaning of the term and its placement in the GO tree the user can click on the link to go to AMIGO [13]. The results are comparable, though we do not offer a P value for the significant nodes. They also offer a list of GO annotations for each gene in the uploaded list. A list of terms is not human friendly and is not a natural method of presenting data that has a tree-like structure. We feel that our SVG based view of the tree is intuitively easier to use and also allows for a quick overview of the data, allowing the user to zoom into relevant sections. Also, our listing of the functions in order of their level in the GO tree allows a user to pay attention to more nodes based on levels (more general corresponds to lower level, while higher levels correspond to more specific terms). The goal of the study determines what level of specificity for a node is preferred.
In addition, the GO browser offered on the GObar website allows an SVG-based exploration of the GO tree, simultaneously showing all the branches and relationships between them, which is different from the text based version offered by the AMIGO website [13]. GObar also allows viewing all the GO annotations of a gene in a single view, as shown in figure 2, which is also a novel feature of this program.
The analysis of genes with U12-dependent splice sites, given in the previous section, is indicative of the power of this approach. We identified functions that are over-represented in the AT-AC-U12 set, which in turn can be the starting point of an investigation into the phylogeny of the genes involved. The phylogeny could explain the evolution and the role of AT-AC-U12 type splice sites.
Conclusion
GObar is a convenient tool for the analysis of large gene lists. It provides useful guidance to biologists on functions and pathways that need further study and is available freely over the web at [2].
Authors' contributions
Ravi Sachidanandam conceived the algorithm and the project and helped with the coding. Gurpreet Katari implemented an initial version of the code before leaving his position and Jason Lee improved and revised much of the code and implemented the web-based front end.
Figure 6 Calculation of the bare and distributed counts of genes at each GO term. The nodes in the figure are GO terms, the arrows are directed from parent to child nodes. The bare count (BC) at each node is the number of genes that are placed there by the annotations of the gene lists. The distributed counts (DC) are the counts transmitted up from the children of the node. Each node contributes its total count = bare count + distributed count, equally up to each of its parents. Thus half of the total number of genes in Node 3 are contributed to the distributed counts of Node 0 and Node 4.
Acknowledgements
Vladimir Grubor, Michele Hastings, Xavier Roca and Susan Janicki gave many detailed suggestions and comments. Andrew Olson implemented the GO term browser available on the GObar website. Cat Eberstark put tremendous effort into improving the figures. The anonymous reviewers helped improve the paper with several suggestions. The Cancer Center of CSHL funded this work and several members of the cancer center provided encouragement.
==== Refs
Gene Ontology Consortium website
GObar website
Smith B Williams J Schulze-Kremer S The Ontology of Gene Ontology Proceedings of AMIA Symposium 2003
Gene E Entrez Gene: unified query environment for genes
FlyBase FlyBase: A database of the drosophila genome
SGD Saccharomyces Genome Database
Graphviz website
Sachidanandam R Curated collection of splice sites in various genomes Unpublished
Burge CB Padgett RA Sharp PA Evolutionary fates and origins of U12-type introns Mol Cell 1998 2 773 85 9885565 10.1016/S1097-2765(00)80292-0
Wu Q Krainer AR AT-AC Pre-mRNA Splicing Mechanisms and Conservation of Minor Introns in Voltage-Gated Ion Channel Genes Molecular and Cellular Biology 1999 19 3225 3236 10207048
Beissbarth T Speed TP GOstat: Find statistically overrepresented Gene Ontologies within a group of genes Bioinformatics 2004 20 1464 1465 14962934 10.1093/bioinformatics/bth088
Beissbarth T Speed TP GOstat Website
The GO consortium Amigo website
|
16042800
|
PMC1190157
|
CC BY
|
2021-01-04 16:27:25
|
no
|
BMC Bioinformatics. 2005 Jul 25; 6:189
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-189
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1941607639910.1186/1471-2105-6-194Methodology ArticleFunctional annotation by identification of local surface similarities: a novel tool for structural genomics Ferrè Fabrizio [email protected] Gabriele [email protected] Andreas [email protected] Manuela [email protected] Boston College, Biology Department, Chestnut Hill MA, USA2 Centre for Molecular Bioinformatics, Department of Biology, University of Rome Tor Vergata, Italy2005 2 8 2005 6 194 194 26 1 2005 2 8 2005 Copyright © 2005 Ferrè et al; licensee BioMed Central Ltd.2005Ferrè et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Protein function is often dependent on subsets of solvent-exposed residues that may exist in a similar three-dimensional configuration in non homologous proteins thus having different order and/or spacing in the sequence. Hence, functional annotation by means of sequence or fold similarity is not adequate for such cases.
Results
We describe a method for the function-related annotation of protein structures by means of the detection of local structural similarity with a library of annotated functional sites. An automatic procedure was used to annotate the function of local surface regions. Next, we employed a sequence-independent algorithm to compare exhaustively these functional patches with a larger collection of protein surface cavities. After tuning and validating the algorithm on a dataset of well annotated structures, we applied it to a list of protein structures that are classified as being of unknown function in the Protein Data Bank. By this strategy, we were able to provide functional clues to proteins that do not show any significant sequence or global structural similarity with proteins in the current databases.
Conclusion
This method is able to spot structural similarities associated to function-related similarities, independently on sequence or fold resemblance, therefore is a valuable tool for the functional analysis of uncharacterized proteins. Results are available at
==== Body
Background
Detection of sequence or fold similarity is often used to infer the function of uncharacterized proteins. By this approach one can tentatively assign a function to approximately 45–80% of the proteins identified by the genomic projects [1,2]. However, function is mostly determined by the physical, chemical and geometric properties of the protein surfaces [3,4], and cases have been described where the same local spatial distribution of residues important for function is achieved with apparently unrelated structures and/or sequences [5]. One of the best known examples is represented by the SHD catalytic triad of serine proteinases [6-8]. Furthermore, surface similarities have been detected in unrelated ATP/GTP binding proteins [9,10] and in the guanine binding sites of p21Ras family GTPases or in the RNA binding site of bacterial ribonucleases [10]. By local structural comparison Hwang et al. [11] were able to infer correctly the nucleotide binding ability of an uncharacterized Methanococcus jannaschii protein.
On the other hand, similar folds can have different functions if their active sites have diverged [12-15]. As a consequence, methods purely relying on sequence and global structure comparison may lead to inaccurate function-related annotations in cases in which few residues are responsible for the specificity of substrate interaction.
The vast majority of well-studied functions (enzymatic activities, binding abilities etc.) are encoded by a relatively small set of residues, often not contiguous in the protein sequence but organized in a conserved geometry on the protein surface that may be used as a marker for reliable functional annotation. Although exposed to the solvent, these function-related residues are often located in surface clefts or cavities [16]. Such residues define functional modules conserved in some proteins sharing a molecular function even if differing in sequence and structure. Several tools for discovering conserved three-dimensional patterns in protein structures have already been proposed [17-20]. Schmitt et al. [21] developed a clique-based method to detect functional relationships among proteins. This approach does not rely on detection of sequence or fold homology and highlights a number of non-obvious similarities among protein cavities. The algorithm, however, is computationally intensive and cannot be applied to an all-against-all analysis of protein surface regions. Binkowski and co-workers [22] recently described an approach for detecting sequence and spatial patterns of protein surfaces: the underlying algorithm is fast, but cannot identify similarities that are independent of the residue order in the compared proteins. Two related papers [23,24] describe a method for local structural similarity detection, which is of great relevance since it is able to evaluate the statistical significance of each match. This method (PINTS) has been then used to analyze protein structures from structural genomics projects [25]. Other recent papers present algorithms able to find structural motifs possibly related to a function and to use them to scan protein structure libraries [26-31].
In a previous work [32] we described the construction of a non redundant library of surface annotated functional sites and a fast comparison algorithm able to find structural similarities independently on the residue sequence order. We report here the analysis of the results of the first all-versus-all comparison of the protein functional sites, the validation of the comparison procedure in a test dataset and its application for annotating a dataset composed of proteins solved in structural genomics projects. The results are available for experimental test at the address .
Results and discussion
Functional sites comparison
We used the compendium of protein surface regions associated to molecular functional sites stored in the SURFACE database [32]. This is a collection of 1521 annotated functional regions obtained following the procedure described in Figure 1 and in the Methods section. Each patch has at least a function-related annotation, that may be the ability to bind a certain ligand, or a match with a PROSITE or ELM pattern [33,34]. Ligand-binding abilities are included among gene ontology (GO) molecular functions [35], as well as many PROSITE patterns and ELM motifs. Some other PROSITE patterns correspond to short motifs that are conserved in all members of certain protein families, which not necessarily are associated to known function-related residues. We chose to include this class of patterns in our annotation system, since they offer a quick way to verify the reliability of a match, and in many cases these motifs do contain functional residues. Hence, our annotations can be classified either as molecular functions or protein signatures. It is worth noticing that the annotation is extended to the whole patch but is also assigned to a subset of specific annotated functional residues.
Figure 1 Description of the experimental procedure. Surface functional sites are automatically located and annotated as described in Methods. Surface clefts, identified by means of SURFNET, are filtered using a volume threshold, and annotated for the binding ability or for the presence of a functional motif from the PROSITE or ELM databases. This library (the SURFACE database) is used to scan a non-redundant collection of protein structures; a semi-automated procedure is used to define conditions for which the structural similarity implies also a functional relationship. Finally, the SURFACE database is used to analyze a list of proteins with unknown function from structural genomic projects, obtaining in several cases significant similarities that could have not been spotted through sequence or fold similarity.
In [32] the structural matches obtained from the comparison of he SURFACE library against the entire collection of surface clefts (both annotated and not annotated) were evaluated by means of the Z-score of each match length against the distribution of the match lengths for any given annotated patch. Here we perform an exhaustive analysis in order to find conditions for which a structural similarity also suggests a function-related similarity. First, only those matches which include annotated functional residues are considered, therefore each structural similarity match is likely to hold a functional meaning. This step is crucial since many matches may be obtained because of general fold similarity, without an underlying functional relationship. Finding a functional match induces an annotation of at least some of the residues, and suggests reasonable hypotheses as to function (we are currently investigating how to use our approach to find novel function-related structural motifs, i.e. recurrent structural matches between proteins that can not be explained only by fold similarity and that may imply a previously undetected functional similarity).
From the comparison of the SURFACE library against the entire collection of surface clefts, we collected a grand total of 65910 stringent matches among patch pairs, about 4.5% of which involve 6 or more residues and 4.5% involve 10 or more residues. A not negligible amount of these matches involve residue pairs whose relative distance is not conserved in the corresponding protein sequences. More interestingly, some of the matches involve residues whose sequence order and/or sequence spacing is different in the two proteins: some of these cases, that may be examples of convergent evolution, are currently under investigation. As an example, metals can interact with proteins by means of similar arrangements of residues that can be found across different folds [36-38]. Scanning our dataset with zinc-binding patches leads to the finding of significant matches to proteins belonging to 42 different folds and 6 different classes as defined by SCOP [39]. Different metal-binding patches lead to similar findings, even though less dramatic. Further analysis would suggest how many of these cases are associated with functional similarities as well.
The fraction of matches validated (as described in the Methods section) sensibly increases with the Z-score (Table 1). At lower Z-scores, the GO terms and SWISS-PROT keywords validation methods are more represented, while, for more significant matches, ability to bind the same ligands, fold similarity and co-presence of PROSITE motifs become more relevant.
The matches that cannot be structurally or functionally justified by these methods and that are characterized by a high Z-score are relatively few (see Table 1). 171 matches out of 2173 (7.9%) having a Z-score higher than 7 are not validated following the above mentioned criteria (Table 1). Of these 171 matches, 130 can be considered as true positive matches, confirmed by literature and information derived from different sources and databases. The remaining 41 matches (1.9%) are not confirmed and should be tested experimentally. About 2% of the highly significant matches can be considered as possible false positive hits or new annotations. Some of these cases are shown and discussed in Figure 2(a,b).
Table 1 Structural matches Z-score distribution and validation. This Table shows the number of structural matches (second column from the left) found as a function of the Z-score of the match. The third column from the left (labeled "validated") reports the number of matches for which at least one of the validation criteria holds. The following columns show a breakdown of the number of matches validated by each validation condition (from the fourth column on the left to the rightmost: same PROSITE pattern annotation; same binding ability; common GO term annotation; same SCOP fold; same Enzyme Classification number; sequence similarity at least 40%; common SwissProt keyword). Note that the sum of the matches validated by the different criteria for each row is higher than the total number of validated matches at that given Z-score, since some matches can satisfy more than one condition. At increasing Z-scores, the ratio of validation condition that we consider less reliable (SwissProt keywords, GO terms) decreases, while the ratio of more reliable annotations (i.e. same binding ability, same PROSITE pattern annotation) increases.
Z-score Total Validated PROSITE Ligand GO Scop E.C. Seq. sim. SwissProt kw
3.0 31341 7066 366 951 3565 765 99 2 5655
3.5 14948 4002 747 830 2222 889 48 3 2944
4.0 9721 2814 557 613 1680 788 44 1 2043
4.5 3942 1346 440 467 841 390 32 1 989
5.0 1549 764 281 234 436 411 5 1 514
5.5 976 612 287 181 320 399 7 0 342
6.0 639 457 177 209 267 271 3 0 323
6.5 621 548 279 258 298 447 4 0 383
7.0 365 328 157 115 180 246 2 0 200
7.5 260 219 105 68 109 176 6 1 152
8.0 270 238 104 87 149 191 0 1 169
8.5 209 195 80 57 129 153 8 1 131
9.0 122 107 54 54 70 87 0 0 63
9.5 137 129 60 48 74 119 0 0 80
10.0 124 113 53 61 75 104 0 1 86
10.5 55 51 17 22 29 43 2 0 36
11.0 106 103 46 40 65 91 4 0 66
11.5 88 88 42 43 65 80 5 0 55
12.0 78 77 33 34 51 75 5 0 52
12.5 71 69 26 32 38 64 5 1 54
13.0 49 47 30 21 24 45 0 0 30
13.5 39 39 9 19 17 39 1 0 24
14.0 29 29 14 16 18 29 3 0 25
Figure 2 Significantly matching residues on proteins sharing no structural or sequence similarity. Similarity detected comparing the SURFACE database of annotated functional sites against a list of annotated monomers (a,b) or proteins with unknown function from structural genomics projects (c,d,e,f); the annotated patch residues are colored in blue, the matching residues in red; whenever possible, the patch annotation (bound ligand or PROSITE pattern) is shown. (a) Similarity detected between the E. coli UDP-galactose 4-epimerase (PDB code 1nah) NADH-binding patch and the H. influenzae YecO methyltransferase (1im8); the NAD co-crystallized with 1nah is shown; the similarity involves 7 residues (with a Z-score 9.06). (b) Structural similarity between the HEXOKINASES PROSITE pattern-annotated patch of the human hexokinase type I (1qha) and the bacteriophage ms2 capsid protein; additional 1qha annotated residues are shown in yellow. (c) Structural similarity detected between the B. subtilis Yqvk protein, and the Wolinella succinogenes fumarate reductase cytochrome B subunit heme group binding patch. (d) Match between Hi1480 protein from Haemophilus influenzae and the bovine cytochrome Bc1 heme-binding patch. (e) Similarity between the B. subtilis protein Yqeu and the E. coli Grea transcript cleavage factor GREAB_1-annotated patch; additional pattern-annotated residues are shown in yellow. (e) Similarity between E. coli lysozyme inhibitor and two ATP-binding patches, the Rattus norvegicus 6-Phosphofructo-2-Kinase/ Fructose-2,6-Bisphosphatase major patch (red) and the mouse Aaa ATPase P97 (green).
From this validation procedure the emerging result is that, using stringent parameters in the comparison step and using the Z-score as a threshold, our algorithm is reliable and able to spot local structural similarities related to functional relationships with only few non confirmed hits, which can be considered as false positives or as testable hypotheses.
An estimation of false negative matches (defining false negative match as the missing detection of structural similarity between two proteins sharing the same function) is not immediate, for the reason that the same or similar molecular function may be achieved in different ways using a different three-dimensional residue arrangement. We estimated the occurrence of false negatives for PROSITE annotated patches, using the list of known true positives (for which the function encoded by the pattern is experimentally verified) for each pattern that is provided by PROSITE. The procedure is done as follows: for all the patches annotated with a given PROSITE pattern, we collect all matches obtained scanning with these patches the entire patches dataset, selecting only those matches having Z-score higher than a fixed threshold. The fraction of known true positives that are not found using the pattern-annotated patches as queries (i.e. the false negatives), when retrieving only those matches having Z-score higher than 5, is 0.3 (meaning that we are able to correctly retrieve the 70% of the occurrences of PROSITE patterns in the dataset), and it raises to 0.35 setting the Z-score threshold to 7.
Benchmark cases
To further test the ability of the procedure in finding known cases of functional similarities among proteins for which sequence and/or structure similarity is not significant, a number of benchmark cases were investigated (Figure 3):
Figure 3 Benchmark cases analysis. (a) Structural superposition of the S. cerevisiae (red) and the E. coli (blue) chorismate mutase (PDB code 4csm and 1ecm, respectively). These two patches align ten residues, with a resulting Z-score of 15.76. (b) Structural superposition of the 4-hydroxyphenylpyruvate dioxygenase (PDB code 1cjx, red), the 2,3-dihydroxybiphenyl 1,2-dioxygenase (1han, blue), catechol 2,3-dioxygenase (1mpy, green) and the methylmalonyl-Coa epimerase (1jc5, yellow). The 1han co-crystallized iron ion is shown. (c) Superposition of the tumor necrosis factor-alpha-converting enzyme (1bkc, red) and the peptide deformylase (1icj, blue). The 1icj co-crystallized nickel ion is shown. (d) Structural superposition of the human P21 ras protein (5p21, red) and HprK/P 1jb1 (blue). (e) Structural superposition of the 1b37 FAD-binding pocket (red) with the highest-score matches obtained in a database search (blue). The 1b37-bound FAD is shown. (f) Bound ligands superposition. Using the three-dimensional transformation used to superpose the residues aligned in (e), also ligands that are bound to some of these proteins are consequently superposed. The ADP molecule bound to the 1djn patch nicely matches the ADP moiety in the similar FAD-binding pockets.
i) The S. cerevisiae and the E. coli chorismate mutase (PDB codes: 1ecm and 4csm, respectively), despite the very low sequence identity, share a similar fold and a similar main functional site [18,21]. The 1ecm largest patch is annotated for the oxy-bridged prephenic acid binding ability. Using this patch as a query, the highest Z-score match is found with the 4csm largest patch (Figure 3a).
ii) The Glyoxalase/Bleomycin resistance protein/Dihydroxybiphenyl dioxygenase fold is common to several unrelated metal ion binding proteins sharing similar catalytic mechanisms, including the bleomycin resistance protein, glyoxalase I, and a family of extradiol dioxygenases [40]. We detected a significant similarity among P. fluorescens 4-hydroxyphenylpyruvate dioxygenase (PDB code 1cjx), B. cepacia 2,3-dihydroxybiphenyl 1,2-dioxygenase (1han), P. putida catechol 2,3-dioxygenase (1mpy) and P. shermanii methylmalonyl-Coa epimerase (1jc5). The comparison algorithm correctly identifies the residues involved in Fe binding (Figure 3b). 1han second largest patch is annotated for the iron binding ability. Structural matches with 1mpy, 1cjx and 1jc5 functional sites are found at high Z-score (7.19).
iii) Metal ions can be coordinated by histidine clusters. We identified a similarity between the human tumor necrosis factor-alpha-converting enzyme (PDB code: 1bkc) Zn binding site and the E. coli peptide deformylase (PDB code: 1icj) Ni binding site, despite their sequence and fold diversity (Figure 3c). The zinc-binding patch of 1bkc shares eight residues in the same structural conformation with the nickel-binding patch of 1icj, with a Z-score of 10.66.
iv) Nucleotide binding abilities can be associated with several unrelated proteins; we detected a high-scoring match between the GTP-binding annotated patch of the human p21 ras protein (5p21) and the L. casei Hpr kinase (1jb1) that aligns eight residues with a Z-score of 9.01 (Figure 3d). These two proteins do not share any significant sequence or fold similarities.
As a further test, we analyzed the flavin-adenine dinucleotide (FAD) binding pockets, known to share structural similarities with other adenine-containing nucleotide binding pockets, despite sequence and fold differences [41,42]. FAD consists of an adenosine monophosphate (AMP) linked to a flavin mononucleotide (FMN) through a pyrophosphate bond and is involved as a cofactor in many biological processes. Using the FAD-binding patch of the Zea mays polyamine oxidase (1b37) as a bait, we selected 9 prey patches with Z-score higher than 12: 8 preys are annotated as being able to bind a FAD molecule and belongs to the same SCOP fold (FAD/NAD(P)-binding domain). The remaining trapped patch is the biggest patch of the trimethylamine dehydrogenase from Methylophilus methylotrophus (1djn), an iron-sulfur flavoprotein, and it is annotated as ADP-binding. 1djn is co-crystallized also with a FMN, which is very similar to FAD, but this ligand is associated to the second largest patch of the 1djn structure. The residues, which were associated by the alignment program, are shown in Figure 3e. These proteins share a very low sequence similarity, which cannot be revealed using BLAST2 [43]. The ADP binding patch of the 1djn structure is nicely superposed to the other patches in the binding pocket (Figure 3e), but shares no evident fold similarity with the other ones, and belongs to a different SCOP fold (the nucleotide-binding domain). When the selected structures in Figure 3f are physically superposed (finding the least-square fitting of the matching residues), also the ligands bound to these structures turn out to be nicely superposed. The procedure could therefore highlight the ability to bind a subset of the FAD molecule, namely an ADP molecule in the 1djn major patch, even with very low levels of sequence and structure similarity. Using each FAD binding patch to scan the dataset, we selected only proteins for which known functional properties are consistent with the FAD or nucleotide binding ability.
Structural genomics proteins analysis
With the stringent parameters described above, we were able to detect only matches linked to function-related similarities, even in cases of non-homologous proteins. For that reason, once proved to be reliable, the procedure can be applied as a predictive tool to obtain clues concerning the function(s) of uncharacterized proteins.
We selected 257 protein structures from the PDB, corresponding to 513 chains that are marked as being of unknown function, or for being a hypothetical protein or for having been solved within a structural genomics project. We analyzed these structures by looking for reliable similarities to our functional sites library and were able to suggest one or more molecular functions to 191 of these chains, for a total of 534 similarity matches. For each match, we checked if the previously described criteria hold (i.e. common GO term, SwissProt keyword, EC number or SURFACE annotation). If not, a literature search has been done to verify the functional relationship. By means of this analysis of the likelihood of each single match, we found that 322 (the 60.3%) of these hits are validated by experimental analysis that have already characterized many of these proteins, while only 29 matches (5.4%) are not found confirmed in previous findings; 107 (20%) hits involve proteins for which the functions are still unknown; 76 hits (14.2%) involve proteins for which a hypothetical function has been assigned by means of sequence or structure global similarity. In this latter case, the function-related annotation obtained from our method can be considered as a new functional annotation that corrects or improves the actual function assignment. Hence, we were able to propose a function by similarity using the annotated patch database 184 times, to 127 different chains (matches with Z-score at least 7 are shown in Table 2). 56% of these new functional annotations are about a PROSITE pattern, the remaining 44% about a ligand binding ability; this is somewhat surprising, since the majority of the patches annotations in the SURFACE library regards binding abilities. A selection of the proposed functional regions is shown in Figure 2(c,d,e,f), while the complete list can be found at . For each match we tested the BLAST2 pair-wise sequence similarity between the sequence of the protein to which the query patch belongs and the target protein sequence, the PsiBLAST sequence similarity matches obtained by running the target sequence versus the non-redundant SwissProt+TrEMBL sequence database, the global structural similarities of the target structure in the PDB using SSM, and the local similarity using PINTS [24]. The match with the highest Z-score (14.29) is between the B. subtilis Yqvk protein (PDB code 1rty), and the Wolinella succinogenes fumarate reductase cytochrome B subunit major patch (1qlaC1), annotated with the heme group binding ability; the structural similarity involves 7 residues. The two proteins do not share any sequence or structural similarity, as checked using BLAST and the structural comparison algorithm SSM [44]. A PsiBLAST run of the Yqvk sequence against the non-redundant SwissProt+TrEMBL shows a significant similarity (E-value 4e-19) with the mouse cobalamin adenosyltransferase (SwissProt entry name MMAB_MOUSE), while the SSM comparison against the whole PDB leads to only one significant similarity, with another uncharacterized protein, the conserved protein 0546 From Thermoplasma acidophilum (1nog). A PINTS comparison [24] of Yqvk, against pre-compiled libraries of structural patterns, retrieves as most significant matches one with the human Small Nuclear Ribonucleoprotein Sm D3 (PDB code 1d3b), aligning 3 pairs of residues with r.m.s.d 0.32 and E-value 0.00481, and another with the pig Dihydropyrimidine Dehydrogenase 1htx (3 pairs aligned with r.m.s.d. 0.337 and E-value 0.00839). The heme binding ability thus may be a new functional annotation of this poorly known protein. The second highest Z-score match (13.32, 9 residues structurally aligned) occurs between Hi1480 protein from Haemophilus influenzae (1mw5) and the bovine cytochrome Bc1 heme-binding patch (1bgyC2). No significant sequence similarity is found in the SwissProt+TrEMBL (the highest match, whose E-value is 2.1, involves the putative E. coli RNA helicase, SwissProt entry name RHLE_ECOLI), as well as no significant matches are found using SSM. PINTS matches involving three residues are found with the virus influenzae Bha/Lsta protein (1mqm) and the Candida tropicalis Enoyl Thioester Reductase 2 (1h0k), whose E-values are 0.401 and 0.451, respectively. Another high-score match (Z-score 10.05, length 7 residues) is found between the B. subtilis protein Yqeu (1vhk) and the E. coli Grea transcript cleavage factor major patch (1grj_1), which is annotated with the GREAB_1 PROSITE pattern, a signature of this class of cleavage factors. Yqeu share SSM-detected structural similarities with another unknown-function protein (namely H. influenzae 1nxz) and significant sequence similarity with a list of hypothetical and uncharacterized bacterial proteins. PINTS reports a local structural similarity with the zinc-binding site of the E. coli CTP-ligated T state aspartate transcarbamoylase (E-value 0.00894, r.m.s.d 0.544 over three pairs of residues).
Table 2 Non-validated functional annotations of non-annotated surface patches. Functional annotated sites have been compared to a collection of surface patches extracted from a non-redundant PDB subset. The reliability of each match was estimated via a series of criteria, as described in the text. The remaining similarities may be new functional annotations of uncharacterized functional sites, or false positive matches, and are shown in this table. Columns:(i) PDB code, chain name and patch number in the annotated query patch; (ii) Description of the protein to which the query patch belongs; (iii) Query patch functional annotation; (iv) Target patch; (v) Description of the protein to which the target patch belongs; (vi) Z-score of the match; (vii) SSM Q score; (viii) SSM P score; (ix) SSM Z score. The SSM Q score takes into account the number of aligned residues, their r.m.s.d. and the size of the proteins; a high Q score means a good similarity. The SSM P score is the log of the pValue (the probability that the match occurred by chance); P scores higher than 3 are considered significant by the authors of the method.
Patch 1 Protein Patch 1 Annotation Patch 2 Protein Z-score SSM Qscore SSM P-value SSM Z-score
3mdeA1 Acyl-CoA dehydrogenase LIG_CO8 1g5bB6 Bacteriophage lambda S/T Protein Phosphatase 9.59 0.01 0 0.5
1qhaA2 Hexokinase I HEXOKINASES 1i78A5 Outer Membrane Protease Ompt 9.44 0.01 0 0.5
1qhaA2 Hexokinase I HEXOKINASES 1zdhA2 Bacteriophage Ms2 Protein Capsid 9.44 0.01 0 0.1
1bp1_1 Bactericidal permeability-increasing protein LIG__PC 1qlwA2 Bacterial esterase 713 9.07 0.01 0 1.5
1nah_1 UDP-galactose 4-epimerase LIG_NAD 1im8A1 YecO methyltransferase 9.06 0.1 0 4
4blcA1 Beef liver catalase LIG_NDP 1io1A5 Phase 1 Flagellin 8.86 0.01 0 1.4
1dbtA1 Orotidine 5'-Monophosphate Decarboxylase OMPDECASE 1dj8A1 E. Coli Periplasmic Protein Hdea 8.76 0.03 0 1.9
1fp2A1 Isoflavone O-Methyltransferase LIG_SAH 1nah_1 UDP-galactose 4-epimerase 8.6 0.05 0 5.5
1fps_1 Prenyltransferase Trimethylamine POLYPRENYL_SY NTHET_1 1h6gA2 Alpha-catenin Molybdopterin Biosynthesis Moeb 8.54 0.04 0 0.3
1djnA1 dehydrogenase LIG_ADP 1jwbB1 Protein 8.51 0.05 0 5.3
19hcA1 Cytochrome C LIG_HEM 1umuB1 UmuD' protein 8.44 0.03 0 4.2
1qhaA1 Type I Hexokinase HEXOKINASES 1e2uA1 Hybrid Cluster Protein 8.34 0.01 0 0.1
256bA1 Cytochrome B562 LIG_HEM 1gpjA1 Glutamyl-tRNA reductase 8.25 0.05 0 0.4
1ep1B1 Dihydroorotate Dehydrogenase B LIG_FAD 1pmi_8 Phosphomannose Isomerase 8.18 0.02 0 0.3
1tsdA1 Thymidylate synthase LIG_U18 1prhA1 Prostaglandin H2 Synthase-1 Formylmethanofuran: Tetrahydromethanopterin 8.16 0.01 0 0.1
2nlrA1 Endoglucanase LIG_G2F 1ftrA1 Formyltransferase 8.05 0.02 0 0.5
1ej0A1 RNA Methyltransferase LIG_SAM 2cmd_1 Malate Dehydrogenase 8.01 0.12 0 3.9
1ecmB1 Chorismate mutase LIG_TSA 1b3qB1 Histidine Kinase Chea 7.96 0.02 0 2.8
1av6A3 Vaccinia Methyltransferase Vp39 LIG_SAH 1b3mA1 Sarcosine oxidase 7.95 0.02 0 2.8
1av6A3 Vaccinia Methyltransferase Vp39 LIG_SAH 1b4vA1 Cholesterole oxidase 7.95 0.02 0 0.9
1qrrA1 Sulfolipid Biosynthesis (Sqd1) Protein LIG_NAD 1g6q12 Arginine methyltransferase HMT1 7.85 0.04 0 1.9
1qrrA1 Sulfolipid Biosynthesis (Sqd1) Protein LIG_NAD 1im8A1 YecO methyltransferase 7.85 0.09 0 2.4
1qrrA1 Sulfolipid Biosynthesis (Sqd1) Protein LIG_NAD 1khhA1 Guanidinoacetate methyltransferase 7.85 0.1 0 2.9
6reqA1 Methylmalonyl-Coa Mutase LIG_3CP 1fepA2 Ferric Enterobactin Receptor 7.79 0.01 0 0
6reqA1 Methylmalonyl-Coa Mutase LIG_3CP 1jihB10 Yeast DNA Polymerase Eta 7.79 0.01 0 1
1bgyC1 Cytochrome BC1 LIG_HEM 1dc1B2 Bsobi Restriction Endonuclease 7.62 0.01 0 0.4
1bgyC2 Cytochrome BC1 LIG_HEM 1k92A4 Argininosuccinate Synthetase 7.62 0.01 0 0.2
1bgyC2 Cytochrome BC1 LIG_HEM 5r1rA2 Ribonucleotide Reductase R1 7.62 0.01 0 0.9
1qanA1 Rrna Methyltransferase Ermc' RRNA_A_DIMETH 1b37B1 Flavin-dependent polyamine oxidase 7.54 0.04 0 5.3
1qanA1 Rrna Methyltransferase Ermc' RRNA_A_DIMETH 1b3mA1 Sarcosine oxidase 7.54 0.04 0 4.3
1qanA1 Rrna Methyltransferase Ermc' RRNA_A_DIMETH 1gpeA1 Glucose oxidase 7.54 0.03 0 3.2
1qanA1 Rrna Methyltransferase Ermc' RRNA_A_DIMETH 1i8tA1 UDP-galactopyranose mutase 7.54 0.04 0 4.1
2cut_1 Serine esterase LIG_DEP 1jfrA1 Exfoliatus Lipase 7.43 0.17 0 5.3
1bp1_2 Bactericidal Permeability-increasing protein LIG__PC 1fuoA10 Fumarase C 7.42 0.01 0 0.1
1hcy_4 Hexameric haemocyanin LIG_NAG 2kinA2 Kinesin 7.42 0.01 0 2.2
1cpq_1 Cytochrome C LIG_HEM 1wpoB1 Human Cytomegalovirus Protease 7.41 0.01 0 1.3
1inp_1 Inositol polyphosphate 1-phosphatase IMP_2 1bgxT6 TAQ polymerase 7.38 0 0 0
1ksaA1 Bacteriochlorophyll A Protein LIG_BCL 1xvaA1 Glycine N-Methyltransferase 7.27 0.02 0 1.3
1b63A1 MutL DNA mismatch repair protein LIG_ANP 1wpoB1 Human Cytomegalovirus Protease 7.22 0.03 0 0.6
1e7uA1 Phosphoinositide 3-Kinase Inhibition PI3_4_KINASE_1 1qi9B1 Vanadium Bromoperoxidase Soluble Quinoprotein Glucose 7.15 0.01 0 0.6
1a12A1 Regulator Of Chromosome Condensation (Rcc1) RCC1_2 1cruB1 Dehydrogenase 7.06 0.08 0 0.4
In some cases we found a structural similarity between a protein with unknown function and two patches annotated with the same function, giving strength to the hypothesis of function-related similarity. The conserved hypothetical protein (Tm0667) from Thermotoga maritima (PDB code 1j6o) shows a structural similarity with surface patches of E. coli nucleotidyltransferase (1gupA2) and Desulfovibrio gigas rubredoxin:oxygen oxidoreductase (1e5dA4), both annotated with the iron binding ability. The E. coli lysozyme inhibitor (1gpq), whose function is still uncharacterized, may bind ATP given the similarity to the Rattus norvegicus 6-Phosphofructo-2-Kinase/ Fructose-2,6-Bisphosphatase major patch (1bif_1) and the mouse Aaa ATPase P97 (second patch (1e32A2)).
For each described match we propose that the detected structural similarity reveals a function-related similarity. For each match we checked whether the similarity could have been detected by means of sequence similarity, as checked using BLAST and PsiBLAST, or structural comparison, as checked by means of SSM and PINTS. Our approach, that is based on comparison of local functional surface residues, independently on their sequence order, may overcome the limitations of current methods possibly due to our incomplete knowledge of the sequence/structure/function relationship or to convergent evolution. Even using PINTS, which is a tool similar in philosophy to our approach, the findings are different, suggesting that different tools may be complementary in the difficult task of protein functional annotation; on the other hand, this may also highlight the difficulty in evaluating the significance of local similarities that in many cases are restricted to a very small number of residues.
Conclusion
The expected burst in the number of protein structures that are not associated to a biological function, stimulated by the structure genomics programs, has emphasized the need for tools to reveal structural regularities even in proteins that do not share sequence or fold similarity [1,45]. Protein structures selected in structural genomics projects usually share very little sequence similarity with the dataset of already characterized proteins [46]. Sequence analysis tools are therefore unsuitable for inferring their functions. Moreover, cases are known where active site residues are not conserved in proteins sharing a common structural fold; therefore, "traditional" structure comparison tools are also not always able to help in function-related annotation.
Using a fully automated procedure, we obtained a reliable library of protein annotated functional sites. A fast structural comparison algorithm allows the rapid scanning of one or more protein structures with the library looking for local structural similarities. This method is designed to help in functional annotation in difficult cases. Our annotated surface patches determination and comparison method offers a new and powerful resource for detecting related function among unrelated proteins, for proteins solved in structural genomics projects or for identifying new function-related sites on the surface of already characterized proteins. We have been able to provide one or more functional clues to a large set of novel proteins, and, where functional evidences are already known, our findings confirm them. Moreover, just as proteins with different sequence and fold can share a similar functional site, proteins with similar sequence and/or fold can have small local differences leading to a completely different function [1,21]. Our method, which is focused on a detailed analysis of functional sites, is able to successfully predict protein functions in these difficult cases. Therefore, it can be used in analyzing the complex evolutionary relationships among protein sequence, structure and function [47-49]. The complete list of the functional predictions that we obtained is accessible at URL ; the structurally similar residues are shown for each match, and the structural superposition can be viewed through the browser plug-in Chime or RasMol. A novel publicly available web server, PdbFun [50], has been developed to allow the on-line structural comparison of user-defined subsets of residues of protein chains, and pre-defined subsets, like the SURFACE library of annotated functional sites, will be provided.
Methods
Functional site library extraction and annotation
The SURFACE database [32] stores a library of 1521 annotated function-related surface regions obtained using the following procedure (described in Figure 1): first, the SURFNET algorithm [51] is applied to a non-redundant, representative list of around 2000 protein chains from the PDB database [52] (downloadable at ) in order to find all the surface clefts with a volume higher than an arbitrary threshold (200 Å3); then for each cleft, a surface patch is identified as a collection of solvent-exposed residues using the MASK algorithm (that is part of the SURFNET package); finally, we infer the function of such surface patches using two kinds of annotations: ability to bind (associated to surface patch residues that are contacting a bound ligand), and match with PROSITE or ELM [33,34] functional motifs. The ability to bind annotation is carried out selecting those residues within 3.5 Å distance from any of the atoms of a ligand found in the crystal structure. Whenever a single patch contains more than 75% of the ligand-contacting residues (62% of the cases), we assign the ligand binding ability to this surface cleft. Considering only large organic molecules and metal ions, the ratio of the ligands that can be unequivocally associated to a single patch raises to 78%. PROSITE annotations are achieved scanning the sequences of monomers in our dataset using the ScanProsite algorithm [53], finding 928 matches. 12 matches were found with the ELM [34] experimentally verified instances. We did not consider those patterns marked by PROSITE as unspecific. Moreover, we annotated only those residues that correspond to non-X positions in the regular expression and that are exposed to the solvent according to the NACCESS procedure [54,55]. Once the dataset chains have been annotated, we map the annotated residues on the structure and in the surface patches. Whenever a single patch contains more than 75% of the pattern exposed residues, we assign the function encoded by this pattern to the patch (43% of the cases).
Structural comparison
A sequence/fold-independent algorithm was used for local surface comparison [32]. The algorithm starts from a seed match (a pair of residues in the query that can be found in the target, at the same distance and with similar physical and chemical characteristics). The structural superposition, obtained by the quaternions method [56] and assessed at each step by residue similarity and root mean square deviation (r.m.s.d.) of the matching residues, is extended adding neighboring residues to the seed match until r.m.s.d and residue similarity are under user-defined thresholds (we used a similarity at least equal to 0.3 for each added pair of residues, and an average similarity at least equal to 1.2, using the Dayhoff substitution matrix [57] and 0.8Å as maximum r.m.s.d.). We consider only structural matches that include at least a fixed fraction (50%) of functional annotated residues, to increase the likelihood that the structural match is a function-related match as well. The algorithm is very fast and explores all the combinations of similar/identical residues in a sequence-independent way. The score of the match is the number of residues that can be superposed within the defined similarity thresholds. The significance of the score is evaluated by calculating the Z-score over the score distribution of the query patch comparison with the whole dataset: for each match, the Z-score is computed as the difference between the score of the match and the average score of all the matches for the query patch, divided by the standard deviation.
In order to obtain an estimate of the number of true positive matches, defining a true positive match as a structural similarity that implies also a functional similarity, we checked if the two matching proteins share also: (i) a common Gene Ontology (GO) term; (ii) a common SwissProt keyword; (iii) the same Enzyme Classification (EC) number; (iv) the same functional annotation (i.e. the binding of the same ligand or a match with the same PROSITE or ELM pattern). Gene Ontology terms search is limited to molecular function or biological process annotations linked to PDB structures from the GOA project [35]. SwissProt [58] keywords were extracted from the SwissProt entries corresponding to the DBREF field in the PDB [52] files header. If this was not available, we extracted the sequence from the order of residues in the structure, then we looked for a close homolog (sequence similarity higher than 95% using BLAST) in the SwissProt database. Some keywords were excluded because not referring to protein functions (i.e. Structural protein, Polymorphism, Alternative promoter usage, etc.). Furthermore, we checked whether the two matching proteins share more than 40% of sequence similarity or the same fold using the SCOP structural classification [39] at the superfamily level. Our database is composed of patches extracted from a non-redundant list of structures, therefore these cases are infrequent.
Authors' contributions
FF carried out the patches definition, extraction and annotation, and the structural genomics protein functional prediction strategy, and drafted the manuscript. GA is the author of the structural comparison algorithm and participated in the design of the project. AZ participated in the procedure for the validation of structural matches, and in the creation of a relational structure to store and spread the project results. MHC participated in the project design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.
Table 3 Function prediction for uncharacterized proteins. Functional annotated sites have been used to infer the function(s) of a large set of uncharacterized proteins, using similarity threshold values that have been successfully tested on a training dataset. Columns: (i) PDB code and chain name of structural genomics proteins; (ii) PDB code, chain name and surface patch serial number of the functional annotated patch; (iii) Functional annotation of the matching patch; (iv) Z-score of the match; (v) Number of aligned residues; (vi) Blast2 bitscore; (vii) Sequence similarity evaluated by means of the Needleman-Wunsch global alignment (using the EMBOSS package 59 application needle). (viii) SSM Q score; (ix) SSM P score; (x) SSM Z score.
Str.gen SURFACE patch Annotation Z-score Score BLAST2 Seq Sim SSM Q SSM P SSM Z
1rtyC0 1qlaC1 LIG_HEM 14 7 0 0.5 0.06 0 1.5
1mw5A0 1bgyC2 LIG_HEM 13 9 0 0.8 0.04 0 1.8
1vhqA0 1ct9A1 LIG_AMP 13 8 13.9 1.2 0.02 0 1.8
1vhsB0 1cjwA1 LIG_COT 13 9 13.1 0.6 0.45 3.2 5.6
1vhsA0 1qsmD1 LIG_ACO 12 9 11.9 35.3 0.41 2.3 4.7
1j2rC0 19hcA1 LIG_HEM 12 8 12.7 0.4 0.01 0 1.5
1oz9A0 1fy7A1 LIG_COA 11 7 14.6 1.5 0.04 0 0.4
1vimA0 1dqrA1 LIG_6PG 10 7 15.4 1.1 0.07 0 3.1
1vj1A0 1tsdA1 LIG_UMP 10 6 13.9 3.2 0.01 0 0.2
1rtyA0 1fps_1 POLYPRENYL_SYNTHET_2 10 7 16.2 15 0.05 0 3.6
1vhnA0 2dorA1 DHODEHASE_2 10 8 13.5 0.7 0.23 0 5.7
1vhkA0 1grj_1 GREAB_1 10 7 0 2.5 0.02 0 2.2
1k7kA0 1qd1B1 LIG_FON 10 6 12.7 1 0.03 0 0.5
1vhkC0 1qd1B1 LIG_FON 10 6 13.5 2.1 0.04 0 2.5
1vhcA0 1bmtA2 LIG_COB 10 8 15 3.7 0.06 0 2
1uf9A0 1esmA1 LIG_COA 10 8 13.9 0.4 0.11 0 4.2
1h2hA0 1ezfA1 SQUALEN_PHYTOEN_SYN_1 10 7 13.1 1.3 0.02 0 0.4
1j5pA0 1ezfA1 SQUALEN_PHYTOEN_SYN_1 10 7 13.1 1.5 0.02 0 1
1rcuB0 2tpsB1 LIG_TPS 10 7 13.9 4.2 0.08 0 1.8
1vhcA0 2tpsB1 LIG_TPS 10 7 16.2 7.9 0.32 0.1 4.2
1jriC0 1atiA1 AA_TRNA_LIGASE_II_1 10 6 14.2 6.6 0.02 0 2.2
1j9jA0 1ft1A6 PPTA 10 7 14.2 1.9 0.02 0 0.6
1j9kB0 1ft1A6 PPTA 10 7 14.2 1.9 0.01 0 0.7
1i36A0 1eluA5 LIG_PDA 9 6 13.9 0.5 0.04 0 1.2
1j6pA0 1bxoA1 ASP_PROTEASE 9 6 13.9 2.7 0.02 0 0.3
1p5fA0 1eyrA1 LIG_CDP 9 6 21.6 33.2 0.06 0 1.9
1kytA0 1drmA1 LIG_HEM 9 6 12.3 0.9 0.02 0 1.6
1l6rB0 1drmA1 LIG_HEM 9 6 0 0.9 0.02 0 0.8
1j6rA0 1pprM1 LIG_DGD 9 6 0 3.7 0.01 0 1.4
1p99A0 1dik_1 LIG_SO4 9 6 14.2 1.7 0.07 0 0.7
1j2rD0 1dbtA1 OMPDECASE 9 6 15 2.5 0.07 0 1.6
1ni9A0 1pkp_1 RIBOSOMAL_S5 9 6 15.4 18.8 0.03 0 2.6
1lxnA0 1eg7A4 FTHFS_1 9 6 13.5 3.4 0.02 0 2.1
1rtyA0 1cpcB2 LIG_CYC 8 6 0 3.3 0.06 0 0.9
1vhnA0 1rblA1 LIG_CAP 8 6 14.2 1.5 0.09 0 2.9
1rtyA0 2cmd_1 MDH 8 6 13.9 19.8 0.02 0 0.9
1vj1A0 1hdoA1 LIG_NAP 8 6 14.6 3.2 0.07 0 3.3
1nc5A0 1aorA1 LIG_PTE 8 6 14.6 0.5 0.01 0 0.5
1rtwA0 1ft1A2 PPTA 8 6 13.1 11.7 0.03 0 1.7
1pg6A0 1qs0A1 LIG_TDP 8 6 13.5 0.2 0.02 0 0.9
1vizA0 1ho4B1 LIG_PXP 8 6 13.9 0.2 0.02 0 4.4
1l5xA0 1knyA1 LIG_APC 8 5 0 9.7 0.03 0 1.5
1vh6B0 1b72B1 HOMEOBOX_1 8 5 0 21.6 0.06 0.4 2.9
1mwqB0 19hcA1 CYTOCHROME_C 8 6 0 3.5 0.02 0 0.8
1s0uA0 1tplA1 BETA_ELIM_LYASE 8 6 15 0.6 0.02 0 2.6
1ixlA0 1ksaA1 LIG_BCL 8 6 0 2.2 0.04 0 1.9
1ufaA0 1nstA1 LIG_A3P 8 6 15.8 2.3 0.02 0 0.7
1rvkA0 2mnr_1 LIG__MN 8 6 14.2 39.5 0.05 9.3 9.8
1rvkA0 2mnr_1 MR_MLE_2 8 6 14.2 39.5 0.05 9.3 9.8
1vh6A0 1rdzA2 LIG_AMP 8 6 13.1 1.7 0.02 0 1
1ns5A0 1qjbB4 LIG_SEP 8 5 0 0.8 0.02 0 1.6
1rtyA0 1bcfA1 BACTERIOFERRITIN 8 6 16.2 7.3 0.14 0 0.9
1vi3A0 1a44_2 PBP 8 6 38.9 31.7 0.24 1 4.7
1j74A0 1dat_1 FERRITIN_1 8 6 15.8 5 0.00 0 0
1j7dA0 1dat_1 FERRITIN_1 8 6 15.8 0.7 0.00 0 0
1pc6A0 1qq8A1 HEME_OXYGENASE 8 6 0 0.9 0.03 0 1
1htwA0 1a4sA1 ALDEHYDE_DEHYDR_GLU 7 6 15 1.7 0.04 0 2
1vhmA0 1f5mB1 UPF0067 7 6 120 52.7 0.64 10 9.3
1vhmB0 1f5mB1 UPF0067 7 6 121 53.3 0.63 11.6 10.1
1rvkA0 2mnr_4 MR_MLE_1 7 6 14.2 39.5 0.05 9.3 9.8
1j6oA0 1e5dA4 LIG_FEO 7 6 14.2 0.3 0.03 0 0.1
1vhmA0 9icwA8 DNA_POLYMERASE_X 7 6 0 0.8 0.03 0 1.5
1qyiA0 2scpA1 EF_HAND 7 6 0 5.7 0.02 0 1.1
1nkvA0 1dhs_2 LIG_NAD 7 6 0 2 0.04 0 0.3
1nigA0 1c8zA1 TUB_2 7 6 0 0.7 0.01 0 4
1gpqB0 1bif_1 ATP_GTP_A 7 6 0 2.5 0.03 0 1
1p9vA0 1cjcA1 LIG_FAD 7 6 14.2 3.5 0.01 0 0.6
1vhmA0 1cjcA1 LIG_FAD 7 6 14.6 0.5 0.02 0 1
1vhmB0 1cjcA1 LIG_FAD 7 6 14.6 0.7 0.01 0 0.2
1lqlA0 1i78A5 OMPTIN_2 7 7 0 2.2 0.03 0 0
Acknowledgements
The authors thank Gianni Cesareni and Arthur Lesk for helpful support and discussion. We gratefully acknowledge the support of Telethon GGP04273, GENEFUN, a PNR 2001–2003 (FIRB art.8) and a PNR 2003–2007 (FIRB art.8).
==== Refs
Shapiro L Harris T Finding function through structural genomics Curr Opin Biotechnol 2000 11 31 35 10679341 10.1016/S0958-1669(99)00064-6
Whisstock JC Lesk AM Prediction of protein function from protein sequence and structure Q Rev Biophys 2003 36 307 340 15029827 10.1017/S0033583503003901
Fischer D Norel R Wolfson H Nussinov R Surface motifs by a computer vision technique: searches, detection, and implications for protein-ligand recognition Proteins 1993 16 278 292 8394000 10.1002/prot.340160306
Norel R Fischer D Wolfson HJ Nussinov R Molecular surface recognition by a computer vision-based technique Protein Eng 1994 7 39 46 8140093
Kauvar LM Villar HO Deciphering cryptic similarities in protein binding sites Curr Opin Biotechnol 1998 9 390 394 9751636 10.1016/S0958-1669(98)80013-X
Lesk AM Fordham WD Conservation and variability in the structures of serine proteinases of the chymotrypsin family J Mol Biol 1996 258 501 537 8642605 10.1006/jmbi.1996.0264
Fischer D Wolfson H Lin SL Nussinov R Three-dimensional, sequence order-independent structural comparison of a serine protease against the crystallographic database reveals active site similarities: potential implications to evolution and to protein folding Protein Sci 1994 3 769 778 8061606
Contreras JA Karlsson M Osterlund T Laurell H Svensson A Holm C Hormone-sensitive lipase is structurally related to acetylcholinesterase, bile salt-stimulated lipase, and several fungal lipases. Building of a three-dimensional model for the catalytic domain of hormone-sensitive lipase J Biol Chem 1996 271 31426 31430 8940153 10.1074/jbc.271.49.31426
Kobayashi N Go N ATP binding proteins with different folds share a common ATP-binding structural motif Nat Struct Biol 1997 4 6 7 8989312 10.1038/nsb0197-6
Via A Ferre F Brannetti B Valencia A Helmer-Citterich M Three-dimensional view of the surface motif associated with the P-loop structure: cis and trans cases of convergent evolution J Mol Biol 2000 303 455 465 11054283 10.1006/jmbi.2000.4151
Hwang KY Chung JH Kim SH Han YS Cho Y Structure-based identification of a novel NTPase from Methanococcus jannaschii Nat Struct Biol 1999 6 691 696 10404228 10.1038/10745
Wistow G Piatigorsky J Recruitment of enzymes as lens structural proteins Science 1987 236 1554 1556 3589669
Holm L Sander C An evolutionary treasure: unification of a broad set of amidohydrolases related to urease Proteins 1997 28 72 82 9144792 10.1002/(SICI)1097-0134(199705)28:1<72::AID-PROT7>3.0.CO;2-L
Ganfornina MD Sanchez D Generation of evolutionary novelty by functional shift Bioessays 1999 21 432 439 10376014 10.1002/(SICI)1521-1878(199905)21:5<432::AID-BIES10>3.0.CO;2-T
Todd AE Orengo CA Thornton JM Plasticity of enzyme active sites Trends Biochem Sci 2002 27 419 426 12151227 10.1016/S0968-0004(02)02158-8
Laskowski RA Luscombe NM Swindells MB Thornton JM Protein clefts in molecular recognition and function Protein Sci 1996 5 2438 2452 8976552
Kleywegt GJ Recognition of spatial motifs in protein structures J Mol Biol 1999 285 1887 1897 9917419 10.1006/jmbi.1998.2393
Rosen M Lin SL Wolfson H Nussinov R Molecular shape comparisons in searches for active sites and functional similarity Protein Eng 1998 11 263 277 9680188 10.1093/protein/11.4.263
Preissner R Goede A Rother K Osterkamp F Koert U Froemmel C Matching organic libraries with protein-substructures J Comput Aided Mol Des 2001 15 811 817 11776292 10.1023/A:1013158818807
Kinoshita K Nakamura H Identification of protein biochemical functions by similarity search using the molecular surface database eF-site Protein Sci 2003 12 1589 1595 12876308 10.1110/ps.0368703
Schmitt S Kuhn D Klebe G A new method to detect related function among proteins independent of sequence and fold homology J Mol Biol 2002 323 387 406 12381328 10.1016/S0022-2836(02)00811-2
Binkowski TA Adamian L Liang J Inferring functional relationships of proteins from local sequence and spatial surface patterns J Mol Biol 2003 332 505 526 12948498 10.1016/S0022-2836(03)00882-9
Stark A Sunyaev S Russell RB A model for statistical significance of local similarities in structure J Mol Biol 2003 326 1307 1316 12595245 10.1016/S0022-2836(03)00045-7
Stark A Russell RB Annotation in three dimensions. PINTS: Patterns in Non-homologous Tertiary Structures Nucleic Acids Res 2003 31 3341 3344 12824322 10.1093/nar/gkg506
Stark A Shkumatov A Russell RB Finding functional sites in structural genomics proteins Structure (Camb) 2004 12 1405 1412 15296734 10.1016/j.str.2004.05.012
Singh R Saha M Identifying structural motifs in proteins Pac Symp Biocomput 2003 228 239 12603031
Chen BY Fofanov VY Kristensen DM Kimmel M Lichtarge O Kavraki LE Algorithms for structural comparison and statistical analysis of 3D protein motifs Pac Symp Biocomput 2005 334 345 15759639
Schmollinger M Fischer I Nerz C Pinkenburg S Gotz F Kaufmann M Lange KJ Reuter R Rosenstiel W Zell A ParSeq: searching motifs with structural and biochemical properties Bioinformatics 2004 20 1459 1461 14962930 10.1093/bioinformatics/bth083
Torrance JW Bartlett GJ Porter CT Thornton JM Using a library of structural templates to recognise catalytic sites and explore their evolution in homologous families J Mol Biol 2005 347 565 581 15755451 10.1016/j.jmb.2005.01.044
Wangikar PP Tendulkar AV Ramya S Mali DN Sarawagi S Functional sites in protein families uncovered via an objective and automated graph theoretic approach J Mol Biol 2003 326 955 978 12581652 10.1016/S0022-2836(02)01384-0
Pal D Eisenberg D Inference of protein function from protein structure Structure (Camb) 2005 13 121 130 15642267 10.1016/j.str.2004.10.015
Ferre F Ausiello G Zanzoni A Helmer-Citterich M SURFACE: a database of protein surface regions for functional annotation Nucleic Acids Res 2004 32 D240 244 14681403 10.1093/nar/gkh054
Falquet L Pagni M Bucher P Hulo N Sigrist CJ Hofmann K Bairoch A The PROSITE database, its status in 2002 Nucleic Acids Res 2002 30 235 238 11752303 10.1093/nar/30.1.235
Puntervoll P Linding R Gemund C Chabanis-Davidson S Mattingsdal M Cameron S Martin DM Ausiello G Brannetti B Costantini A ELM server: A new resource for investigating short functional sites in modular eukaryotic proteins Nucleic Acids Res 2003 31 3625 3630 12824381 10.1093/nar/gkg545
Camon E Magrane M Barrell D Lee V Dimmer E Maslen J Binns D Harte N Lopez R Apweiler R The Gene Ontology Annotation (GOA) Database: sharing knowledge in Uniprot with Gene Ontology Nucleic Acids Res 2004 32 D262 266 14681408 10.1093/nar/gkh021
Alberts IL Nadassy K Wodak SJ Analysis of zinc binding sites in protein crystal structures Protein Sci 1998 7 1700 1716 10082367
Tainer JA Roberts VA Getzoff ED Protein metal-binding sites Curr Opin Biotechnol 1992 3 378 387 1368439 10.1016/0958-1669(92)90166-G
Barondeau DP Getzoff ED Structural insights into protein-metal ion partnerships Curr Opin Struct Biol 2004 14 765 774 15582401 10.1016/j.sbi.2004.10.012
Andreeva A Howorth D Brenner SE Hubbard TJ Chothia C Murzin AG SCOP database in 2004: refinements integrate structure and sequence family data Nucleic Acids Res 2004 32 D226 229 14681400 10.1093/nar/gkh039
McCarthy AA Baker HM Shewry SC Patchett ML Baker EN Crystal structure of methylmalonyl-coenzyme A epimerase from P. shermanii: a novel enzymatic function on an ancient metal binding scaffold Structure (Camb) 2001 9 637 646 11470438 10.1016/S0969-2126(01)00622-0
Fraaije MW Mattevi A Flavoenzymes: diverse catalysts with recurrent features Trends Biochem Sci 2000 25 126 132 10694883 10.1016/S0968-0004(99)01533-9
Dym O Eisenberg D Sequence-structure analysis of FAD-containing proteins Protein Sci 2001 10 1712 1728 11514662 10.1110/ps.12801
Tatusova TA Madden TL BLAST 2 Sequences, a new tool for comparing protein and nucleotide sequences FEMS Microbiol Lett 1999 174 247 250 10339815 10.1016/S0378-1097(99)00149-4
Krissinel E Henrick K Protein structure comparison in 3D based on secondary structure matching (SSM) followed by Ca alignment, scored by a new structural similarity function Proceedings of the 5th International Conference on Molecular Structural Biology, Vienna, September 3–7 2003 2003 88
Teichmann SA Murzin AG Chothia C Determination of protein function, evolution and interactions by structural genomics Curr Opin Struct Biol 2001 11 354 363 11406387 10.1016/S0959-440X(00)00215-3
Chance MR Bresnick AR Burley SK Jiang JS Lima CD Sali A Almo SC Bonanno JB Buglino JA Boulton S Structural genomics: a pipeline for providing structures for the biologist Protein Sci 2002 11 723 738 11910018 10.1110/ps.4570102
Todd AE Orengo CA Thornton JM Evolution of protein function, from a structural perspective Curr Opin Chem Biol 1999 3 548 556 10508675 10.1016/S1367-5931(99)00007-1
Thornton JM Todd AE Milburn D Borkakoti N Orengo CA From structure to function: approaches and limitations Nat Struct Biol 2000 7 991 994 11104008 10.1038/80784
Irving JA Whisstock JC Lesk AM Protein structural alignments and functional genomics Proteins 2001 42 378 382 11151008 10.1002/1097-0134(20010215)42:3<378::AID-PROT70>3.0.CO;2-3
Ausiello G Zanzoni A Peluso D Via A Helmer-Citterich M pdbFun: mass selection and fast comparison of annotated PDB residues Nucleic Acids Research 2005 33 W133 7 15980442 10.1093/nar/gki499
Laskowski RA SURFNET: a program for visualizing molecular surfaces, cavities, and intermolecular interactions J Mol Graph 1995 13 323 330 307–328 8603061 10.1016/0263-7855(95)00073-9
Berman HM Battistuz T Bhat TN Bluhm WF Bourne PE Burkhardt K Feng Z Gilliland GL Iype L Jain S The Protein Data Bank Acta Crystallogr D Biol Crystallogr 2002 58 899 907 12037327 10.1107/S0907444902003451
Gattiker A Bienvenut WV Bairoch A Gasteiger E FindPept, a tool to identify unmatched masses in peptide mass fingerprinting protein identification Proteomics 2002 2 1435 1444 12422360 10.1002/1615-9861(200210)2:10<1435::AID-PROT1435>3.0.CO;2-9
Hubbard S Thornton JM NACCESS, Computer Program Department of Biochemistry and Molecular Biology 1993 University College London
Hubbard SJ Campbell SF Thornton JM Molecular recognition. Conformational analysis of limited proteolytic sites and serine proteinase protein inhibitors J Mol Biol 1991 220 507 530 1856871 10.1016/0022-2836(91)90027-4
Coutsias EA Seok C Dill KA Using quaternions to calculate RMSD J Comput Chem 2004 25 1849 1857 15376254 10.1002/jcc.20110
Schwartz R Dayhoff M Matrices for detecting distant relationships 1979 Foundation NBR. Washington DC
Boeckmann B Bairoch A Apweiler R Blatter MC Estreicher A Gasteiger E Martin MJ Michoud K O'Donovan C Phan I The SWISS-PROT protein knowledgebase and its supplement TrEMBL in 2003 Nucleic Acids Res 2003 31 365 370 12520024 10.1093/nar/gkg095
Rice P Longden I Bleasby A EMBOSS: the European Molecular Biology Open Software Suite Trends Genet 2000 16 276 277 10827456 10.1016/S0168-9525(00)02024-2
|
16076399
|
PMC1190158
|
CC BY
|
2021-01-04 16:27:24
|
no
|
BMC Bioinformatics. 2005 Aug 2; 6:194
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-194
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1971607899810.1186/1471-2105-6-197SoftwareSquid – a simple bioinformatics grid Carvalho Paulo C [email protected]ória Rafael V [email protected] Miranda Antonio B [email protected] Wim M [email protected] Laboratory for Functional Genomics and Bioinformatics, Oswaldo Cruz Institute, Fiocruz, Rio de Janeiro, Brazil2005 3 8 2005 6 197 197 18 2 2005 3 8 2005 Copyright © 2005 Carvalho et al; licensee BioMed Central Ltd.2005Carvalho et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
BLAST is a widely used genetic research tool for analysis of similarity between nucleotide and protein sequences. This paper presents a software application entitled "Squid" that makes use of grid technology. The current version, as an example, is configured for BLAST applications, but adaptation for other computing intensive repetitive tasks can be easily accomplished in the open source version. This enables the allocation of remote resources to perform distributed computing, making large BLAST queries viable without the need of high-end computers.
Results
Most distributed computing / grid solutions have complex installation procedures requiring a computer specialist, or have limitations regarding operating systems. Squid is a multi-platform, open-source program designed to "keep things simple" while offering high-end computing power for large scale applications. Squid also has an efficient fault tolerance and crash recovery system against data loss, being able to re-route jobs upon node failure and recover even if the master machine fails. Our results show that a Squid application, working with N nodes and proper network resources, can process BLAST queries almost N times faster than if working with only one computer.
Conclusion
Squid offers high-end computing, even for the non-specialist, and is freely available at the project web site. Its open-source and binary Windows distributions contain detailed instructions and a "plug-n-play" instalation containing a pre-configured example.
==== Body
Background
Bioinformatics includes some highly repetitive and computing intensive applications, such as comparison of nucleotide or peptide sequences in search for similarities. The BLAST algorithm (Basic Local Alignment and Search Tool) is well known for its performance [1]. Even though BLAST is "fast", it is an increasingly time-consuming operation when many sequences are to be queried against large databases.
Web pages offering BLAST capabilities are limited in the number of query sequences and available databases to search, while local facilities can easily get overloaded due to limited computing resources when dealing with data intensive operations in smaller research centers. Grid computing permits usage of idle resources and is an inexpensive alternative to large multiprocessor machines or dedicated clusters.
"A computational grid is a hardware and software infrastructure that provides dependable, consistent, pervasive and inexpensive access to high-end computational capabilities" [2]. There are high-quality grid initiatives such as Globus [3], N1 Grid Engine 6, MyGrid [4] and EUROGRID [5] among others. Such software acts as middleware interlinking resources of multiple computers within or between institutions using open and general purpose protocols for secure high performance distributed computing. Although some of these solutions are very complete, their installation is quite complex, and may require specialist management, or have limitations regarding the operating system. Currently, a few non-commercial packages specialized in performing distributed BLAST are available. Among them stand out S-BLAST [6], BeoBlast [7], CondorBLAST [8], Soap-HT-BLAST [9], W.ND BLAST [10], and mpiBLAST [11]. However, only SOAP- HT-BLAST and W.ND BLAST present GUIs, and the latter is only Windows based. Thus a few drawbacks for these solutions are: GUI absence, the need for extensive networking skill for implementation, low fault tolerance [10] and not being multi-platform, resulting in limited utility for the end-user.
In this work a simple, low-cost, platform independent, easy to install and use computational grid interface was constructed and applied to BLAST implementation. This environment called Squid is based on TCP/IP [12], automatically manages available computing resources and makes large BLAST queries possible, even for small laboratories with limited computing resources.
Implementation
The Squid environment is programmed in Perl (version 5.8.6) and is composed of two programs, Squid and Tentacle, each having their configuration file in text format, which can be edited by the user to reflect the local configuration. The graphic user interface (GUI – images can be seen at project web site) can be used to view the currently available computing resources, select query sequence files and choose databases to confront. Squid can also be fully controlled through the command prompt allowing experienced users to encapsulate it within other programs and pipelines. Each grid node must have a copy of the BLAST and Tentacle programs installed and, ideally, a copy of the database. Although remote DB copies can be accessed, this can heavily overburden the connecting network and degrade performance. Squid can be installed under Linux, Unix, Mac OS and Windows operating systems. As soon as the user submits a job, Squid will test which of the nodes are up and properly configured to receive jobs. Squid also timely checks the availability of new up-nodes.
The user-submitted file containing the query sequences is split into smaller files called "work fragments" with user preconfigured size (ex. a file with 10000 query sequences can be split in 200 files with 50 query sequences each), and kept in a work directory. By knowing the up time and availability of nodes, fragment size can be adapted to best suit various working environments. Each fragment is sent to an available node for BLAST execution. Only when the BLAST results file is successfully returned, the respective fragment file is removed from the work directory. Squid will continue to send jobs to nodes until there are no more fragments located in the work directory. This approach is a simple yet highly efficient job control and crash recovery system.
Tentacle is responsible for a single nodes internal administration. The communication semantics consists in receiving a command, (i.e. blast, reset_node, erase_work_files, authenticate, etc.), followed by required complementary data. The command is validated, processed and an answer is always sent back to the central administration node running Squid.
Squids' remote node administration core works by managing three lists: the up_node (nodes that are ready to receive work), down_node (nodes that are not responding), and busy_node (nodes that are currently processing). After creation of the "work fragments" in the work directory, node classification as belonging to the up_node or down_node list is performed. This is quickly accomplished by sending preset messages to every remote node and validating if the proper response is returned, much like the ping command in most operational systems.
Successfully validated nodes are added to the up_node list. Subsequently, "work fragment" files are selected from the work directory and sent to nodes in the up_node list, together with the blast command. The receiving nodes are moved from the up_node list to the busy_node list. Squid continues to send jobs and manage them in different threads until no more nodes are available in the up_node list. When a BLAST result is fully received from a remote node, it is written to disk and its corresponding work fragment is then deleted from the work directory. Finally the node is changed from the busy_node to the up_node list. It is important to note that a small overhead occurs for every communication established, therefore larger work fragments can result in less overhead.
To check if processing nodes are still up, validation commands are timely sent. If validation fails, indicating that the node has stopped serving the grid, it is immediately removed from the up_node or busy_node list and placed in the down_node list. A thread timely sends validation requests to nodes in the down_node list. If the node responds again, a node reset command is sent, clearing local work files, and it is again moved to the up_node list. The "lost fragment" will be re-routed to a new node, since Squid only erases the "work fragment" from the work directory after receiving and saving its complete BLAST result.
A fault recovery routine is also implemented to handle occasions where the main node goes down. Before reactivating Squid, the crash recovery button in the GUI can be clicked. This makes Squid jump the routine where the initial FASTA file is read and fragmented into "work fragments" that are placed in the work directory; thus only the work fragments that haven't been processed will remain. Squid will once again go through node validation and pick up right from where it left.
Data can be lost in intermediate networks or in unstable connections, but the TCP/IP communication protocol is capable of detecting such errors and automatically trigger retransmission until data is correctly and completely received. Squid's node communication is fully based on TCP/IP addressing and data transfer verification through a user configured port. Even though Squid authenticates a remote node before receiving input, enhanced security can be obtained by setting up a virtual private network (VPN) when working across unsafe networks. A VPN performs data tunnelling (making sure that it cannot be intercepted) and encryption; linking it with Squid should guarantee a secure and reliable data transmission, especially when sensitive data is involved. Since there is always a loss in performance when using encrypted data, such degree of safety should be evaluated.
Results and conclusion
Squid is designed to "keep things simple" offering grid power for large scale applications (BLAST in the current configuration) for smaller labs, so the user gets to worry about analyzing results while Squid worries about distributed computing and getting the job done. Squid stands out among other software because it can simultaneously work with Windows / Linux nodes and efficiently manage job control but above all, it is meant to be user friendly. One can subsequently use BioParser, also available at our lab page to further process large BLAST outputs. Various tests were carried out to analyze Squids' performance and robustness where the following should be noted:
1. A grid containing N nodes is able to execute multiple BLAST queries almost N times faster than if working with only one node.
2. Overhead occurs due to computer communication, network latency and initiating new computer processes. Thus the size of fragment files should not be too small.
3. For maximum performance, each node should have a copy of the database, but remote copies can also be used.
4. Squid successfully handles problems such as unexpected remote node shutdown or even main node accidental shut down. Squid picks up right from were it left.
Every time a node is available for a job, Squid sends it a work fragment. If the node stops serving the grid while processing, its uncompleted work file will be eliminated and Squid will eventually re-route the "skipped fragment" to another available node. Being so, by knowing the up time and availability of nodes, fragment size can be adapted to best suit for various working environments.
Further help and instructions are included within the distribution. Open source and binary Windows versions also come with a pre-configured example for evaluation purposes.
Availability and requirements
• Project name: Squid – A simple bioinformatics grid
• Project home page: Squid is available for download at the projects website [13]. Once in the site select softwares and Squid to view the project page and download links.
• Operating system(s): Platform independent
• Programming language: Perl 5.8.6
• License: Creative Commons – Commons Deed [14].
• Any restrictions to use by non-academics: license needed
Authors' contributions
PCC performed software engineering, coding, elaboration of manuscript. RVG, ABM and WD performed software testing, benchmarks, helped in GUI coding, debugging and manual / manuscript revisions.
Acknowledgements
We thank CNPq, FAPERJ, CYTED-RIB, LACBioNet, RNP-GIGA, FIOCRUZ-PAPES/PDTIS, IOC for financial support.
==== Refs
Altschul SF Warren G Webb M Eugene WM David JL Basic local alignment search tool J Mol Biol 1990 215 403 10 2231712 10.1006/jmbi.1990.9999
Foster Ian What is the Grid? A Three Point Checklist 2002
Foster I Kesselman C The Globus Project: A Status Report. I Proc IPPS/SPDP'98 Heterogeneous Computing Workshop 1998 4 18
MyGrid
EUROGRID
S-BLAST: Federated BLAST Using Sorcer
Grant JD Dunbrack RL Manion FJ Ochs MF BeoBLAST: distributed BLAST and PSI-BLAST on a Beowulf cluster Bioinformatics 2002 18 765 6 12050075 10.1093/bioinformatics/18.5.765
Condor BLAST
Soap-HT-BLAST
Dowd SC Zaragoza J Rodriguez JR Oliver MJ Payton PR Windows .NET Network Distributed Basic Local Alignment Search Toolkit (W.ND-BLAST) BMC Bioinformatics 2005 6 93 15819992 10.1186/1471-2105-6-93
Darling A Carey L Feng W The Design, Implementation, and Evaluation of mpiBLAST ClusterWorld 2003 conference 2003
Introduction to TCP/IP
Squid project homepage
Creative Commons – Commons Deed
|
16078998
|
PMC1190159
|
CC BY
|
2021-01-04 16:27:25
|
no
|
BMC Bioinformatics. 2005 Aug 3; 6:197
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-197
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-2021609553810.1186/1471-2105-6-202Methodology ArticlePALSSE: A program to delineate linear secondary structural elements from protein structures Majumdar Indraneel [email protected] S Sri [email protected] Nick V [email protected] Howard Hughes Medical Institute, University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd. Dallas, TX 75390, USA2 Department of Biochemistry, University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX 75390, USA2005 11 8 2005 6 202 202 28 1 2005 11 8 2005 Copyright © 2005 Majumdar et al; licensee BioMed Central Ltd.2005Majumdar et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The majority of residues in protein structures are involved in the formation of α-helices and β-strands. These distinctive secondary structure patterns can be used to represent a protein for visual inspection and in vector-based protein structure comparison. Success of such structural comparison methods depends crucially on the accurate identification and delineation of secondary structure elements.
Results
We have developed a method PALSSE (Predictive Assignment of Linear Secondary Structure Elements) that delineates secondary structure elements (SSEs) from protein Cα coordinates and specifically addresses the requirements of vector-based protein similarity searches. Our program identifies two types of secondary structures: helix and β-strand, typically those that can be well approximated by vectors. In contrast to traditional secondary structure algorithms, which identify a secondary structure state for every residue in a protein chain, our program attributes residues to linear SSEs. Consecutive elements may overlap, thus allowing residues located at the overlapping region to have more than one secondary structure type.
Conclusion
PALSSE is predictive in nature and can assign about 80% of the protein chain to SSEs as compared to 53% by DSSP and 57% by P-SEA. Such a generous assignment ensures almost every residue is part of an element and is used in structural comparisons. Our results are in agreement with human judgment and DSSP. The method is robust to coordinate errors and can be used to define SSEs even in poorly refined and low-resolution structures. The program and results are available at .
==== Body
Background
Protein secondary structure was first predicted on the basis of stereo-chemical principles to adopt α-helical and β-strand conformations due to the periodicity of their inter-backbone hydrogen bonds [1]. Subsequent experimental determination of protein structures confirmed these predictions revealing the presence of α-helices and β-strands as the predominant secondary structure elements (SSEs) in proteins. Other minor SSEs such as 310-helix [2], π-helix [3], β-turns (type I-IV) [4], γ-turns [5], and β-bulges [6] have been defined based on the stereochemistry of the polypeptide chain. However, deviations from ideal geometry in experimentally determined structures due to interactions between the elements and possible errors in coordinates can make an algorithmic definition that is consistent with manual assignment challenging.
The first algorithm for the automatic delineation of secondary structure was proposed by Levitt and Greer [7]. They defined secondary structures based on peptide hydrogen bonds (using i, i+3 Cα distances and i, i+1, i+2, i+3 Cα torsion angles). A more comprehensive algorithm, DSSP, was subsequently developed and is based on a careful analysis of backbone-backbone hydrogen bond energies and geometrical features of the polypeptide chain [8]. It is very accurate in its residue-based definition of the available coordinates and does not attempt to interpret the data in any predictive way. Another well-known secondary structure definition program is STRIDE [9]. Like DSSP, this program also uses hydrogen bond energy as well as main chain dihedral angles φ and ψ. In addition, it relies on a database of derived recognition parameters that uses the crystallographer's definition of secondary structures as the standard. Other programs to assign the secondary structure states of proteins exist. Xtlsstr defines secondary structure in the same way a person assigns secondary structure visually [10]. It uses two angles and three distances computed from the protein backbone atoms. Sstruc is a reimplementation of the classic DSSP method (Smith DK, Thornton J; unpublished). The P-Curve algorithm allows the helicoidal structure of a protein to be calculated starting from the atomic coordinates of its peptide backbone [11].
Secondary structures in experimentally determined protein coordinate data often deviate from the ideal geometry and thus methods of secondary structure assignment that use different logic and cutoffs can vary significantly in their assignments. Maximum variation is seen near the edges of SSEs and consensus secondary structures defined by different algorithms have been proposed in order to define SSEs accurately [12]. Attempts have been made to define secondary structures consistently and in agreement with visual inspection by recognizing errors in protein coordinates. In the Stick algorithm, line segments become the primary data elements and can then be used to define secondary structure. By contrast, previous approaches have used secondary structure definitions to specify line segments [13]. Our algorithm retains the former approach, as Cα geometry allows locating breakpoints in both α-helices and β-strands and allows generation of residue based pairing information helpful in determining edges of β-sheets.
A strong correlation exists between hydrogen bonding patterns and Cα distances and torsion angles [14]. Algorithms such as those described by Levitt [7], DEFINE_S [15], VOTAP [16] and P-SEA [17] assign secondary structure not on the basis of actual hydrogen bonding patterns, but using an interpreted residue pairing based on the Cα coordinates of the protein structure. However, most of these programs assign secondary structure properties to individual residues of a protein chain. For the purpose of vector-based structural similarity searches, a secondary structure definition of the linear segments (elements) that can be used to approximate the protein structure in a simplified form as a set of interacting SSEs is required. Using DSSP [8] assignments to define SSEs for protein structure comparisons may lead to problems in identifying linear elements and element edges [18]. Our analysis of DSSP assignments reveals that they are not well adapted to the definition of SSEs for several reasons. DSSP finds a secondary structure state for every individual residue in a protein chain. When these states are found, the consecutive residues that belong to one state can be unified to form a SSE. Such a strategy of element definition has several disadvantages. First, one might wrongfully unify two consecutive elements into one. Alternatively, if there is a gap in the secondary structural state of a residue inside an element, due to disorder, refinement error or some other irregularity, this element will be unjustifiably split into two. Residue coverage for DSSP assignments is poor when the structure is not well refined or not well ordered and the stringent criteria set for hydrogen bonds are not met. DSSP misses some short helices and β-strands in which hydrogen bonding criteria are violated. Moreover, the edges of elements are not always defined accurately. Thus, although DSSP can identify a secondary structure state as a property of each residue, it not well suited to outline SSEs.
Here we describe a method "Predictive Assignment of Linear Secondary Structure Elements (PALSSE)" to identify SSEs from the three-dimensional protein coordinates. The method is intended as a reliable predictive linear secondary structure definition algorithm that could provide an element-based representation of a protein molecule. Our algorithm is predictive in that it attempts to overlook isolated errors in residue coordinates and is geared towards defining SSEs of proteins relevant to vector-based protein structure comparison. For the purpose of similarity searches, use of just the major SSEs, namely the α-helix and the β-strand that can be approximated by vectors will suffice, as they typically incorporate the majority of the residues in a protein.
Our program delineates α-helices, and β-strands participating in β-sheets. Helices are broadly defined to include right-handed α-helices, 310 helices, π-helices and turns that show a helical propensity. We do not distinguish between these helices since many linear helical elements in proteins combine 2 or more helical types. For example, the first turn of a α-helix might be a 310 helix and the last turn could be a π-helix. If one would like to differentiate between the three types of helices, DSSP [8] or STRIDE [9] should be used. Similarly, the β-strand elements are broadly defined by our program to include β-strands, β-bridges, β-bends and the residues of β-hairpins.
Results and discussion
Brief overview of the algorithm
α-helices and β-strands are the predominant and most distinct types of secondary structures observed in proteins [8,9,19]. Using only Cα coordinates, our algorithm delineates two types of secondary structural elements, namely helices (includes α, 310 and π-helices) and β-strands (includes β-strands, β-bridges, β-bends and the residues of β-hairpins).
Our method was developed for predictive assignment of linear SSEs. Preliminary helix and β-strand categories are assigned to residues, based on i, i+3 Cα distance and i, i+1, i+2, i+3 Cα torsion angle (step 1 in methods). Next, probable helix and β-strand elements are generated by selecting consecutive residues that belong to the same category (helix or strand, step 2). Quadruplets of residues, formed by two pairs of hydrogen-bonded consecutive residues that satisfy criteria of distances and angles (steps 3 and 4), are constructed from residues that do not meet the strict criteria for helix definition in step 1. A quadruplet is the smallest unit for defining potential β-sheets and is formed from a set of four Cα atoms that are linked with two covalent bonds and two pseudo-hydrogen bonds (see step 3 in methods). The quadruplets are joined together, end-to-end in the direction of covalent bonds, to form ladders of consecutive pairs of residues (steps 5, 6, 7). The ladders of paired residues are joined to form paired β-strands. Helices defined previously are split, using root mean square deviation (RMSD) of constituent residues about the helix axis, so that they can be represented as linear elements (step 8). β-strands are split using various geometrical criteria and pairing of neighboring residues (step 9).
The program's main output is in the PDB [20] file format. HELIX and STRAND records are added or substituted by our definition.
Defining and extending core α-helices and β-strands
Cα-Cα distance (i, i+3) and torsion angle (i, i+1, i+2, i+3) are used to select core regions of secondary structures. The parameters are then made less restrictive to identify and assign residues that do not follow the idealized pattern of α-helices and β-strands. Residues that individually might fail the test for a secondary structure state, due to either hydrogen-bonding criteria or φ and ψ angles, or both, may be placed in an element if the geometric parameters and pairing conditions of the neighboring residues support the inclusion. This makes the algorithm predictive in nature; therefore helices and β-strands might be defined in regions that show a helical tendency or have neighboring β-strands respectively, even if the polypeptide model at that region is erroneous. We have found the criteria of a minimum of 3 residues with at least 2 residues pairing with a neighboring β-strand [7] to perform well in identifying β-sheets that match human judgment. Thus, anti-parallel pairs of β-strands with two residues each never arise from loop regions. A method based on quadruplets of residues linked with two covalent bonds and two hydrogen bonds, approximated using a set of three parameters based on distances and angles between the residues, have been used as the seed unit for β-sheets. A quadruplet-based approach has also been used by Levitt [7]. β-strands are defined only when one or more neighboring paired β-strands are available, thus reducing the chance for errors in our predictive definition of β-strands. The smallest helix by our definition has a length of 5 residues, which is a single turn of a α-helix. Our algorithm assigns small 5-residue turns showing helical propensity as helices. Other programs that are Cα-based also fail to distinguish between such turns and helices [21]. As shown in fig. 1a, only a few of these assignments are actually incorrect.
Figure 1 Ramachandran angles from helices and β-strands defined by DSSP and our program. The culled PDB set (described in methods) was used for this calculation. Figs. a and b show the Ramachandran angles obtained respectively from helices and β-strands defined by DSSP. Figs. c and d show the Ramachandran angles obtained respectively from helices and β-strands not defined by DSSP but defined by our program. φ and ψ angles for figs. a and b were obtained from DSSP output. φ and ψ angles for figs. c and d were calculated from output of our algorithm such that φ is torsion angle between residues i-1 and i, and ψ is torsion angle between residues i and i+1 where residues at positions i-1, i and i+1 are part of the same SSE. α, π and 310 helices were used for obtaining data shown in fig. a (DSSP definition 'H', 'I', 'G' respectively). β-Strands were used for obtaining data shown in fig. b (DSSP definition 'E'). Three regions of over-predicted points by our method are shown with an example from each region. Figs. e, f and g show stereo diagrams of parts of three helices respectively from "1b × 4" (chain A, residue 175 in red), "1iom" (chain A, residue 73 in cyan) and "1 × 7d" (chain A, residue 180 in magenta). φ and ψ angles from the residues under study are marked red, cyan and magenta in fig. c. The residues with φ and ψ points highlighted in fig. c are shown as spheres in the same colors in figs. e, f and g.
Definition of linear elements
Our method attempts to provide a definition of SSEs that can be approximated using vectors. It is possible for edges of elements to overlap, leading to more than one secondary structure state for a particular residue. If required, these elements are broken using geometric criteria and directionality changes over a stretch of residues, including pairing for β-strands, to obtain linear elements. Specialized algorithms to detect curved, kinked and linear helices based on local helical twist, rise and virtual torsion angle are known [22]. We have used simpler rule-based methods to improve computational speed. Single residues existing between two bent helices fail cutoffs for Cα distance and torsion angles. Bent helices thus remain separate even though their edges might overlap. Gently curved helices are not broken unless the angle between vectors representing the broken sections is greater than 20°. The broken sections are chosen using an approach (described in methods) to minimize the number of acceptably linear elements, while retaining the maximum number of residues from the original helix. A similar breaking angle of 25° has been observed by Richards and Kundrot [15]. This angle and RMSD are used to break kinked helices correctly. Kinked helices, however, are retained if the angle between their representative vectors is less than 20°.
Reliability of secondary structure definitions
The reliability of secondary structure definition by our algorithm was checked by plotting the main chain torsion angles φ and ψ [23], individually for all helix and β-strand regions defined by DSSP [8] (figs. 1a, 1b) and defined only by our program when compared with DSSP (figs. 1c, 1d). Examples of helices (fig. 1c) and β-strands were inspected manually.
Most of the φ and ψ angles (fig. 1c) are found to be in the regions of the Ramachandran plot commonly accepted to be representative of α-helices [23,24]. An example from the region bounded by -150°<φ-125° and 50<°ψ<100° is shown as a red dot in figs. 1c and 1e. Residues with φ and ψ angles in this region occur mostly at overlaps between two helix elements. Residues with φ and ψ angles in the region -50°<φ<-75° and 100°<ψ<150° (shown as a cyan dot in figs. 1c and 1f) are mostly edge residues of helices. Secondary structure for residues with φ and ψ angles in the -100°<φ<-125° and 125°<ψ<150° region (shown as a magenta dot in figs. 1c and 1g) are not immediately obvious. These are mostly at the edge of helices and could potentially be part of a β-strand or extended region. However, for the examples that we studied, it would not be wrong to define these residues as belonging to part of a helix as no consecutive β-strands were present and the edge residues showed helical tendency. The region of the plot at 50°<φ<100° and -50°<ψ<50° consists of residues taking part in 310-helices.
Nearly all φ and ψ angles for β-strand residues over-predicted by our method (fig. 1d) fall in the same regions as that of DSSP [8] defined β-strands, and most were found to fall in the commonly accepted region of the Ramachandran plot for β-strands [23,24]. Examples of β-strand residues with φ and ψ angles outside this region were inspected manually. A large number of residues with angles in the region -100°<φ<-50° and -50°<ψ<0° were found to be at overlaps between helices and β-strands. Residues with angles in the region 50°<φ<100° and -50°<ψ<50° are mostly bulges and regions of overlap between two β-strand elements. Some residues were found to be included in our element definition but would not be assigned a secondary structure by a residue-based method. These had angles in the region 50°<φ<150° and -180°<ψ<-125° and also 50°<φ<150° and 150°<ψ<180°. These residues are rare in β-strands and were not found to occur successively. They sometimes form part of bulges or element edges.
Residues that fail strict secondary structure assignment when explicit hydrogen-bonding criteria are considered, cannot be used to properly form SSEs. Therefore, predictive assignment with our algorithm is preferred. The following reasons may account for the necessity of predictive assignments. Protein structures are intrinsically flexible. Hydrogen bonds that are present in some family members might be absent in other homologues. Domain interactions, loops, insertions and deletions can all influence the secondary structure around them. Crystal packing and solvent interaction can also account for changes observed in residue coordinates. Further, models based on X-ray data are not always as accurate as they are believed to be [25]. Therefore, the absence of a strictly defined hydrogen bond does not mean that the hydrogen bond is actually absent in the molecule in all its accessible conformations.
Robustness towards coordinate errors
Our algorithm shows a higher degree of robustness than either DSSP [8] or P-SEA [17] when input data contains errors of up to 1.5Å deviation for individual residues (fig. 2a). Secondary structure definition, by DSSP, using main-chain atoms deteriorate sharply when coordinates are non ideal for α-helices or β-strands. Other methods based on recurrence quantifications have been proposed [26] that aim to replicate DSSP definitions while being more robust to coordinate errors. These results are meaningful only at high residue coverage. Our algorithm is able to delineate more helical regions, than DSSP or P-SEA, even when the coordinates are randomly shifted by as much as 2Å (fig. 2b). Close to 80% of all residues are assigned to SSEs by our method. Defining a core of strict α-helix and β-strand forming regions and extending these with neighboring residues to obtain elements ensures that residues are not lost from SSEs when they fail strict definitions, like in low-resolution X-ray and NMR structures.
Figure 2 Secondary structure assignment reliability for DSSP, P-SEA and our program using randomly shifted PDB coordinates. The culled PDB set (described in methods) was used for this calculation. Gaussian random numbers were used to randomly shift coordinates of residues from 0.2Å to 2Å in steps of 0.2Å, in the PDB files. 100 files were generated for every file for every data-point leading to a total of 1,00,000 randomly shifted coordinate files. 2a: Mean and standard error of assignment consistency compared with assignment by the same program on the original coordinates. A percentage match was calculated by comparing definitions for the coordinate shifted file with the program output from actual file on a per residue basis. Means for the percentage match are shown. Standard errors were about 1% in each case (not shown). 2b: Average secondary structure content defined by each program for PDB files at different levels of perturbations are shown. The files used are the same as for fig. 8a. The number of residues assigned as helices or β-strands are shown as a percentage of total residues. Spaces and coils in the program output are counted for calculating percentages. 2c: Percentage of residues over-predicted by each program (DSSP, P-SEA, our method) with respect to the other two is shown. 100 files from the culled PDB set were used for these calculations. 35,670 residues were considered. Results shown are for over-predictions by program names in the column heads when compared with program names in the row heads. Actual number of helices and β-strands assigned by the program are shown on the diagonal (bracketed values).
Keeping the above results in perspective, the amount of secondary structure missed by the two other programs with respect to each of DSSP, P-SEA and our method was studied (fig. 2c). Our program is able to assign about 14% more residues as helices and 13% more residues as β-strands when compared with either DSSP or P-SEA. DSSP assigns less than 1% residues that differ from our definition. P-SEA assigns less than 3% residues that differ from our definition. Most residues defined by DSSP but not assigned by our algorithm did not contribute to long SSEs.
Comparison with other programs
Definitions from our program were compared with assignments by DSSP [8], P-SEA [17], DSSPCont [27], SSTRUC (Smith DK, Thornton J; unpublished), STRIDE [9], DEFINE_S [15], STICK [13] and PROSS [28]. A numerical comparison of residue assignment between secondary structure delineation methods, irrespective of whether they define elements or are residue-based, is not very useful. Element lengths may not be optimum or elements may be too curved for representation by vectors. Residues assigned to α-helix or β-strand regions may not overlap between different programs, thus leading to a wrong comparison if only numbers are used.
In this article, we show two examples of our study (fig. 3), with the secondary structure definitions colored cyan or yellow respectively for helices and β-strands. Two-residue β-strand definitions have not been considered as β-strands for this comparison. Figs. 3a–d are obtained by processing the coordinates of an averaged NMR structure (PDB ID: 1ahk [29]) containing an immunoglobulin sandwich made up of 7 β-strands in two β-sheets. Figs. 3e–h are from a low-resolution (3.0Å) X-ray structure (PDB ID: 1fjg, chain "J" [30]) which has a ferredoxin-like fold made up of two β-α-β units.
Figure 3 Two examples of secondary structure assignment by different programs. We chose an averaged NMR structure "1ahk" and a low-resolution X-ray structure (3.0Å) "1fjg" to show as examples since over-prediction by our method is maximum for such structures. Only chain "J" of "1fjg" is shown. Figs. a, b, c show cartoon diagrams of "1ahk", prepared using MOLSCRIPT [46], for β-strand and α-helix definitions by our program, DSSP [8] and P-SEA [17] respectively. β-Strands are shown in yellow and helices in cyan. The N- and C- termini are marked for each structural diagram. The elements produced by our program are labeled. Fig. d shows the secondary structure assignment for "1ahk" by PALSSE (our program), DSSP, P-SEA, DEFINE_S, STRIDE, SSTRUC and PROSS. Our interpretation of β-strands and helices as defined by the different programs are colored in yellow and cyan respectively. The starting positions of each element labeled in fig. a are shown on the first line. The sequence is numbered on the second line with black letters denoting units, red denoting tenths and blue denoting the hundredths places. The protein sequence is shown in the third line. Figs. e, f, g shows cartoon diagrams of chain "J" of "1fjg", prepared using MOLSCRIPT [46], highlighting β-strand and helix definitions by our program, DSSP, and P-SEA respectively. β-Strands are in yellow and helices are shown in cyan. Elements are labeled in fig. e. Fig. h shows the secondary structure alignment for "J" chain of "1fjg". Definitions produced by the same programs as that used for fig. d are shown. Yellow color is used for our interpretation of β-strands and cyan denotes our interpretation of helices. Green has been used to denote overlaps between helix and β-strand elements defined by our program. The first, second, and third lines show start of each element in fig. e, residue number and sequence respectively, similar to fig d. DSSP, P-SEA, DEFINE_S, SSTRUC, STRIDE and PROSS assignments were generated by obtaining the programs, and then compiling and running them with default parameters on the example PDB files.
Our algorithm shows a marked difference as compared to other programs when low-resolution and NMR structure coordinates are processed. Residue coverage is greater for our definition when compared with DSSP [8] and P-SEA [17], and is also correct with respect to identified elements when used for a similarity search. Fig. 3a shows that our method has been able to correctly identify all 7 β-strands of the immunoglobulin fold, compared to only 2 by DSSP (fig. 3b) and 6 by P-SEA (fig. 3c). Our method has also been able to correctly identify all 4 β-strands and 2 helices in the ferredoxin like fold (fig. 3e), compared to only 2 α-helices and parts of 2 β-strands by DSSP. A vector-based similarity search system will produce incorrect results with the DSSP assignments shown, as the β-strands will be found to be located too far away. Assignment by P-SEA shortens the helix and incorrectly adds a β-strand to the definition. Our interpretation of helix and β-strands from PDB files 1ahk and 1fjg (chain "J") by various programs are respectively colored in figs. 3d and 3h. It is interesting to note the wide variation in assignments of secondary structures by different programs. The only other program that compares favorably in residue coverage to P-SEA and our program is DEFINE_S. However, DEFINE_S sometimes misses helices and wrongly defines parts of helices as β-strands. We have not found the DSSPCont [27] output to be any different from that of DSSP, unless the results are interpreted using probability scores in the output file. STRIDE [9] results are not significantly different from that of DSSP for the examples shown, even though the STRIDE algorithm attempts to improve upon DSSP definitions using a common criterion of hydrogen bonding energy as well as Cα distances. Results from SSTRUC (Smith DK, Thornton J; unpublished) are clearly different and residue coverage is poor when compared with other programs for the average NMR structure "1ahk". Residue coverage and element identification for SSTRUC is similar to that by DSSP and STRIDE for "1fjg". Residue coverage for helix and β-strand definition by PROSS [28] is low and it fails to identify all elements.
Conclusion
The algorithm developed by us can assign linear helix and β-strand SSEs, from only Cα atoms. Our method is predictive in nature and SSEs defined by us can include residues that do not form ideal α-helices and β-strands. Assignments are similar to helix and β-strands defined by a residue-based approach, like DSSP [8], for high-resolution X-ray structures, although our elements include more edge residues. For NMR and low-resolution X-ray structures, our predictive algorithm delineates more SSEs than is possible with a residue-based approach. Elements defined by our method can be approximated using vectors, and we have retained longer elements wherever possible.
This method has been developed for simplified representation of protein structures for similarity searches with other proteins. It should not be used if an accurate residue level definition is necessary. Compared to other programs, we have found our algorithm to perform well in terms of defining linear elements reliably for both helices and β-strands and yet yield a high residue coverage. Visual judgment of results supports our definitions.
The algorithm has been implemented as a computer program "PALSSE" (Predictive Assigner of Linear Secondary Structure Elements). It is written in Python and C and has been tested on the GNU/Linux platform on the i386 processor architecture. The software is available online [31].
Methods
Datasets used for generation of statistics and for expert judgment
The sequences from the SEQRES records of PDB [20] files, current until March 2002, and solved using X-ray diffraction (resolution better than 4.0 Å), and NMR techniques were clustered using BLASTCLUST [32] to obtain clusters with sequence similarity less than 25% and overlap greater than 90%. After screening the resulting chains for errors, format violations, and cases that would prevent either DSSP [8] or our program from processing the file, we obtained a dataset of 2787 polypeptide chains (Statset). This set was used to determine appropriate cutoff values for various parameters in our program.
Algorithm development was monitored by manual inspection of the results produced by the implemented code. For this, a dataset of 295 domains (checkset) consisting of randomly chosen representative structures for every fold in the SCOP database (version 1.63) [33,34] belonging to the 'all α proteins' and 'α and β proteins' classes were chosen.
A set of high-resolution structures (culled PDB set) was used for comparing the final program output with that from other programs, with respect to reliability and robustness towards coordinate errors. For this, a list of the 100 longest non fragmented PDB chains having resolution better than 1.6Å and sequence similarity less than 20% were obtained from the culled PDB database [35].
Programming platform
We used the Python programming language [36] (v2.3) to implement our algorithm. We also used the Biopython [37] Bio. PDB [38] framework to parse the PDB [20] files and store data in its internal data structures. This is a robust method of parsing PDB files and can satisfactorily handle many common problems with parsing NMR models, multiple chains, alternate locations and insertion codes. The "Polypeptide" module enabled handling of chain segments. The included functions were used to store and retrieve data from the constituent objects. Our project extends the Biopython framework to define secondary structures. The implemented code was tested on GNU/Linux [39] systems on the i386 and opteron hardware running Linux i386 kernel. All data plots shown in this article have been prepared using "gnuplot" [40]. All protein structure images were prepared using PyMOL [41] and POV-Ray [42] unless specified otherwise.
Brief overview of methodology
Our method has five major steps that are used to sequentially process the PDB [43] coordinates and assign secondary structure. These are: 1: Assignment of helix and β-strand propensities to individual residues based on i, i+3 Cα distance and i, i+1, i+2, i+3 Cα torsion angle. 2: Delineation of probable core regions of helix and β-strand elements from residues that pass strict criteria. 3: Formation of quadruplets of residues connected by two covalent and two hydrogen bonds as seeding units for β-sheets. 4: Initiation and extension of paired β-strands using quadruplets of paired residues. 5: Breaking consecutive non-single helices and β-strands taking into consideration residue pairing and neighboring elements. Steps of our algorithm need to be run sequentially for the results to be meaningful.
Step 1: Assignment of helix and β-strand propensities to individual residues
Our algorithm first conservatively estimates the propensity of each residue to be part of a secondary structural state (helix, β-strand, both, or none). This is the only step in which our algorithm deals with secondary structure as a property of the individual residue and not as that of an element. Cα coordinates of every residue of the molecule are processed from the N- to C- terminal end and a simple Cα-Cα (i, i+3) distance (fig. 4a) and Cα torsion angle (i, i+1, i+2, i+3) (fig. 4b) are used to mark residues based on whether they have a propensity to assume helical or β-strand conformation. It is possible for a residue to be part of both a helix and a β-strand, or none of them (coil). We decided to use i, i+3 Cα-Cα distances and i, i+1, i+2, i+3 Cα torsion angles as these parameters have been studied extensively [7,8,15,21] with regards to their applicability to secondary structure definition.
Figure 4 Parameters for assignment of helix and β-strand property to individual residues. Cα distance and Cα torsion angle were calculated for defining helices (fig. a) and β-strands (fig. b). The distance between residues i, i+3 (shown joined by a blue line) and torsion angle between residues i, i+1, i+2, i+3 (shown as angle between two colored planes; yellow plane between residues i, i+1, i+2 and orange plane between residues i+1, i+2, i+3) are used to assign loose-helix, strict-helix and loose-strand secondary structure property to individual residues. 4c: Distance between i, i+3 Cα residues from helix and β-strand definitions obtained from DSSP [8] output. Distances were binned in 0.2 Å intervals. Cutoff distance c1 (8.1 Å) is the maximum distance allowed for assigning loose-helix property to a residue. Cutoff c1 is also the minimum distance allowed for assigning loose-strand property to a residue. Residues at i and i+3 positions are allowed in the same SSE template (SSET) only if the cutoff distance c1 passes. Cutoff c2 (6.4 Å) is the maximum i, i+3 Cα distance for strict helix definition. 4d: Torsion angle between i, i+1, i+2, i+3 Cα atoms for helix and β-strand definitions obtained from DSSP output. Angles are binned in 5° intervals. A loose-helix definition is assigned to a residue only if the torsion angle for the residue falls between c1 (-35°) and c2 (115°). A loose-strand is assigned only if the torsion angle is -180° to c1 or c2 to 180°. c3 is the optimal torsion angle for helices and is used to define strict-helix residues if the torsion angle is within a 2 sigma deviation from c3.
The output of DSSP [8] was used to generate cutoff statistics for initial residue-based definition, since it is a conservative program that delineates secondary structures based on hydrogen bonding energy. i, i+3 Cα-Cα distances (fig. 4c) and 4i, i+1, i+2, i+3 Cα torsion angles (fig. 4d) were calculated separately for α-helices and β-strands from DSSP output generated from the "Statset". In this calculation, only residues corresponding to the letters 'H' and 'E' in the 'structure' column of DSSP output were considered as helices and β-strands respectively. π, 310-helices, and β-turns were not used as we decided to initiate secondary structural elements conservatively, using these cutoffs, and then extend them rather than to start from spurious elements for later removal.
DSSP [8] provides assignment for individual residues and does not assemble them into SSEs. We used DSSP data to generate elements, and slightly different approaches were taken for α-helices and β-strands. All consecutive residues belonging to the helical state were considered as part of the same α-helix. For β-strands, DSSP provides bonded pair information (column 'BP1' and 'BP2' in DSSP output) and this was used to locate continuous stretches of paired residues. Each set of these continuous residues was considered as an element. Distance and torsion angle data were calculated from the 'N-' to 'C-' terminal end of the molecule. The data was used to determine cutoff conditions for identifying potential helix and β-strand-rich regions in this step of our algorithm. The following functions were used to decide a particular residue's propensity to form part of helix or β-strand.
δ<8.1 Å, -35°≤τ≤115° ⇒ ρ=loose-helix (1)
δ>8.1 Å, -180°≤τ<-35° | 115°<τ≤180° ⇒ ρ=loose-strand (2)
δ≤6.4 Å, τ within (50.1° ± 2σ) ⇒ ρ=strict-helix (3)
where δ is Cα-Cα distance, τ is Cα torsion angle (i, i+1, i+2, i+3), σ is standard deviation of torsion angle (= 8.6°), ρ is propensity (cutoff values from figs. 4c, d).
Strict-helix residues thus form a subset of the loose-helix residues. This eliminates β-bends from interfering with helix definition even though it also removes π-helices, which have a greater i, i+3 Cα distance than α-helices. Left-handed helices are also removed as they have a lower i, i+1, i+2, i+3 Cα torsion angle than α-helices.
The secondary structure definitions at the end of this step are in agreement with the findings of several other groups who have interpreted secondary structures from just the Cα coordinates [7,15,17,21]. In our program, we use this step as an initial criterion for defining secondary structure states of individual residues. This definition is further refined and extended in subsequent steps of our program to define SSEs. Estimation of secondary structure of individual residues is an important step of the algorithm, however the rest of the algorithm is also designed to correct for errors that may have been introduced at this step.
Step 2: Delineation of probable helix and β-strand elements
Consecutive residues with the same secondary structural propensity are joined to initiate seed-SSEs. Overlapping elements are considered in this step. The overlap can be between elements of the same (H-H, E-E) or different type (H-E, E-H). Since we always process PDB [20] files from the N- to the C- terminal end of the polypeptide for generating cutoff data and also generating Cα distance and torsion angles for this algorithm, extension of seed-SSEs using the distance and torsion angle data was done from the C-terminal end of the seed-SSE towards the C- terminal end of the polypeptide.
We use a group of at least two consecutive loose-helix residues to generate a loose-helix seed-SSE. It is possible to get a single loose-helix residue in β-hairpins whereas a set of consecutive loose-helix residues is more likely to be a part of a helix as it signifies that four (i, i+1, i+3, i+4) residues out of a five residue group (i – i+4) have passed cutoffs for i, i+3 Cα distance and i, i+1, i+2, i+3 Cα torsion angle. Loose-helix SSE templates (SSE template henceforth referred to as SSET) consisting of at least five residues are formed from every set of consecutive loose-helix residues and the three residues immediately succeeding them. This process ensures that a helical element is generated from only a single continuous region of loose-helix residues. At this stage, it is possible that the third residue of a five residue loose-helix SSET has not passed the cutoffs for distance and torsion angle with any other residue. Strict-helix seed-SSE is defined by a group of three consecutive strict-helix residues. Strict-helix SSETs are formed from a strict-helix seed-SSE and the three residues immediately after it. This implies that every residue in a strict-helix SSET passes the strict cutoffs of distance and torsion angle with at least one other residue making the minimum length of a strict-helix SSET six residues. All loose-helix SSETs that do not contain at least one strict-helix residue, other than in the last three residues, are discarded. The remaining loose-helix SSETs denote possible helix templates.
A loose-strand-forming seed-SSE is defined by a group of loose-strand residues, with at least one residue in the group. Extending every loose-strand seed-SSE to the i+3 position at the C- terminal end gives rise to loose-strand SSETs. Overlaps between elements are formed during extension of the C- terminal end of the seed-SSE to the i+3 residue leading to a maximum overlap of two residues between any two SSETs.
Step 3: Quadruplets as seeding units of paired β-strands
In this step, paired β-strands are generated to identify and represent sheet-forming β-strand ladders (including β-bends and β-bridges) which are paired stretches of consecutive residues. We start by defining and identifying the smallest unit of such a network of residues, namely a quadruplet, which is formed by four Cα residues, linked by a pair of covalent bonds and a pair of hydrogen bonds (fig. 5a).
Figure 5 The three parameters on which quadruplets scoring is based. In figs. a, d, g, a quadruplet is formed from residues i, i+1, j-1 and j. The score is used to select the best quadruplets to join and form β-sheets. Each scoring parameter has been chosen such that they least influence each other. Residues i-1, i+2, j+1 and j-2 are required to calculate angles for quadruplet scoring. The first parameter is Cα-Cα distance between paired residues (fig. a). Blue lines joining i, j and i+1, j-1 show the distance being scored. This parameter approximates the deviation of the triangle apex i with reference to the triangle apex j in fig. b due to rotation of the plane i-1, i, i+1 on the X axis. Fig. c shows the Cα-Cα distances for parallel and antiparallel β-strands obtained from DSSP [8] output. Data is binned at 0.1 Å intervals and fit to a normal distribution using "gnuplot" [40]. Distribution for parallel β-strands has a mean at c1 (4.81 Å) with a sigma of 0.22. Distance for antiparallel β-strands follows a bi-modal distribution with means (μ) at c2 (4.46 Å) and c3(5.24 Å) and a standard deviation (σ) of 0.26. These μ and σ values were used to calculate the probability of occurrence of Cα-Cα pairing distances while scoring quadruplets by our algorithm. A Cα-Cα maximum distance of 7.5 Å (not shown) was used to limit pairing between residues. The second parameter is angle between lines (shown in blue) joining the vertices i, j and the base j-1, j+1 of the imaginary triangles j-1, j, j+1 and i+1, i, i-1 (fig. d). Only one of the four cases is shown. The other angles are between lines j, i and i+1, i-1; j-1, i+1 and i, i+2; i+1, j-1 and j, j-2. Deviation of this angle approximates the deviation of the triangle apex i-1 with reference to the triangle apex j+1 in fig. b due to rotation of the plane i-1, i, i+1 on the Y axis. Fig. e shows the distribution of angles, binned at 5° intervals, obtained from parallel and antiparallel β-strands defined by DSSP where c1 (87°) and c2 (82.2°) are the respective means. Fig. f shows the probability of obtaining a parameter-2 angle at different multipliers of the standard deviation for data shown in fig. e. The probability obtained is used for scoring quadruplets. The third parameter is a torsion angle (fig. g) between the points j, mj, mi, i. mj is the midpoint between j+i, j-1. mi is the midpoint between i-1, i+1. Lines joining residues and the midpoints are shown in blue. A similar torsion angle involving residues j-1, i+1 as end points and midpoints between j, j-2 and i, i+2 is computed (not shown). Deviation of the torsion angle approximates the deviation of vertex i in fig. b with respect to vertex j due to rotation of the plane i-1, i, i+1 on the Z axis. Fig. h shows the distribution of torsion angles (binned at 5° intervals) obtained from DSSP output where c1 (-20.9) and c2 (-27.9) are the respective means for data from parallel and antiparallel β-strands. Fig. i shows the probability of obtaining a torsion angle at different multipliers of the standard deviation for the data in fig. h.
Since the covalent bonds that link a quadruplet of residues are easy to define confidently from their sequence and coordinates, and their hydrogen bonds are not always clear, we decided to use parameters that depend on covalently linked rather than hydrogen bonded residues neighboring the quadruplet (fig. 5). At most, a single covalently linked atom is required from outside the quadruplet at each of the four corners for calculation of quadruplet-scoring parameters. DSSP [8] output, obtained from the "Statset" described above, was used to identify parallel and anti-parallel quadruplets and to score the individual parameters that were considered. Only residues marked "E" in the DSSP "structure" column and having pairs under either or both "BP1" and "BP2" columns were used for generation of statistics. The neighboring atoms, if required by the parameter, were also required to be marked "E" under the column for "structure". We describe the three parameters used for scoring the quadruplets.
The first parameter Cα-Cα distance between paired residues (fig. 5a), is equivalent to the deviation of vertex i of the imaginary triangle i-1, i, i+1 from the vertex j of the triangle j+1, j, j-1 (fig. 5b). Cα-Cα distances from binned DSSP [8] data follow a normal distribution (fig. 5c). The distances for parallel and anti-parallel β-strands were computed. The distance data were fitted to the following equations for parallel and anti-parallel quadruplets respectively.
where ρ is probability of x, α is a constant, μ is mean and σ is standard deviation. GNUPLOT [40] was used to fit the data and obtain values for mean (μ) and standard deviation (σ).
Probability of obtaining a particular distance, for scoring quadruplets, is calculated by using the mean and standard deviation obtained above. Two distances are calculated for each quadruplet. For parallel quadruplets, both distances are scored individually using the mean and standard deviation from equation 4. For anti-parallel quadruplets, the larger distance is scored using μ2 (c3 in fig. 5c) and the smaller distance using μ1 (c2 in fig. 5c), and standard deviation obtained from equation 5 above. Two scores are used for each quadruplet (fig. 5a: distance between residues i, j and i+1, j-1).
The second parameter is an angle to determine the deviation of a residue pair with respect to the neighboring one. The angle between the line joining potential hydrogen-bonded residues and the line joining the previous and the next residues of one of the pairing residues was measured, for each of the pairing residues (fig. 5d). Angles were measured for parallel and anti-parallel β-strands. Mean (μ; c1 and c2 for parallel and anti-parallel β-strands respectively in fig. 5e) and standard deviation (σ) were calculated from the raw data. Times-sigma deviation was calculated for each data point (fig. 5f) and the probability of obtaining a value at each point was calculated and used for scoring quadruplets. Four angles are calculated and scored for each quadruplet (fig. 5d: angle between lines i, j and j-1, j+1; j, i and i+1, i-1; j-1, i+1 and i, i+2; i+1, j-1 and j, j-2). From this parameter, only half the total score is used to score the quadruplet.
The third parameter is the torsion angle between pairing residues calculated as an angle between planes joining the pairing residues and the midpoints between the neighboring residues (Fig. 5g: residue j; midpoint mj between residues j+1, j-1; midpoint mi between residues i-1, i+1; and residue i for the first torsion angle. Residue j-1; midpoint between residues j, j-2; midpoint between residues i, i+2; and residue i+1 for the second torsion angle.). Angles were measured for parallel and anti-parallel β-strands (fig. 5h) and 5a probability vs. times-sigma deviation curve (fig. 5i) was prepared similar to that of parameter 2 and used for scoring quadruplets. Two torsion angles are calculated and scored for each quadruplet.
Quadruplet parameters are scored based on the probability of obtaining individual Z-scores. Deviation of each parameter data from its respective mean value, obtained from DSSP, is divided by its respective standard deviation, obtained from DSSP, to obtain a Z-score (times-sigma value). Probability of obtaining a particular Z-score is used for scoring quadruplets.
As the probability values are very small and could be subject to floating point errors over multiple operations, a negative logarithm was used to convert them to positive numbers. Thus, lower numbers represent better scores. The total score of the quadruplet is obtained by adding the individual parameter scores. Equal weights (2 scores for Cα distance, half of 4 scores for the second parameter angle and 2 scores for Cα torsion angle) are used for the three parameters. Technical limitations of the computer's ability to work with numbers close to zero were carefully avoided by rejecting probabilities close to zero (less than six decimal places). The table of times-sigma and negative logarithm of the probabilities were kept for lookup during scoring of actual quadruplets found by DSSP [8] during determination of quadruplet scoring cutoffs, and by our algorithm during β-strand definition.
Sets of true and false quadruplets were generated from DSSP [8] output (figs. 6a, 6b). A true quadruplet was created using two pairs of consecutive residues with a link in the "BP1" or "BP2" column of DSSP output. For every true quadruplet that was created, four false quadruplets were generated as shown in fig. 6 by considering alternative quadruplets that these residues could potentially form with neighboring residues. Each quadruplet was scored using all three parameters (fig. 5) as described above. Scores for true and false quadruplets were analyzed (figs. 6c, 6d).
Figure 6 True and false quadruplets generated from DSSP-defined β-strands. 6a, 6b: Residues i, j and i+1, j-1 shown paired with green broken lines form the true quadruplet. For every such quadruplet four false quadruplets (shown with red and blue broken lines) are possible: j, i+1, i+2, j-1 and j-1, i, i+1, j-2 in fig. a; i, j-1, j, i-1 and i, i+1, j, j+1 in fig. b. These quadruplets were scored to find the difference in scores between true and false quadruplets. True and false scores for quadruplets generated from DSSP output for parallel β-strand coordinates (fig. c) and antiparallel β-strand coordinates (fig. d). Cutoff c1 (45) and c2 (46) in fig. c are the scores of the best false quadruplet and the worst correct quadruplet respectively for parallel β-strand data. Cutoff c1 (39) and c2 (44) are the scores of the best false quadruplet and worst correct quadruplet respectively from antiparallel β-strand data. Cutoffs c1 and c2 are used as cutoffs to differentiate between grade 1 and grade 2 quadruplets in our algorithm.
Our algorithm uses quadruplets formed from all available residues with the restriction that both covalently linked residues do not exist in any strict-helix SSET and allows a maximum overlap of one residue with a strict-helix SSET. Pairing residues are limited by Cα-Cα distance (7.5 Å; fig. 5c), before scoring the other two parameters to improve computation speed. Since quadruplet-based β-sheet definition is one of the most computationally intensive steps of the algorithm, we critically examined distance scores to reject quadruplets that will not be part of a β-sheet. The quadruplets are scored similar to quadruplets obtained from DSSP [8], as described above. Based on the final score and cutoffs determined from the scores of true and false quadruplets generated from DSSP, quadruplets are placed in one of three groups. The grade 1 quadruplets are those with scores better than the best false quadruplets located from DSSP in the previous step (quadruplets scoring less than c1 in fig. 6c and 6d). The grade 2 quadruplets are those scoring in between the best false DSSP quadruplets and worst true DSSP quadruplets (quadruplets scoring between c1 and c2 in figs. 6c and 6d). Lastly, the grade 3 quadruplets are those, which passed cutoffs of distance and have non-zero probabilities for each of the three parameters (fig. 5). These three grades of quadruplets are sorted based on their scores and used as seeding and extending units for all paired β-strands.
Step 4: Calculation of cutoffs to disallow neighboring quadruplets with wrong residue pairings
Use of relaxed criteria such as those described in the previous section to determine the quality of quadruplets can lead to errors in some cases. For example, it is possible to obtain a low scoring grade 1 or a grade 2 quadruplet in cases where an extended region comes close to a β-sheet, leading to quadruplets connecting the β-sheet to the extended region. Depending on the local structure, these quadruplets may score better than those of the β-sheet in the surrounding region. While it is possible that one or a couple of good hydrogen bonds might actually exist in such a situation between the β-sheet and the extended region, this region will now get incorrectly assigned to the sheet. In such a case, the secondary structure obtained in a residue-correct manner will not be beneficial for the purpose of approximating β-sheets as a set of interacting linear elements.
A rule based on multi-strand β-sheet geometry was implemented to determine correct quadruplet extension and pairing in later steps. True quadruplets generated in the previous step from DSSP [8] output files (figs. 6a, 6b) were studied with respect to the bending angle of neighboring quadruplets having a common pair of covalently bonded residues. Angles were calculated for each of the common pair of residues with its pairs from both quadruplets (figs. 7a, 7b). An angle of 70° was chosen (c1 in fig. 7b) as the cutoff to determine if neighboring quadruplets are part of the same β-sheet. Both residues of the common covalent quadruplet edge have to pass the cutoff for both neighboring quadruplets to exist in a β-sheet. This ensures that no residue in any β-sheet has an angle >70° with its pairs and solves the problem of incorrectly assigning an isolated extended region that happens to be in the vicinity of a sheet as being a part of that sheet.
Figure 7 Initiation and extension of ladders of paired residues using quadruplets. 7a: Ladders of paired residues are initiated and extended using quadruplets. The initiation quadruplets i+2, i+3, j+5, j+6 and i+6, i+7, j+1, j+2 are shown with green pairing. Quadruplets are attached on either side to extend the arms of the ladder. Addition of quadruplet i+4, i+5, j+3, j+4 joins the two ladders i, i+4, j+4, j+8 and i+5, i+8, j, j+3 to form the complete unit. Depending on the position of the best quadruplets any number of quadruplets might be responsible for seeding a ladder. Smaller ladder fragments get joined by worse scoring quadruplets. 7b: Residue pairing angle between residues on three β-strands (residue i+1, j-1, k-1 in fig. c) from DSSP output. Cutoff c1 (70°) is close to the largest angle observed. This was used to check new residue pairings formed while adding quadruplets. 7c: Checks performed during quadruplet addition and ladder extension. Quadruplet k, k-1, j-1, j and i, i+i, j-1, j share the common residues j-1 and j. Pairing and angle between pairs are checked for residues j-1 and j when worse scoring quadruplets are added. Quadruplet k, k-1, j-1, j scores better than i, i+1, j-1, j. While adding i, i+1, j-1, j it was found that the angle i, j, k fails the cutoff of 70° (fig. b). Quadruplet i, i+1, j-1, j is not added. Insertion of bulge residues is handled during joining of quadruplets. Quadruplets i+1, i+2, j-2, j-1 and i+2, i+3, j-3, j-2 share the common residues i+2, j-2. The quadruplets are simply added end to end. However, quadruplets j, j-1, k-1, k and j-2, j-3, k-2, k-1 (pairing between j-2, k-1 not shown) share only a single residue k-1. As j-2, j-3, k-2, k-1 scores worse than j, j-1, k-1, k, the pairing between j-1, k-1 is retained and residue j-2 becomes a bulge with respect to residue k-1.
Step 5: Initiation of β-ladders (paired β-strands)
A quadruplet of residues can potentially be extended by joining neighboring quadruplets in the direction of hydrogen bonds and/or in the direction of covalent bonds, to form a β-sheet. However, extension in the direction of hydrogen bonds poses some problems in the case of bulges. Therefore, we decided to extend quadruplets only in the direction of the covalent bonds in order to obtain β-strand ladders (fig. 7c). The quadruplet extension process leads to the formation of ladders that are a sequence of paired residues. Thus, all residues on one arm of the ladder are paired with neighboring residues on the other arm of the ladder (fig. 7a), except for bulges (fig. 7b). Presence of a bulge is considered during ladder generation, and it is possible to have isolated residues on one arm of the ladder that do not have a pair. Such a bulge residue is incorporated into a ladder at this step.
Generation of paired β-strands is initiated using quadruplets from the sorted list of grade 1 quadruplets obtained in step 3. Use of only grade 1 quadruplets for seeding ensures that the core region of paired β-strands is less prone to errors. Each grade 1 quadruplet, starting with the best scoring one, is checked for its ability to either join a previously selected better scoring quadruplet or initiate a new ladder in this single pass method. In case of a conflict with a previously chosen quadruplet, the current worse scoring quadruplet is rejected. In every step irrespective of whether the current quadruplet is chosen or rejected, all quadruplets in the grade 1, 2 and 3 lists having scores worse than the current quadruplet and which can possibly conflict with the current quadruplet based on its constituent residues and pairing are rejected. This ensures that conflicting quadruplets do not initiate ladders on their own. Attempts are made to join quadruplets end to end in one of two ways. A check is first made if both edge residues (fig. 7a) of one quadruplet are the same as both edge residues of the neighboring quadruplet. This allows two consecutive quadruplets to be joined end to end to extend the ladder by a pair of residues. Upon failure to join a quadruplet by this method, the second method is applied. If only one corner of a quadruplet has the same residue as a corner of the other quadruplet, and the pairs to that corner residue in the different quadruplets are consecutive, the quadruplets are joined end to end to extend the ladder by one pair of residues. The pair for the common corner in the new quadruplet is designated as a bulge with respect to the common corner residue. This method of generating bulges also ensures that bulges arise from the worse scoring quadruplet. We have found such an assignment to be correct by manual inspection. In case no suitable existing quadruplet is located at either edge of any preformed ladder, the current quadruplet is designated as a ladder by itself and can be extended by quadruplets scoring lower than it.
Joining of quadruplets is attempted for both edges of a quadruplet. Thus, a quadruplet can potentially join with the edge quadruplets from two different ladders, with its own two edges. This joins the two existing ladders to form a single longer ladder of quadruplets. These ladders of paired residues are checked for errors before further processing. It is possible to obtain quadruplets with both residues on its diagonal as bulges. All such quadruplets are removed after breaking the links at their edges with neighboring quadruplets. Quadruplets are selectively removed even if a single common residue belonging to a pair of neighboring quadruplets fails to pass the cutoff angle for wrong residue pairing as described in step 4 (fig. 7a). Cases where residues have two and three pairing residues are treated differently. A single residue, pairing with more than three other residues was not observed, thus showing that quadruplet parameters are able to distinguish between gross errors in geometry. A residue having only two pairs indicates two neighboring quadruplets. If the residues involved fail to pass the cutoff scores for the angle, the worse scoring quadruplet is rejected. A residue pairing with three other residues indicates three neighboring quadruplets having at least one common residue. It is possible for a set of three residues at this location to pass cutoff scores for angle, in real structures, so we choose and eliminate the wrong quadruplet at this location. Two distinct cases are handled. A quadruplet is removed if it does not pass the angle cutoff with another neighboring quadruplet. If two quadruplets do not pass cutoff within themselves, but both pass cutoffs with a third quadruplet, we keep only one of the first two. This is chosen based on the length of the ladder of paired residues which the quadruplet takes part in. The quadruplet from the longer ladder is retained while the other one is rejected, consistently with the goal of obtaining longer elements. All isolated single quadruplets, which are formed by residues between two strict-helix SSETs, are removed, as they cannot extend a β-strand in any direction.
Step 6: Removal of short helices between consecutive β-strands
In our program, β-turns tend to show up as five residue helices. This has also been observed in other algorithms that define secondary structure on the basis of Cα coordinates [7,15,21]. Since we prefer fewer long elements instead of many short connected elements, wherever possible, β-turn regions are shown as part of the β-strand instead of short isolated helices. A step was added to remove all small helix SSETs at β-hairpin regions. We define a helix as being a minimum of five residues, as this could suitably represent a single turn of a helix. Moreover, for removal at this step, a helix is required to have at least five consecutive residues that are not part of any β-ladders formed in the previous step. We consider a maximum overlap of two residues between elements, so all helices of eight residues or longer always pass this criterion. Thus for this step, all helices up to seven residues long are considered for removal. Only helices inside β-hairpin bends as detected by Cα pairing between the β-strands are removed.
Step 7: Extension of β-ladders
Quadruplets with grade two scores are used to extend ladders formed from at least two grade 1 quadruplets. All ladders therefore have a core region of grade 1 quadruplets, which are then extended to the regions that are more distorted. This approximates stretches of paired residues with definite hydrogen bonding extended by residues where strict hydrogen bonds may be absent, and prevents defining entirely incorrect β-strands. Extension is very similar to the initiation step using grade 1 quadruplets, except that more checks are put in place when each quadruplet is added, to prevent unreasonably distorted regions from being joined. Our algorithm attempts to approximate elements as linear vectors at every step of the process. We incorporated steps to geometrically define the end of a ladder of paired residues to prevent generating ladders that span more than a single linear element. Quadruplets are not added if these criteria are violated. As grade 2 quadruplets are formed using less stringent criteria than the grade 1 quadruplets, a conservative approach in their use prevents delineation of highly distorted regions as β-strands. Extension is attempted on a list of ladders sorted by length. This ensures that longer elements have a better chance of ending due to geometric criteria rather than due to an extension quadruplet already being allocated to a shorter ladder. A pair of residues are added to the ladder only if neither of the new residues already have two other residue pairs each. Thus, ladder extension depends both on ladder length as well as on quadruplet grade. Extension of a ladder is also terminated if this causes residue overlap with residues already participating in a strict-helix SSET. Only residues that are a part of a loose-strand SSET are used in the extension step. Extension is also terminated upon reaching a β-hairpin bend. No residues are added if the pair about to be added does not fall within the distance cutoffs of 3.5 Å to 7.5 Å (fig. 5a). The ladder is terminated, if any of the residues about to be added have an i-3, i Cα-Cα distance less than 8.1 Å (eq 1, fig. 4c) with a residue already in the same ladder. This ensures that we do not extend a ladder over a β-bend.
Grade 3 quadruplets are used next, to extend only single grade 1 quadruplets that have not been extended by any of the above methods. As grade three quadruplets are the worst scoring quadruplets that are used, two important constraints are utilized to prevent spurious extensions of lone grade 1 quadruplets. Firstly, we decided to use a criterion of exactly three consecutively paired residues to denote paired β-strands that arose from single grade 1 quadruplets. This ensures that only a single unreliable grade 3 quadruplet is used for extension of any particular β-ladder. Upon inspection of structures, it was observed in that keeping a minimum of two consecutively paired residues, namely a single grade 1 quadruplet, to delineate β-strands, was too relaxed. This constraint alone was able to eliminate most cases of chance interaction between pairs of residues. However, in highly distorted regions that are too twisted to represent linear elements, more than a pair of consecutive residues seem to form isolated ladders. We decided not to include them as well. So, secondly, only grade 1 quadruplets sharing common residues with an existing β-ladder, formed by previous initiation and extension steps, are extended by a single grade 3 quadruplet, such that it maintains residue pairing with the longer β-ladder. These two constraints together ensure that we retain short, three residue, β-strands at edges of existing β-sheets even if they are slightly distorted, but at no other region unless they are initiated by a pair of consecutive grade 1 quadruplets.
At this stage, due to the possibility of bulges at the overlapping edges of consecutive β-ladders, we might obtain consecutive ladders of residue pairs whose one arm is separated from the other by a single residue. These ladders are joined with a single bulge residue if consecutive pairing arms on the other side share a common residue and restrictions imposed during ladder extension are not violated. Our method of ladder generation also makes it possible to obtain overlapping ladders such that one arm of one ladder overlaps with the arm of a different ladder. Due to the complicated nature of β-sheets, it is also possible to not have residue pairing between neighboring β-strands in the middle of a β-sheet. Dealing with multiple fragments of overlapping ladder-arms prevents the ability to distinguish between true edges of β-strands. Thus, we join all ladder arms that are formed from consecutive stretches of residues or with common residues between them. These joined ladder arms are used to denote β-strands. The residue pairing information generated during ladder formation is retained to generate pairing information between the newly formed β-strands. Consistent with our minimum requirement of three pairing-residues for ladders (described above), β-strands are considered as part of the same β-sheet if at least three residues are paired.
Step 8: Generating and breaking α-helices to form linear elements
Helix SSETs representing α, π and 310-helices, defined in the above steps, are based on relaxed criteria to avoid missing out residues that could potentially be part of a helical element. Although a rudimentary form of element edge delineation is obtained by the use of Cα distance and torsion angles during delineation of probable helical elements in step 2, these methods are capable of detecting only drastic changes in the helix axis. Our relaxed criteria of helix definition designed to include π and 310 helices, also allows curved, kinked and bent helices [44] to be included as being a part of a single element. In order to approximate these elements by vectors suitable for motif search, we decided to split these helices into linear elements while considering overlap between them. Some helix SSETs are found to overlap with parts of ladders generated and extended in the previous step. As the residues that overlap are not part of strict helix SSETs we prefer a state of 'β-strand' to 'α-helix' for the residues. Helix SSETs having β-strand overlaps are checked and shortened so as not to have more than two residue overlap, and only at the edges, with any residue that is part of a β-strand. As it is possible for a residue to form only two hydrogen bonds, we simulate this constraint by rejecting from the edge of the helix any residues that pair with more than one other residue in the formation of β-strands in the above step. The helices are again checked for the presence of at least five residues before being accepted for processing by this step.
Manual judgment of the results at this point indicated that our program's delineation of helices were acceptable in terms of residue coverage. Presence of bent helices was noticed in the checkset (described previously) and we decided to split them into linear elements without loss of constituent residues. We used helices defined by our program for calculation of parameters for breaking helices. Since DSSP [8] does not distinguish between consecutive helices, it was not possible to derive parameter-based cutoffs from DSSP helix definition. Using data from helices defined by our program, enabled us to implement a system that could properly handle loosely defined helices and not to break them more frequently than needed. The results were visually inspected for errors.
The helix breaking method relies on an analysis of the RMSD of helix residues around the helix axis. Two different methods were used to generate the helix axis. Only one of the methods was finally adopted (fig. 8a). We explain both methods, as they are equally suitable for long (>7 residue) α-helices. However, the method chosen for our algorithm works for helices defined by us which encompass α, π and 310 helices as distinct or part of the same element and also short (<8 residue) α-helices. The first method calculates the principal moment of the helix residues and used the eigenvector corresponding to the largest eigenvalue as the helix axis. This method thus depends only on the spread of the residues in space and does not take into account the linear connectivity of the helix residues. Errors in helix axis assignment were observed for π-helices and short α-helices. Due to the spread of residues being more on the diametrical plane of these helices, the axis found using the eigenvector method lay closer to the plane of the diameter instead of being normal to it. This method gave good results for longer helices. However, as we considered short α, π and other opened up helices for breaking and final definition, this method was not used.
Figure 8 Helix endpoints redefined based on RMSD and angle between their axial vectors. 8a: Vectors representing a short opened up helix by two different methods. The red arrow shows the axis obtained by using the largest spread of the Cα atoms (vector corresponding to the largest eigenvalue). The green arrow shows the rotational axis obtained when the helix that is shifted by one residue, is aligned to the original helix. The first method is unsuitable for representing this helix and does not work for π and short helices. Our algorithm uses the rotational-fit method (described below, 8b) for all helices. RMSD of residues are calculated over this vector. Angles between vectors, calculated from residues of consecutive helices, are used to determine whether to break them so as to appropriately define the helices as linear elements. 8b: Helix RMSD data calculated using the rotational fit vector. Average RMSD of unbroken helices from our algorithm varies widely. The helices were broken multiple times and the angle of break was analyzed (data not shown). The mode of the angle of break (22°) for long (>15 residue) helices was used to determine the break point of consecutive helices. Helices that break at >22° were chosen for the dataset for calculation of RMSD and angle of break (fig. c). Average RMSD of broken helices is shown in this figure. A line was fitted using "gnuplot" [40] to approximate the RMSD of broken helices. A Z-score of 2.5 is used to limit breaking helices that deviate less than 2.5 times sigma around the approximated RMSD value for broken helices at a particular helix length. 8c: Angle of helix break calculated from dataset of helices used in fig. b. Data were collected from helices broken once, twice and thrice. The normalized data are shown. Helices that show an angle greater than c1 (20°) between broken parts are split. 8d: Helix split by our algorithm. All possibilities of broken pieces are assessed with respect to the RMSD of the pieces and angle of break. Helices i, i+5 and i+4, i+15 are finally chosen as correctly broken. Helices i, i+8 and i+15; and i, i+11 and i+10, i+15 are also possibilities that are analyzed but not chosen as the optimum break. Residues i+4, i+5 are shared by the two helix pieces (Cα shown as spheres).
Based on the observations above, a rotational fit method was used to determine helix axis [22,45] (fig. 8a). This involves shifting the helix by one residue and aligning it with the original helix. The rotation axis corresponds to the helix axis. This was found to precisely determine the helix axis and was not dependent upon the length or under-winding of the helix. The axis determined by this method and that determined by the eigenvector method correlate very well for helices greater than 8 residues (data not shown).
Our program was run on the "statset", described previously, and all helices were extracted for study. The RMSD of helix residues from the helix axis were calculated and analyzed (fig. 8b). All helices were processed using the helix-breaking method described below. Each helix was split in three ways such that case 1 gave rise to 2 pieces, case 2 gave rise to 3 pieces and case 3 gave rise to 4 pieces from the original unbroken helix. The mode of breaking angles was calculated for every helix length (data not shown). The highest mode (22°) obtained was used in our breaking method to obtain a set of all helices that broke at least once at an angle greater than the mode. Broken helices that arose from this set were used to calculate the RMSD of broken helices around the rotational fit axis (fig. 8a). The standard deviation was calculated for every helix length and the results showing dependency of the standard deviation on the length were fitted to a straight line (data not shown).
The average RMSD and average standard deviation calculated above were used to obtain broken helices (as described below) that were analyzed manually. Our algorithm does not break a helix showing a slight curvature in its structure, or containing a few distorted residues. For bent helices, we decided to use two cutoffs to determine breakpoints. Flexibility is represented by a Z-score representing the allowed deviation as multiples of the standard deviation of helix residues (as calculated above). A sharp bend in the helix axis is measured by the angle between the axes of two neighboring helices as calculated by the rotational fit method (described above). The broken helices observed were manually inspected to determine the optimum Z-score and the break angle (fig. 8c).
A breaking method for helices was developed in order to determine the correct cutoffs for Z-score and angle of break. The method considers every possible breakpoint in a helix and attempts to choose the optimal result. We define broken helices as single elements only if the piece with the highest RMSD is still acceptable (fig. 8d). Multiple breaks are considered with no overlaps and with overlaps of up to two residues at every position of the helix. The minimum helix length allowed is five residues. Broken helices are considered starting from the shortest helix length (5 residues), with maximum overlap (2 residues), to the longest possible helix (the unbroken helix). The helix breaking process analyses the RMSD and angle of break for every combination of helical fragments (instance) derived from the unbroken helix. Thus, every "instance" is the set of broken helical fragments produced during the iterative breaking process. Every possible case of broken helices that can be formed by an unbroken helix is considered and a single helix is broken into as many long fragments as is optimal for correct representation as linear elements. RMSD along the helix axis, computed by the rotational fit method described above, is used to determine overall suitability of broken helical pieces. For every unbroken helix, RMSD of every possible broken helical fragment is considered. For all "instances" of broken helices, the fragment with the highest RMSD is located. The "instance" where the highest RMSD is minimum is considered the optimum break. Thus, we aimed to minimize, over all possibilities in which a single helix may be broken, the maximum RMSD of the broken helices generated by a single "instance". A Z-score cutoff is used to determine if a helix or helix-fragment needs to be broken. We tried out different values of Z-scores in an attempt to find the optimum value. The results for every set of broken helices corresponding to different Z-scores were manually inspected to disallow breaking helices that looked like a single element. Consecutive fragments are treated as a single helix if the angle between their axes is less than the cutoff angle, even if the resulting helix has a Z-score that fails cutoffs. A Z-score cutoff of 2.5 and an angle cutoff of 20° (fig. 8c) were finally chosen. Results were judged manually for correctness. Linear helices, according to our algorithm, thus never fail the axial bending angle of 20° and are not broken. This is close to the 25° angle obtained by Kundrot and Richards [15]. Curved helices are broken only if they are long and the axial bending angle is above 20°. Bent and kinked helices are already broken in step 2 (by i, i+3 Cα distance and i, i+1, i+2, i+3 Cα torsion cutoffs; fig. 4c, 4d) as they have a sharp angle and never need to be broken by this step.
This step of the calculation is computationally intensive due to the large number of possibilities that are considered. To prevent the algorithm from taking abnormally long to complete in special cases, we avoid checking for breaks in single helices that are longer than 50 residues.
Step 9: Breaking β-strands to generate linear elements
Bent β-strands are split to obtain linear elements using both geometric criteria (Cα distance and angle) and by using neighbor pairings for constituent residues. As β-strands have a natural tendency to curve, we take care not to break short gently curved β-strands, nor break them at bulges. Sharp bends in β-strands, gently curved but long β-strands, which cannot be optimally represented by a single linear vector, and β-strands which do not have at least two residues shared by two different β-sheets are considered for breaks. β-strands are broken using geometric criteria after checking for the possibility of a bulge being located near the potential breakpoint. We prefer to retain large regions of connected β-strands rather than split them into isolated pieces, and therefore place more importance on residue pairings than on geometric criteria of individual β-strands. Restricting the minimum length of a β-strand to three residues (as described above) makes β-strand breaking a sensitive operation, as it is possible to lose small β-strands completely if breaks are located either within the small β-strands themselves or on the β-strand pairing with the small β-strand. Thus, we try to retain residues in short β-strands if they are linked to part of a larger β-sheet. However, it is more likely for short β-strands to arise due to chance proximity of extended regions. These short β-strands may hinder correct representation of the entire β-sheet. Although our program does not aim to perfectly describe β-sheets, we do try to detect and remove short β-strands that are either not well connected to a larger β-sheet or located close to breaks in neighboring β-strands of the β-sheet. The β-strand breaking methods are designed not to depend on the length of the β-strands being broken, however, they do depend on the order in which they are applied on the original β-strand and its paired neighbors. β-Strands are broken in four steps, where each step works on the complete set of β-strands obtained after applying the previous breaking method. Bulges are taken into consideration at every step.
β-hairpin regions of β-strands are broken first, and these breaks are permanent. The use of quadruplets to generate β-ladders and joining ladder-arms to obtain paired β-strands does not allow demarcation of β-hairpin regions perfectly. Also, use of β-hairpin residues to form part of β-ladders in previous steps is not consistent with our aim of using a quadruplet to represent pairs of hydrogen bonded residues, since the last pair of residues of a hairpin bend may be consecutive and thus covalently bonded. Further, residues of hairpin bends, even when separated by a single residue score low when included in quadruplets. We identify all β-hairpins using residue pairing. This enables β-hairpin identification even if the last pair of paired residues leading to the hairpin is separated by up to three residues. This method allows detection of a β-hairpin correctly if the β-hairpin residues are paired with neighboring β-strands but not between themselves. All such cases of β-hairpins are resolved by placing the residues in one of the two β-strands based on residue pairing of consecutive residues. In case of an odd number of residues being in the β-hairpin, the middle residue is left out of the β-strands on either side of the β-hairpin.
Geometric breaks of β-strands using i, i+3 Cα-Cα distance are considered (fig. 9a). These account for the presence of bulges and are not marked as permanent breakpoints. Depending on pairing of and breakpoints in neighboring β-strands, the breakpoint may be disregarded in later steps. If the i, i+3 Cα distance is less than the cutoff of 8.1 Å (c1 in fig. 4c) a break is inserted. Presence of a bulge, originating from β-ladder initiation and extension steps above, within any of the residues at i, i+1, i+2, i+3 positions, prevents a break. The residues at positions i+1 and i+2 are treated as overlapping residues and included in both β-strands. Although breaks generated by this method indicate structurally bent regions, they are discarded upon absence of other potential breakpoints in the vicinity, at later steps. Breaks are not initiated if they lead to less than three-residue long β-strands.
Figure 9 β-strands redefined to obtain linear elements using different methods. 9a: Strands broken based on i, i+3 Cα distance (fig. 4a). Distance between i-1, i+2 residues and i+13, i+16 residues fail the cutoff distance of 8.1 Å. The residues i, i+1 (shown in red) are shared by β-strands i-9, i+1 and i, 1+7. Residues i+15, i+14 (shown in red) are shared by the β-strands i+24, i+14 and i+15, i+8. 9b: Angles for β-strand breaking while accounting for bulges. Angles were calculated from all β-strands defined by our algorithm before the β-strand-breaking step. The angle between i-2, i, i+2 Cα atoms is used to determine if the β-strand is bent. An average pseudo-point (pp) was generated from the j, j+1, j+2, j+3 atoms and the angle between j-1, pp, j+4 was found. β-strands were broken when i-2, i, i+2 angle was greater than c1 (45°) and j-1, pp, j+4 angle was greater than c2 (70°). j = i-1 showed the best correlation between the two angles (data not shown). 9c: Strand breaking using pseudo-point to find distorted regions. Residues i-1, i, i+1, i+2 (shown in red) are used to generate an average point. Angle between i-2, the average point and i+3 locates a distorted region if the cutoff angle of 70° fails. The β-strand is broken if the i-2, i, i+2 angle also fails at the same location. The β-strand i-3, i+4 is split to generate two β-strands i-3, i+1 and i, i+4. 9d: Strand breaking using pairing information between neighboring β-strands. Residue i (shown in red) is paired to residue j on one side and to residue k on the other. Residue i+1 is paired to residue j-1 however residue i-1 is not paired to β-strand j. Also, residue i-1 is paired to residue k+1 but residue i+1 is not paired to β-strand k. Lack of a pair of common residues pairing between β-strand j and k splits the sheet, with residue i shared between both sheets.
Geometric breaks of β-strands are also considered based on i-1, i, i+2 Cα angle. The i ± 2 Cα angle was observed to be a reliable indicator of bent β-strands in non-bulge regions. This break, as in the previous case, is treated as a potential break point only. A permanent break is initiated only if no bulges are observed in the β-strand region. All β-strands defined by our program were extracted from PDB [20] chains in the "statset" and the i-2, i, i+2 angles were calculated (fig. 9b). An angle of 45° (c1 in fig. 9b) was used as cutoff to determine a breakpoint. This was found to be too restrictive in that regions with only a single distorted residue could get broken. A method was implemented to locate β-strand distortions in the surrounding region using a less restrictive criterion. Data for finding cutoffs for bulge detection was found by processing the "statset" with our program. All β-strands defined by our program were extracted. A pseudo point was created for each set of four consecutive residues belonging to the same β-strand using residues i-1, [i, i+1, i+2, i+3], i+4 (fig. 9c). The angle between the lines joining the pseudo-point to the residues before and after the four residues used for the pseudo-point was calculated for all residues of the β-strands (fig. 9b). An angle cutoff of 70° (c2 in fig. 9b) was chosen to allow gradual bends in the structure. A break is allowed only if the i-2, i, i+2 Cα angle and the pseudo-point angle both fail for a given residue i. Breaks are not initiated if they lead to β-strands less than three residues long.
The final method used for β-strand breaking considers residue pairing between β-strands. Using residue pairing for detecting the end positions of β-strands allows splitting a β-sheet correctly based on the whole structure and not on minor distortions in geometry. Breaks by this method are permanent and are used as deciding factors in splitting an entire β-sheet into smaller pieces. Although our program does not aim to produce a perfect definition of all β-sheets, with respect to their boundaries as appropriate for correct domain definition, we do verify the SSE pairings to include them as part of the correct β-sheet. Due to the way in which ladders of residue pairs (described previously) were generated and joined, it is possible for two ladder arms to be formed from consecutive residues or be with a maximum of only one residue common between them, with the residues on the other arms having no connectivity (fig. 9d). A minimum of two residues overlap is required for neighboring β-strands to be part of the same β-sheet. Thus, a break is initiated if residues of a β-strand pairing with one of its neighbors lose contact with it for more than a single residue, unless the other pairing β-strand continues to maintain residue pairing on the other side. Only one common residue is allowed at the break, thus allowing a single residue to pair with both the neighbors, which are placed in different β-sheets. A β-strand is not considered for breaks if there are at least two residues pairing with both neighboring β-strands. Flexibility in allowing lost contact for a single residue ensures that bulges are retained.
Step 10: Generation of helices from residues not in any element
In previous steps, residues not part of a strict-helix SSET were considered for generation of ladders of residue pairs and β-strands. Some of these residues do not finally participate in any β-strand formation. These residues, not part of α-helices or β-strands, can potentially contribute to formation of π and 310 helices, or may be distorted while still showing helical tendency. Overlap with previously defined α-helices and β-sheets were considered for this step.
All loose-helix SSETs having no residue in a previously defined α-helix or a β-strand are considered at this step. To loosen the criteria and to allow over and under-wound helices to be detected, any presence of a strict helix-forming residue, other than in the last three residues of the previously defined element, leads to rejection of the template from consideration, as these have already been considered for helices in above steps. The templates are checked for a maximum overlap of two residues with β-strand elements and shortened at the edges if required. The templates are also checked so that no residue overlapping with a β-strand at the edge, pairs with more than one β-strand residue. Finally, helix templates occurring at β-hairpins are removed. Any helix template that overlaps with β-strands on both edges and has less than five non-overlapping helix residues is rejected. All remaining loose-helix SSETs are included in our final helix definition.
Step 11: Assignment of β-strands to β-sheets
As described previously, we consider a minimum of two pairs of residues to determine linked β-strands which themselves are at least 3 residues long. Consecutive residues of β-hairpins are considered linked for this purpose, even though they do not actually form a hydrogen bond. Sheets are assigned for each group of β-strands that can be traversed by a consecutive pair of linked residues. Breaks in β-strands located in previous steps are taken into consideration for this step. Breaks caused by changes in pairing (described previously) are treated as permanent, and the β-strands on either side are kept in separate β-sheets. Breaks caused by geometric evaluation of the local region are treated as potential breaks. A geometric break is ignored unless it affects all paired residues on either side of it. Thus, geometric breaks are used only if multiple β-strands need to be broken to split the β-sheet. Keeping a rigid criterion for the use of geometric breaks for individual β-strands ensures flexibility for the entire β-sheet, as larger β-sheets tend to show a gradual bending. A more drastic bending shows up as a sequence of geometric abnormalities, thus allowing proper use of the potential geometric breaks detected previously.
Assignment of β-strands to β-sheets is for ease of motif searches only. Our program does not define β-sheets for the purpose of domain definition. As such, ambiguity regarding whether two β-sheets should be linked or separate might arise in cases where there is a gradual bending of the sheet or where two or more β-strands link them at the edges. It is also possible for two different β-sheets to be linked together if a few β-strands in each sheet are distorted and are in close proximity to each other. In the majority of cases, however, our program defines β-sheet boundaries correctly.
Robustness of secondary structure assignment towards coordinate errors
Robustness towards coordinate errors was estimated by checking consistency of definition using randomly shifted PDB [20] coordinates. Gaussian random numbers were used to shift coordinates of the 100 files in the culled PDB set. Every residue (all atoms together) was individually shifted in all the three axes. 100 randomly shifted coordinate files were generated for every PDB file, starting each time with the original coordinates. 10 sets of 10,000 files each were generated with average RMSD from 0.2Å to 2.0Å in steps of 0.2Å. Secondary structure definition produced by DSSP [8], P-SEA [17] and our algorithm for each of these 1,00,000 files was compared with the definition produced by the same program from the unaltered coordinates. Mean percentage of correct definition over all residues, assuming definition from the unaltered definition to be fully correct, and standard error was calculated for each set of 10,000 files for each program (fig. 2a). Mean percentage of total residues reported as helices and β-strands for each of these 10 sets were also noted (fig. 2b).
Authors' contributions
IM designed and implemented the algorithms, tested program performance, analyzed the results, and drafted the manuscript. SSK contributed to algorithm development and provided expert judgment of program output. NVG conceived of the study and participated in its design and coordination. All authors read and approved the final manuscript.
Acknowledgements
This project was supported by NIH grant GM67165 to NVG. We thank James O. Wrabl for reading the manuscript and providing valuable comments.
==== Refs
Pauling L Corey RB Configurations of polypeptide chains with favoured orientations around single bonds: two new pleated sheets Proc Natl Acad Sci U S A 1951 37 729 740 16578412
Donohue J Hydrogen bonded helical configurations of the polypeptide chain Proc Natl Acad Sci U S A 1953 39 470 478 16589292
Low BW Grenville-Wells HJ Generalized mathematical relationships for polypeptide chain helices. The coordinates of the pi helix Proc Natl Acad Sci U S A 1953 39 785 802 16589334
Venkatachalam CM Stereochemical criteria for polypeptides and proteins. V. Conformation of a system of three linked peptide units Biopolymers 1968 6 1425 1436 5685102 10.1002/bip.1968.360061006
Rose GD Gierasch LM Smith JA Turns in peptides and proteins Adv Protein Chem 1985 37 1 109 2865874
Richardson JS The anatomy and taxonomy of protein structure Adv Protein Chem 1981 34 167 339 7020376
Levitt M Greer J Automatic identification of secondary structure in globular proteins J Mol Biol 1977 114 181 239 909086 10.1016/0022-2836(77)90207-8
Kabsch W Sander C Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features Biopolymers 1983 22 2577 2637 6667333 10.1002/bip.360221211
Frishman D Argos P Knowledge-based protein secondary structure assignment Proteins 1995 23 566 579 8749853 10.1002/prot.340230412
King SM Johnson WC Assigning secondary structure from protein coordinate data Proteins 1999 35 313 320 10328266 10.1002/(SICI)1097-0134(19990515)35:3<313::AID-PROT5>3.0.CO;2-1
Sklenar H Etchebest C Lavery R Describing protein structure: a general algorithm yielding complete helicoidal parameters and a unique overall axis Proteins 1989 6 46 60 2608659 10.1002/prot.340060105
Colloc'h N Etchebest C Thoreau E Henrissat B Mornon JP Comparison of three algorithms for the assignment of secondary structure in proteins: the advantages of a consensus assignment Protein Eng 1993 6 377 382 8332595
Taylor WR Defining linear segments in protein structure J Mol Biol 2001 310 1135 1150 11502001 10.1006/jmbi.2001.4817
Oldfield TJ Hubbard RE Analysis of C alpha geometry in protein structures Proteins 1994 18 324 337 8208725 10.1002/prot.340180404
Richards FM Kundrot CE Identification of structural motifs from protein coordinate data: secondary structure and first-level supersecondary structure Proteins 1988 3 71 84 3399495 10.1002/prot.340030202
Dupuis F Sadoc JF Mornon JP Protein secondary structure assignment through Voronoi tessellation Proteins 2004 55 519 528 15103616 10.1002/prot.10566
Labesse G Colloc'h N Pothier J Mornon JP P-SEA: a new efficient assignment of secondary structure from C alpha trace of proteins Comput Appl Bio Sci 1997 13 291 295 9183534
Mizuguchi K Go N Comparison of spatial arrangements of secondary structural elements in proteins Protein Eng 1995 8 353 362 7567920
Hutchinson EG Thornton JM PROMOTIF--a program to identify and analyze structural motifs in proteins Protein Sci 1996 5 212 220 8745398
Berman HM Battistuz T Bhat TN Bluhm WF Bourne PE Burkhardt K Feng Z Gilliland GL Iype L Jain S Fagan P Marvin J Padilla D Ravichandran V Schneider B Thanki N Weissig H Westbrook JD Zardecki C The Protein Data Bank Acta Crystallogr D Biol Crystallogr 2002 58 899 907 12037327 10.1107/S0907444902003451
Ramakrishnan C Soman KV Identification of secondary structures in globular proteins--a new algorithm Int J Pept Protein Res 1982 20 218 237 7129755
Bansal M Kumar S Velavan R HELANAL: a program to characterize helix geometry in proteins J Biomol Struct Dyn 2000 17 811 819 10798526
Ramachandran GN Sasisekharan V Conformation of polypeptides and proteins Adv Protein Chem 1968 23 283 438 4882249
Laskowski RA Rullmannn JA MacArthur MW Kaptein R Thornton JM AQUA and PROCHECK-NMR: programs for checking the quality of protein structures solved by NMR J Biomol NMR 1996 8 477 486 9008363 10.1007/BF00228148
DePristo MA de Bakker PI Blundell TL Heterogeneity and inaccuracy in protein structures solved by x-ray crystallography Structure (Camb) 2004 12 831 838 15130475 10.1016/j.str.2004.02.031
Webber CLJ Giuliani A Zbilut JP Colosimo A Elucidating protein secondary structures using alpha-carbon recurrence quantifications Proteins 2001 44 292 303 11455602 10.1002/prot.1094
Carter P Andersen CA Rost B DSSPcont: Continuous secondary structure assignments for proteins Nucleic Acids Res 2003 31 3293 3295 12824310 10.1093/nar/gkg626
Srinivasan R Rose GD A physical basis for protein secondary structure Proc Natl Acad Sci U S A 1999 96 14258 14263 10588693 10.1073/pnas.96.25.14258
Ichikawa S Hatanaka H Yuuki T Iwamoto N Kojima S Nishiyama C Ogura K Okumura Y Inagaki F Solution structure of Der f 2, the major mite allergen for atopic diseases J Biol Chem 1998 273 356 360 9417088 10.1074/jbc.273.1.356
Carter AP Clemons WM Brodersen DE Morgan-Warren RJ Wimberly BT Ramakrishnan V Functional insights from the structure of the 30S ribosomal subunit and its interactions with antibiotics Nature 2000 407 340 348 11014183 10.1038/35030019
PALSSE
Wolf YI Dondoshansky I BLASTCLUST
Murzin AG Brenner SE Hubbard T Chothia C SCOP: a structural classification of proteins database for the investigation of sequences and structures J Mol Biol 1995 247 536 540 7723011 10.1006/jmbi.1995.0159
Lo Conte L Ailey B Hubbard TJ Brenner SE Murzin AG Chothia C SCOP: a structural classification of proteins database Nucleic Acids Res 2000 28 257 259 10592240 10.1093/nar/28.1.257
Wang G Dunbrack RLJ PISCES: a protein sequence culling server Bioinformatics 2003 19 1589 1591 12912846 10.1093/bioinformatics/btg224
Python Programming Language
Biopython
Hamelryck T Manderick B PDB file parser and structure class implemented in Python Bioinformatics 2003 19 2308 2310 14630660 10.1093/bioinformatics/btg299
Debian
gnuplot
DeLano WL The PyMOL Molecular Graphics System
POV-Ray
Berman HM Westbrook J Feng Z Gilliland G Bhat TN Weissig H Shindyalov IN Bourne PE The Protein Data Bank Nucleic Acids Res 2000 28 235 242 10592235 10.1093/nar/28.1.235
Kumar S Bansal M Geometrical and sequence characteristics of alpha-helices in globular proteins Biophys J 1998 75 1935 1944 9746534
Christopher JA Swanson R Baldwin TO Algorithms for finding the axis of a helix: fast rotational and parametric least-squares methods Comput Chem 1996 20 339 345 8673326 10.1016/0097-8485(95)00075-5
Kraulis PJ MOLSCRIPT: A Program to Produce Both Detailed and Schematic Plots of Protein Structures Journal of Applied Crystallography 1991 24 946 950 10.1107/S0021889891004399
|
16095538
|
PMC1190160
|
CC BY
|
2021-01-04 16:27:24
|
no
|
BMC Bioinformatics. 2005 Aug 11; 6:202
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-202
|
oa_comm
|
==== Front
BMC BiolBMC Biology1741-7007BioMed Central London 1741-7007-3-171605353310.1186/1741-7007-3-17Research ArticleEcholocation calls and communication calls are controlled differentially in the brainstem of the bat Phyllostomus discolor Fenzl Thomas [email protected] Gerd [email protected] Max Planck Institute of Psychiatry, Kraepelinstr. 2-10, Munich, 80804, Germany2 Department Biology II, Ludwig-Maximilians-Universitaet, Grosshaderner Str. 2, Planegg-Martinsried, 82152, Germany2005 1 8 2005 3 17 17 8 4 2005 1 8 2005 Copyright © 2005 Fenzl and Schuller; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Echolocating bats emit vocalizations that can be classified either as echolocation calls or communication calls. Neural control of both types of calls must govern the same pool of motoneurons responsible for vocalizations. Electrical microstimulation in the periaqueductal gray matter (PAG) elicits both communication and echolocation calls, whereas stimulation of the paralemniscal area (PLA) induces only echolocation calls. In both the PAG and the PLA, the current thresholds for triggering natural vocalizations do not habituate to stimuli and remain low even for long stimulation periods, indicating that these structures have relative direct access to the final common pathway for vocalization. This study intended to clarify whether echolocation calls and communication calls are controlled differentially below the level of the PAG via separate vocal pathways before converging on the motoneurons used in vocalization.
Results
Both structures were probed simultaneously in a single experimental approach. Two stimulation electrodes were chronically implanted within the PAG in order to elicit either echolocation or communication calls. Blockade of the ipsilateral PLA site with iontophoretically application of the glutamate antagonist kynurenic acid did not impede either echolocation or communication calls elicited from the PAG. However, blockade of the contralateral PLA suppresses PAG-elicited echolocation calls but not communication calls. In both cases the blockade was reversible.
Conclusion
The neural control of echolocation and communication calls seems to be differentially organized below the level of the PAG. The PLA is an essential functional unit for echolocation call control before the descending pathways share again the final common pathway for vocalization.
==== Body
Background
Bats use echolocation calls for orientation in space and hunting for prey. Communication calls serve the purpose of inter- and intraspecific social communication. Neural control of both types of calls has access at the level of the medulla to the final common pathway for vocalization. Motoneurons controlling the larynx, the vocal tract and the respiratory muscles are accessed via interneurons from the nucleus ambiguus/retroambiguus complex [1]. One major subcortical structure for the control of vocalizations above the level of the medulla is the periaqueductal gray matter (PAG). Here, communication calls can be elicited by electrical or pharmacological stimulation in many mammalian species like squirrel monkey [2], rhesus monkey [3], rat [4], guinea pig [5], gibbon [6] and bat [7,8]. While in most areas where vocalizations can be elicited, long persisting stimulation causes a fading of the vocal reaction, the PAG is one of the very few regions where natural communication calls can be elicited without changes in the emotional state of the animal and without habituation to the stimulus, even over extended stimulation periods [9]. PAG stimulation in bats also evokes echolocation calls [10,11] without motor concomitants (except ear and nose leaf movements), and without habituation to the stimulus [7]. The PAG in bats is thus involved in the control of communication calls as well as echolocation calls. The different types of calls are elicited at anatomically distinct PAG locations [7].
In monkeys, the PAG has been shown to be part of the vocalization controlling system [12]. Although the author found the majority of vocal- eliciting structures within the limbic system, only stimulation in the PAG yielded all types of communication calls used by the animal [12].
Further findings suggest a hierarchical organization of the vocalization controlling system including the PAG. Bilateral lesions of vocalization-eliciting areas in the forebrain and diencephalon do not affect vocalizations triggered in the PAG, whereas lesions made at brain levels below the PAG, e.g., in the dorsolateral pons and the ventrolateral medulla suppress PAG-triggered vocalizations [13]. The PAG would thus serve a gating function controlling the release of vocal patterns that themselves are constituted in networks below the PAG level [14].
In the bat (Rhinolophus rouxi), three areas below the midbrain level in addition to the PAG have been identified that yield echolocation calls (but not communication calls) on electrical stimulation without temporal or spectral distortions [18]: a) the deep layers of the superior colliculus, b) the deep mesencephalic nucleus in the reticular formation and c) an area medial to the rostral parts of the dorsal nucleus of the lateral lemniscus, the paralemniscal area (PLA). All three areas share most of the stimulus-response conditions found in the PAG, such as very short latencies of vocalizations on stimulation, no habituation to stimulus, no body movements accompanying the vocalizations and no distortion of the call pattern. However, only echolocation (and not communication) calls can be elicited from these areas. Moreover, echolocation calls from the PLA can also be elicited by microinjection of a glutamate agonist (kainic acid), indicating that it is the neurons in the PLA rather than fibers passing through this area that are responsible for the activation of vocal utterances [19].
The PLA displays similar properties in other bat species, e.g., Pteronotus p. parnellii [20] and Phyllostomus discolor [7,21]. The interest in the PLA as an important component for vocal control in echolocation was supported by the finding that neurons in the rostral portion of the PLA of R. rouxi were active both before and during vocal emissions and also showed simultaneous responses to auditory stimuli. The PLA could therefore serve as an audio-vocal interface between auditory processing and motor control of vocalization [22,23]. The physiological data on the PLA available from animals other then bats are very sparse and anatomically corresponding areas have been investigated under different functional aspects [23-28].
Vocal control in other, non-echolocating mammals is characterized by a rather direct connection from PAG to the ambiguus/retroambiguus complex [1,15-17] or to the reticular formation [13,14,29], which would not necessitate an interaction from other brain structures (e.g., PLA) below the PAG level. But these findings seem to be insufficient to explain findings from Schuller and coworkers and Fenzl on the involvement of the PLA in motor control of vocalizations. A parallel organization of the descending control system for vocalizations could be assumed at the level of the PAG and PLA, and it is of interest whether this organization is a general mammalian feature.
The present study investigates whether the PAG and PLA interact during the PAG-induced emission of communication and/or echolocation calls in P. discolor using reversible blockade of PLA function. We found that PAG-induced communication calls could not be blocked either through ipsilateral or contralateral blockade of the PLA. However, PAG-induced echolocation calls could be temporarily blocked only through contralateral blockade of the PLA. These differences display partly differentiated functional organizations of vocal controlling pathways for communication calls and echolocation calls below PAG level.
Results
Implants – Stability of chronical microstimulation
Echolocation calls and communication calls were reliably elicited by electrical stimulation at specific stimulation sites in the PAG of P. discolor using chronically implanted stimulation electrodes. The thresholds for eliciting vocalizations varied within relatively narrow limits from probe to probe and from one experimental day to the next. In Fig. 1A the temporal variation of thresholds are plotted for the four implants over the entire experimental duration of 22 (animal 1) and 33 days (animal 2), respectively. The thresholds remained relatively stable between the day 4 and day 13, whereas stronger changes occurred in the first days after implantation or in some cases later than 13 days. Overall deviation between the day of implantation and the last experimental day was -30%, +33%, +56% and +95% in experiments A, B, C and D, respectively. During experiments the stimulation current was adjusted at supra-threshold levels to provide a reliable one-to-one relationship between electrical stimuli and vocal responses. The median values of supra-threshold currents used with the chronical implants A, B, C and D were 20% (P25 = 20, P75 = 25), 14% (P25 = 11, P75 = 15), 13% (P25 = 8, P75 = 14) and 14% (P25 = 7, P75 = 17) above threshold, respectively (Fig. 1B).
Echolocation calls versus communication calls
In order to test whether the production of echolocation or communication calls could be suppressed by neuropharmacological blockade, the nonselective glutamate antagonist kynurenic acid (KA) was iontophoretically injected into the PLA. Injections were either ipsilateral or contralateral to the PAG stimulation site. Twelve blockade experiments were conducted with consistent results in animal one (implants A and B), and seven experiments were conducted in animal two (implants C and D).
Blocking the PLA with KA suppressed PAG-induced echolocation calls whereas PAG-induced communication calls were less affected or not affected at all. The degree of suppression of echolocation calls depended on whether the PLA blockade was applied ipsilaterally or contralaterally to the PAG stimulation sites as shown in Fig. 2. An 85 min application of KA to the left PLA, i.e., the PLA ipsilateral to the PAG electrode eliciting echolocation calls and contralateral to the electrode triggering communication calls, did not completely suppress either echolocation calls (Fig. 2A, gray) or communication calls (Fig. 2A, black). A reduction of response probability for echolocation calls could, however, be seen around 50 min after the onset of iontophoresis. The activation probability for communication calls varied largely around 50%, and showed only a relative persistent lowering to around 40% between 50 and 90 min after the onset of KA application (Fig. 2B, gray).
A much different pattern of vocal responses appeared if the block was applied to the PLA contralateral to the PAG electrode eliciting echolocation calls. The response probability for echolocation calls started to decrease about 10 min after the onset of KA application and showed great variability, until complete suppression at around minute 80 (Fig. 2B, black). The blockage persisted beyond the termination of the experiment after 280 min.
When KA was applied bilaterally to both PLA sites, the suppression of PAG-induced echolocation calls occurred with a much shorter delay (about 24 min). However, PAG-induced communication calls could not be completely suppressed, although the probability of eliciting communication calls showed generally greater variability in response to KA application (Fig. 2C). A slight effect of KA could be discerned around 30 min after onset of the application; however, the effect did not persist. The slight mean decrease in elicitability, i.e., not falling below about 50%, and the high variability of elicitation probability persisted throughout the entire duration of the experimental run. These effects lasted beyond the 240 min, after which the experimental sessions had to be terminated.
After each blocking experiment a recovery period was inserted. Subsequently, a control experiment with electrical microstimulation only was started 24 hrs after the beginning of the previous experiment. In all control runs the elicitability of both types of calls had fully recovered. Calls could be elicited at standard stimulation currents, and no lesioning of PLA structures by iontophoretic currents could be detected. Typically the next iontophoresis experiment started within 24 hrs after the control experiment.
Ipsilateral vs. contralateral blockade of echolocation calls
The finding that a PLA-block by KA primarily affects contralaterally PAG-induced echolocation calls was supported by the results from following experiments, shown in Fig. 3. First, KA was applied to the left PLA for 43 min. Echolocation calls elicited in the PAG contralateral to the iontophoresis site were completely depressed about 32 min after the onset of iontophoresis and partially recovered about 10 min after termination of KA application. In contrast, the ipsilateral PAG-triggered echolocation calls were not affected at all. Similarly, the right PLA was blocked by KA application for 15 min. Again, the contralateral PAG-induced echolocation calls were blocked consistently about 7 min after iontophoresis onset and started to recover about 80 min after cessation of KA application. Echolocation calls elicited in the ipsilateral PAG were temporarily impeded, but elicitability was far from being totally blocked.
Although the depression of PAG-induced echolocation calls persisted beyond the end of the experimental session (140 min; Fig. 3A), elicitability from both PAG electrodes was fully recovered in the control experiments 24 hrs later.
Influence of sedation on animal during experiments
To ensure a stable position of the animal during the iontophoresis experiments, Rompun® was chronically applied subcutaneously during the sessions. Sedative was adequate to achieve optimal stability of the animal. Both echolocation calls and communication calls were elicited consistently over a long stimulation period of up to 33 days without any change in spectral composition of the calls. Motor reactions associated with vocalizations, such as pinna and nose leaf movement with echolocation calls or mouth movements with communication calls, also did not change during the experiments.
To ensure that sedation had no influence on activation of electrically induced vocalizations, a pure stimulation experiment under identical depressant conditions as applied at the blockade experiments was carried out (Fig. 4). Neither the initial dose of 0.5 ml sedative (0.04% Rompun® in 0.9% NaCl) prior to the experiment nor the continuous application of 6 μl/min of sedative showed any influence on the ability to trigger vocalizations. The vocalizations triggered under the influence of the sedative in general showed no difference to vocalizations triggered without sedation or to vocalizations emitted spontaneously, either in spectral or temporal patterns. Also the percentage of vocal answers triggered by one pulse train (duration: 2s, 12 single stimuli, 12 vocal answers correspond to 100% vocal answer) did not decrease under the influence of the sedative, as compared to experiments without sedation (data not shown). A decline or even depression of vocal answers of the type seen in the blockade experiments could not be detected.
Histological verification of electrode positions
Stimulation sites in the PAG and in the PLA were identified by tissue lesions (Fig. 5). The lesions in both structures could easily be detected in anatomical sections.
Discussion
This study has demonstrated that the neural control of echolocation calls and communication calls must have access to at least partly different neural substrates for vocalization.
Sedation of animals, stability of chronic implants and iontophoretic efficiency
The stability of the chronically implanted electrodes for electrical micro-stimulation was very satisfactory. Both echolocation calls and communication calls were elicited consistently over a long stimulation period of up to 33 days without any change in spectral composition of the calls. The slight increase of stimulation thresholds for the individual electrodes (Fig. 1) could be attributed to an accumulation of glia cells and debris caused by the presence of implants. Motor reactions associated with vocalizations also did not change during the experiments.
It is noteworthy that the onset of a contralateral PLA blockade is extraordinarily variable between 3 to almost 80 min under comparable experimental conditions (Fig. 2 and Fig. 3). Slightly different positions of the iontophoresis probe at the PLA site may be responsible for this, as even deviations as small as 100 μm correspond to almost 15% of the mediolateral dimension (≈ 800 μm, [7]) of the PLA in P discolor. At marginal application sites, KA would have taken longer to influence the necessary number of neurons in the PLA than when injected the geometrically optimal PLA location. The onset time differences indicate that the suppressive effect of PLA inactivation depends on the of PLA neurons.
Differentially organized vocal substrates for the production of echolocation calls and communication calls
Vocalizations are complex motor patterns imbedded into differentiated behaviors of an animal. It is well established that the PAG plays an important role in vocal control of communication calls [14,30], e.g., communication calls can be triggered by stimulating the PAG in several mammalian [7,31,32] and non-mammalian species [33]. In addition, echolocation calls can be elicited in bats within restricted areas of the PAG [8] that are distinct from areas in which communication calls can be triggered [7].
Outside the PAG, echolocation calls can also be elicited in a variety of brainstem areas [18], among which the PLA shows the lowest thresholds and the shortest latencies for eliciting ultrasonic vocalization. However, no communication calls can be elicited from the PLA [7,18,20]. From these findings, the hypothesis was deducted that different types of vocalizations could be modulated via at least partially separate and/or parallel vocal pathways in the bat.
A vocal pathway from the PAG to the nucleus retroambiguus (NRA) for the production of communication calls has been neuroanatomically defined by several authors [1,16,32,34]. The NRA includes a group of premotor neurons which send direct projections to thoracic and upper lumbar motoneurons [17] involved in expiration, and to the nucleus ambiguus containing laryngeal and pharyngeal motoneurons [15,35]. According to current knowledge, the PLA has no direct interferences with components of this pathway. The direct descending vocal path therefore cannot account for the blockade of PAG-induced echolocation calls by inactivation of the PLA as demonstrated in this paper.
Inactivation of confined areas in the brainstem exerts suppressive action on vocalization also in other mammals besides bats, as demonstrated by Jürgens [36]. Here, KA injections into the ventral pons of squirrel monkeys blocked a specific type of PAG-triggered communication call with characteristic frequency modulations, whereas other call types remained unaffected.
Jürgens suggests that vocal patterns are generally controlled in different brainstem regions, and that vocalizations with frequency modulations seem to depend on an intact periolivary region [36]. Echolocation calls of P. discolor are typically frequency-modulated calls covering a range between 45 and 100 kHz with the 3rd – 5th harmonic [37]. Our findings support the assertion by Jürgens that different vocal patterns could be controlled or modulated through activity in different brainstem regions, at least pertaining to echolocation calls and particular types of communication calls.
The effect of PLA-induced blockades on PAG-triggered vocalizations strictly depends on the side of the application, since echolocation calls can only be blocked when KA is applied contralateral to the stimulation site in the PAG. This is in contrast to what Jürgens describes for squirrel monkeys [36] where PAG-elicited vocalizations were only affected by ipsilateral, and not contralateral KA injection into the ventrolateral pons.
This difference in functional laterality may be attributed to the different brain regions involved in both studies and their specific connectivity. Since the neuroanatomical connectivity of the PLA is far from being understood, there is no straightforward explanation of this heterolateral influence. Besides strong reciprocal connections between the PLA [18,20] of both anatomical sides, there are no major midline crossing projections known so far to our knowledge that could account for the contralateral influence. Until today no anatomical data are available to describe connections between the PAG and the PLA and from the PLA towards the region of the nucleus retroambiguus/ambiguous complex in P. discolor. Only very few data are available from other species. In P. p. parnellii, efferents to the PAG were found when wheat germ agglutinin conjugated to horseradish peroxidase (WGA-HRP) was injected into vocally active sites of the PLA [20]. Efferents to the nucleus ambiguus from the PLA were found by Metzner using WGA-HRP [38] and connections from the lateral tegmental area (LTR) to the PAG and contralateral LTR were found using WGA-HRP and fluorescent tracers [27,28]. However, these data are not sufficient to explain the effectiveness of contralateral blockades under anatomical aspects.
The enhancement of the suppression effect by bilateral KA application may have two reasons. First, reciprocal interconnection between the two PLAs has been shown anatomically [20]. Second, ipsilateral blockage of the PLA also leads to reduced probability for eliciting echolocation calls, although it never reaches the level of suppression seen in most contralateral cases. The reciprocal interaction between the two PLAs is predominantly inhibitory as shown in R. rouxi (Schuller, unpublished), but the functional significance of this interaction is unclear. The assumption of a bilateral, but strongly unbalanced, descending control of echolocation calls via the PLA seems to fit the data more closely than does the assumption of a strictly unilateral organization. The presence of a non-functional ipsilateral PLA in addition to the silent contralateral PLA would further and more effectively decrease the probability for eliciting echolocation calls, resulting in a shorter time for onset of supression.
Based on connectional evidence (i.e., direct projections from the lateral part of the caudal PAG to the nucleus retroambiguus in the cat), Holstege proposed that the vocal pattern generation takes place within a final common pathway for vocalization driven by input from the PAG [15-17]. However, these findings and those of Zhang [1] that the vocal control pathway consists only of a direct connection from the PAG to the nucleus retroambiguus in the medulla oblongata are not supported by the results from the monkey [36]. Likewise, while our findings do not rule out a direct connection from the PAG to the nucleus retroambiguus (certain types of communication calls), they also demonstrate that a differentiated control for vocalizations (echolocation calls, frequency modulated) via parallel or at least partly separated pathways for echolocation calls could exist. This evidence of a more complicated network for vocal control at the level below the PAG in the monkey, as well as in the bat, underlines the broader significance of this concept on a mammalian level. The bat vocal control system therefore cannot be considered to be specialized, but is a general mammalian vocalization system with distinct emphasized features.
Conclusion
Communication calls and echolocation calls can be elicited with electrical microstimulation through chronically implanted electrodes at different sites within the PAG. Reversible blockade of the vocally active PLA in the region in which only echolocation calls can be triggered totally blocks PAG-induced echolocation calls but not communication calls. Thus, the PAG-NA/NRA pathway for vocalization described in literature may not be the only pathway processing vocal activity. The PLA seems to be essential for the production of echolocation calls but not for particular types of communication calls elicited in the PAG. This suggests differential pathway organization for particular types of communication calls on one hand and echolocation calls on the other hand. Whether the differentiation of pathways applies to the two classes of echolocation calls and communication calls in general, or whether it is more directly dependent on specific call properties in the acoustic pattern domain, remains open to further experimentation.
Methods
Experimental design
Two male neotropical bats (P. discolor) originating from the departmental breeding colony were used for this study. During the experiments, the animals were kept under semi-natural conditions with a 12:12 hrs light cycle, at 70% relative humidity and 28°C.
Surgical preparation of the animals was done under 4% Isoflurane (CuraMED Pharma, Karlsruhe, Germany) anesthesia. After additional local application of 2% Xylocain® (Astia, Germany), skin and muscles overlying the skull were cut along the midline, retracted to the sides and stabilized with sponge material (Gelastypt®, Hoechst, Germany). Minor bleedings were stanched with the coagulant Orbat® (lege artis Pharma, Germany). The skull surface was freed from tissue remains and a small holding tube was attached with a light-curing dental compound (Microglass®, Kulzer, Germany).
Experiments were conducted in an anechoic chamber, thus reducing acoustical interferences from the environment and reflections of call signals. The animals were placed in a holder that prevented gross body movements and the head was immobilized by attaching the surgically affixed tube to a head holder that allowed accurate repositioning (≤ 10 μm) of the animal in the stereotaxic device throughout repeated experimental procedures.
The orientation of the skull and consequently the brain within the stereotaxic coordinates was determined by scanning the profile lines of the exposed skull in parasagittal and transverse directions at the first post-operative day. Details of the stereotaxic device, the procedures to determine the skull position, and the reconstruction of the stimulation and iontophoresis sites are described elsewhere [39]. This method typically yields accuracy better than150 μm in all three dimensions.
Localization of vocally active sites within the PAG and PLA by electrical microstimulation typically started on the third postoperative day. Thirty minutes before each experimental session, an initial dose of 0.4 ml sedative (0.04% Rompun®/0.9% NaCl) was injected subcutaneously. The sedated state was maintained by continuous subcutaneous infusion of 0.04% Rompun in physiological saline with a rate between 3.5 μl to 5.5 μl/min (syringe pump: TSE-Systems, Bad Homburg, Germany). Sessions were generally limited to a maximum of 5 hrs per day.
Stimulation electrodes and iontophoresis probes were inserted through small holes of a typical diameter of 200 μm. Penetrations were made at different rostrocaudal and mediolateral inclinations in order to reach different locations through the same hole. All coordinates of probe positions were referred to the reference point of the equipment also used to determine skull and therefore brain position, and thus could be mathematically transformed to coordinates of standard frontal sections, corresponding to a specially prepared standard working brain atlas for this bat species (T. Fenzl and A. Nixdorf, unpublished). For further verification, stimulation and iontophoresis sites within the brain were marked using electrical lesions (-4 μA DC for 5 min) through the electrodes implanted in the PAG and through the stimulation electrodes used to localize the vocally active sites in the PLA.
Electrical microstimulation
For acute microstimulation, Parylene-coated tungsten electrodes (type TM33A20, WPI Inc., Sarasota, USA) were used. Teflon-insulated silver wires (AGT0510, WPI, Sarasota, USA, diameter: 125 μm) were implanted for chronic PAG micro-stimulation. A sharpened tungsten wire inserted into the neck musculature served as an indifferent electrode.
Electrical stimuli consisted of 15 ms long bursts with 0.1 ms long negative rectangular current pulses at 1 kHz rate, and were presented at a repetition rate of 6 bursts per s (stimulatorS48 with stimulus isolation unit PSIU6, Grass, Quincy, USA). One stimuli train lasted 2s.
During experiments, the animals were continuously monitored by TV under infrared light illumination (camera: Teli, Tokyo, Japan; monitor: TC-800 E4D, Osaka, Japan)) (LED array: 12 V/28 LED; Conrad Electronics, Germany).
Chronic stimulation electrodes
Sharpened Teflon®-insulated silver wires(type AGT0510, WPI, Sarasota, USA) with a wire diameter of 125 μm and 100–200 μm bare length were secured at the skull with light curing dental compound cement at PAG positions at which appropriate calls (communication and echolocation calls) could be electrically elicited. A pair of IC-socket pins was used as connectors. The electrical stimulus could be switched between electrodes using a remote controlled electric relay, interposed between the isolation unit and the stimulation electrodes.
A graphical overview of the experimental approach is provided in the additional file (Additional file 1).
Iontophoresis within PLA
For reversible blockade of the PLA, the glutamate antagonist kynurenic acid (KA; 75 mM at pH 9; Sigma-Aldrich, Steinheim, Germany) was iontophoretically applied through borosilicate glass microelectrodes with a tip diameter between 3 μm and 5 μm. Retaining current was 30 nA and ejection currents ranged between 200 nA and 250 nA, delivered by a Neurophore BH-2 system (Medical Systems Corp., Greenvale, USA).
PAG-triggered vocalizations were considered to have been blocked by iontophoresis in the PLA when 75% of electrical stimulations failed to elicit vocalizations during a stimulation period of 10 min (coherent).
Sound recording, processing and data analysis
Vocalizations were picked up with an ultrasonic microphone (type 4135, Bruel & Kjaer, Darmstadt, Germany), amplified, digitally converted at 250 kHz sample rate (CIO-DAS16/M1, Computer Boards Inc., Mansfield, USA), and stored on a personal computer. The recording program was written in Agilent-VEE (version 6 pro, Agilent, USA). Spectral analysis of the recorded vocalizations for call identification and evaluation of frequency shifts was performed with the software "Bat Sound" (Petterson Electronic AB, Sweden). Peak amplitude of vocalizations was derived from the power spectrum of vocalizations. Call length was evaluated using the envelopes of the vocalizations.
Neuroanatomical processing
The animals were euthanized at termination of the experiments with Nembutal® (16 mg/100 g BW) and transcardially perfused with 4% paraformaldehyd in 0.05 M phosphate buffer solution (PBS). The brains were soaked in 30% sucrose in 0.05 MPBS for cryoprotection. For embedding with egg yolk, the brains were aligned in a small Perspex chamber so that the section plane would best correspond to that of the reference brain sections of the brain atlas [39]. The brains were cut on a cryostat (Frigocut type 2700, Reichert-Jung, Germany) into 42 μm slices and generally three adjacent series were processed. Routinely, Nissl and fiber stains [40] were applied for identification of stimulation sites. The standardized cutting procedure made it possible to refer anatomical data from individual brains to the sections of the reference brain, permitting a comparison of data from individual animals.
Animal care
Principles of laboratory animal care were followed and experiments were conducted under the regulations of the current version of German Law and Animal Protection. Reference Government of Bavaria (Az.Reg.v.Obb.211-2531-37/98).
Authors' contributions
TF developed chronic implants and designed and carried out all experiments. GS participated in the design of the study and the manuscript. Both authors read and approved the final manuscript.
Additional files
File name: additional_file_1
File format: PDF
Title of data: Graphical illustration of the experimental approach
Description of data:
The additional data (see Additional file 1) provide a graphical overview of the experimental approach by explaining the position of the implanted electrodes within the PAG together with the ipsi- and contralateral alignments of the PLA electrodes. Additionally the removable connectors used on the animals are illustrated.
Supplementary Material
Additional File 1
Graphical illustration of the experimental approach.
Click here for file
Acknowledgements
The authors would like to thank Uwe Firzlaff and Andreas Nixdorf for fruitful discussions during the experiments, Herrmann Schweizer for his kind help on anatomical verifications of brain sections, Claudia Schulte and Horst König for their technical support, Karl-Heinz Esser (University of Hannover, Germany) and Hans Erkert (University of Tübingen, Germany) for helping to establish a breeding colony of P. discolor by supplying bats and Doug Truskowsky for proofreading the manuscript. Funded by Deutsche Forschungsgemeinschaft (DFG) ref. Schu390/5-3 and Schu390/7-1.
Figures and Tables
Figure 1 Stability of implants. Stability of stimulation threshold currents to elicit vocalizations and time course at chronic microstimulation probes. Electrodes were implanted at day0 of the graph A. (A) Thresholds of four chronically implanted stimulation probes (A/B and C/D) are shown. Stimulation through implant A and B (animal 1) and through implant C (animal 2) elicited echolocation calls, while stimulation through implant D (animal 2) triggered communication calls. (B) "Mean percentage above threshold"-values for all four implants (A to D) as plotted in A during the actual blockade experiments. The mean stimulation current was 20% (electrode A), 14% (electrode B), 13% (electrode C) and 14% (electrode D) above threshold to ensure a vocal answer on each stimulus within the pulse train. At threshold level, not all stimuli triggered a vocalization. The median values are indicated as numbers; additionally P25 and P75 values are shown in italics.
Figure 2 PLA-located blockade of PAG-induced echolocation calls and communication calls. The percentage of successful stimulations (y-axis) for eliciting vocalizations in the PAG is represented during and after KA induced blockade of PLA. Echolocation calls are graphed with diamonds, while squares are used for communication calls. The kynurenic acid application is indicated by black horizontal bars below the x-axis. Onset and termination times of iontophoresis are given in italics. Onset and termination times of iontophoresis are indicated by numbers at both ends of the bars in italics. Stimulation success (%) was calculated for intervals of 2 s. Black arrows on the abscissa indicate the 25% blockade boundary. (A) Application of KA to the left PLA. The ipsilateral production of PAG-induced echolocation calls (left electrode) cannot be blocked, although a slight depression can be noticed in the first half of the graph. The contralateral production of PAG-induced communication calls (right electrode) is only rarely influenced by the glutamate antagonist KA. (B) A blockade of the right PLA totally blocks the production of contralaterally PAG-induced echolocation calls (left electrode), while the ipsilateral production of communication calls (right electrode) again is barely influenced. Note that the curve for PAG ipsilateral (communication call) starts around 50%. (C) Bilateral blockade of both PLA sites again leads to a total depression of PAG-induced echolocation calls, while PAG-induced communication calls can be elicited across the entire experimental run. Prior to each experiment, vocal answers were stable for at least 10 minutes at a value comparable to the data shown at the start of the application of the antagonist.
Figure 3 PLA-located blockade of PAG-induced echolocation calls. At both electrodes, echolocation calls were triggered. Refer to Fig. 2 for explanations on graph. (A) Application of kynurenic acid to the left PLA does not influence the elicitability of ipsilaterally PAG-induced echolocation calls (left electrode), whereas it lowers dramatically the efficiency of contralaterally induced echolocation calls (right electrode). The elicitability does not recover to the 75% mark within termination of the experiment at minute 120. (B) A blockade of the right PLA totally blocks the production of contralaterally PAG-induced echolocation calls (left electrode), while ipsilateral-induced echolocation calls (right electrode) are little affected (although some drop in elicitability can be detected). Prior to each experiment, vocal answers had to be stable for at least 10 min at a value comparable to the data shown at the application of the antagonist.
Figure 4 Sedation has no influence on electrical stimulation. Influence of sedation on vocal performances. Prior to electrical stimulation, an initial dose of 0.5 ml sedative (0.04% Rompun® in 0.9% NaCl) was injected subcutaneously. Starting 30 min later at minute 0 of the plot, a continuous dose of 6 μl/min of 0.04% Rompun® was infused subcutaneously for a total of 205 min. The elicitability of electrically triggered PAG vocalizations is shown for a time period of 205 min. Compared to electrically triggered vocalizations before the sedative was injected (data not shown), no difference can be detected. Therefore, the sedative has no influence on electrically elicited vocalizations. Probe 1 and 2 refers to two implanted electrodes.
Figure 5 Electrical lesions. Histological verification of electrode locations A/B show Nissl stained 42 μm frontal sections. (A) Two lesions (L1/L2) caused by repeated electrical stimulation through chronically implanted electrodes placed into vocally active sites within the PAG. AQ, aqueduct; B, boundary between PAG and surrounding tissue. (B) Electrically induced lesion (L) 400 μm below the location of the iontophoresis probes in the PLA. Due to the 400 μm offset of the lesion below the PLA the function of the PLA during further experiments was not influenced. CER, cerebellum; LL, lateral lemniscus; 4V, 4th ventricle.
Table 1 Onset of KA-induced total suppression and recovery of PAG-elicited calls as shown in Figs. 2 and 3.
Call_type (PAG)
KA_I/C (PLA)
T
supp
[min]
T
rec
[min]
Fig.
EC I - - 2A
CC C - - 2A
EC C 75 >280 2B
CC I - - 2B
EC I/C 24 >240 2C
CC I/C - - 2C
EC I - - 3A
EC C 33 >140 3A
EC I - - 3B
EC C 8 110 3B
Call_type is either echolocation (EC) or communication (CC) call. KA application (KA_I/C) was either ipsi- (I) and/or contralateral (C) to the stimulation site in PAG. Tsupp, Trec are the times after the start of KA-application when the probability for eliciting calls was fully suppressed or fully recovered for a period of at least 10 min, respectively.
==== Refs
Zhang SP Bandler R Davis PJ Brain stem integration of vocalization: role of the nucleus retroambigualis J Neurophysiol 1995 74 2500 2512 8747209
Lu CL Jürgens U Effects of chemical stimulation in the periaqueductal gray on vocalization in the squirrel monkey Brain Res Bull 1993 32 143 151 8102315 10.1016/0361-9230(93)90068-M
Larson CR Kistler M Periaqueductal gray neuronal activity associated with laryngeal EMG and vocalization in the awake monkey Neurosci Lett 1984 46 261 266 6738919 10.1016/0304-3940(84)90109-5
Yajima Y Hayashi Y Yoshii N The midbrain central gray substance as a highly sensitive neural structure for the production of ultrasonic vocalization in the rat Brain Res 1980 198 446 452 7407609 10.1016/0006-8993(80)90759-3
Martin JR Motivated behaviors elicited from hypothalamus, midbrain, and pons of the guinea pig (Cavia porcellus) J Comp Physiol Psychol 1976 90 1011 1034 1033210
Apfelbach R Electrically elicited vocalizations in the gibbon Hylobates lar (Hylobatidae), and their behavioral significance Z Tierpsychol 1972 30 420 430 5048149
Fenzl T Schuller G Periaqueductal gray and the region of the paralemniscal area have different functions in the control of vocalization in the neotropical bat, Phyllostomus discolor Eur J Neurosci 2002 16 1974 1986 12453061 10.1046/j.1460-9568.2002.02261.x
Valentine DE Sinha SR Moss CF Orienting responses and vocalizations produced by microstimulation in the superior colliculus of the echolocating bat, Eptesicus fuscus J Comp Physiol A Neuroethol Sens Neural Behav Physiol 2002 188 89 108 11919691 10.1007/s00359-001-0275-5
Jürgens U The role of the periaqueductal grey in vocal behaviour Behav Brain Res 1994 62 107 117 7945960 10.1016/0166-4328(94)90017-5
Suga N Yajima Y Newman JD Auditory-vocal integration in the midbrain of the mustached bat: Periaqueductal Gray and Reticular Formation The physiological control of mammalian vocalization 1988 New York: Plenum Press 87 107
Suga N Schlegel P Shimozawa T Simmons J Orientation sounds evoked from echolocating bats by electrical stimulation of the brain J Acoust Soc Am 1973 54 793 797 4754391
Jürgens U Ploog D Cerebral representation of vocalization in the squirrel monkey Exp Brain Res 1970 10 532 554 4988409 10.1007/BF00234269
Jürgens U Pratt R The cingular vocalization pathway in the squirrel monkey Exp Brain Res 1979 34 499 510 105921 10.1007/BF00239145
Jürgens U Neural pathways underlying vocal control Neurosci Biobehav Rev 2002 26 235 258 11856561 10.1016/S0149-7634(01)00068-9
Holstege G Anatomical study of the final common pathway for vocalization in the cat J Comp Neurol 1989 284 242 252 2754035 10.1002/cne.902840208
Vanderhorst VG Terasawa E Ralston HJ IIIHolstege G Monosynaptic projections from the lateral periaqueductal gray to the nucleus retroambiguus in the rhesus monkey: implications for vocalization and reproductive behavior J Comp Neurol 2000 424 251 268 10906701 10.1002/1096-9861(20000821)424:2<251::AID-CNE5>3.0.CO;2-D
Vanderhorst VG Terasawa E Ralston HJ IIIHolstege G Monosynaptic projections from the nucleus retroambiguus to motoneurons supplying the abdominal wall, axial, hindlimb, and pelvic floor muscles in the female rhesus monkey J Comp Neurol 2000 424 233 250 10906700 10.1002/1096-9861(20000821)424:2<233::AID-CNE4>3.0.CO;2-C
Schuller G Radtke-Schuller S Neural control of vocalization in bats: mapping of brainstem areas with electrical microstimulation eliciting species-specific echolocation calls in the rufous horseshoe bat Exp Brain Res 1990 79 192 206 2311697 10.1007/BF00228889
Pillat J Schuller G Audiovocal behavior of Doppler-shift compensation in the horseshoe bat survives bilateral lesion of the paralemniscal tegmental area Exp Brain Res 1998 119 17 26 9521532 10.1007/s002210050315
Schuller G Fischer S Schweizer H Significance of the paralemniscal tegmental area for audio-motor control in the moustached bat, Pteronotus p. parnellii: the afferent and efferent connections of the paralemniscal area Eur J Neurosci 1997 9 342 355 9058054
Nixdorf A Die prätektale Region bei der Fledermaus Phyllostomus discolor – eine audio-motorische Schnittstelle? Dept Biol II, Ludwig-Maximilians-Universität München 2003
Metzner W Anatomical basis for audio-vocal integration in echolocating horseshoe bats J Comp Neurol 1996 368 252 269 8725305 10.1002/(SICI)1096-9861(19960429)368:2<252::AID-CNE6>3.0.CO;2-2
Metzner W A possible neuronal basis for Doppler shift compensation in echo-locating horseshoe bats Nature 1989 341 529 532 2797179 10.1038/341529a0
Covey E Response properties of single units in the dorsal nucleus of the lateral lemniscus and paralemniscal zone of an echolocating bat J Neurophysiol 1993 69 842 859 8463819
Henkel CK Edwards SB The superior colliculus control of pinna movements in the cat: possible anatomical connections J Comp Neurol 1978 182 763 776 730846 10.1002/cne.901820502
Henkel CK Afferent sources of a lateral midbrain tegmental zone associated with the pinnae in the cat as mapped by retrograde transport of horseradish peroxidase J Comp Neurol 1981 203 213 226 7309921 10.1002/cne.902030205
Gerlach I Thier P Koehler W Visual and vestibular signals in the lateral mesencephalic tegmental region of the cat Exp Brain Res 1991 85 641 649 1915715 10.1007/BF00231750
Gerlach I Thier P Brainstem afferents to the lateral mesencephalic tegmental region of the cat J Comp Neurol 1995 358 219 232 7560283 10.1002/cne.903580205
Thoms G Jürgens U Common input of the cranial motor nuclei involved in phonation in squirrel monkey Exp Neurol 1987 95 85 99 3792484 10.1016/0014-4886(87)90009-4
Jürgens U Neuronal control of mammalian vocalization, with special reference to the squirrel monkey Naturwissenschaften 1998 85 376 388 9762689 10.1007/s001140050519
Kirzinger A Jurgens U Vocalization-correlated single-unit activity in the brain stem of the squirrel monkey Exp Brain Res 1991 84 545 560 1864326 10.1007/BF00230967
Zhang SP Davis PJ Bandler R Carrive P Brain stem integration of vocalization: role of the midbrain periaqueductal gray J Neurophysiol 1994 72 1337 1356 7807216
Kennedy MC Vocalization elicited in a lizard by electrical stimulation of the midbrain Brain Res 1975 91 321 325 1164681 10.1016/0006-8993(75)90556-9
Ennis M Xu SJ Rizvi TA Discrete subregions of the rat midbrain periaqueductal gray project to nucleus ambiguus and the periambigual region Neuroscience 1997 80 829 845 9276498 10.1016/S0306-4522(97)00051-1
Vanderhorst VG Terasawa E Ralston HJ Monosynaptic projections from the nucleus retroambiguus region to laryngeal motoneurons in the rhesus monkey Neuroscience 2001 107 117 125 11744252 10.1016/S0306-4522(01)00343-8
Jürgens U Localization of a pontine vocalization-controlling area J Acoust Soc Am 2000 108 1393 1396 11051465 10.1121/1.1289204
Rother G Schmidt U Der Einfluss visueller Information auf die Echoortung bei Phyllostomus discolor Z Säugetierkunde 1982 47 324 334 16414822
Metzner W Anatomical basis for audio-vocal integration in echolocating horseshoe bats J Comp Neuro 1996 368 252 269 10.1002/(SICI)1096-9861(19960429)368:2<252::AID-CNE6>3.0.CO;2-2
Schuller G Radtke-Schuller S Betz M A stereotaxic method for small animals using experimentally determined reference profiles J Neurosci Methods 1986 18 339 350 3540473 10.1016/0165-0270(86)90022-1
Gallyas F Silver staining of myelin by means of physical development Neurol Res 1979 1 203 209 95356
|
16053533
|
PMC1190161
|
CC BY
|
2021-01-04 16:02:56
|
no
|
BMC Biol. 2005 Aug 1; 3:17
|
utf-8
|
BMC Biol
| 2,005 |
10.1186/1741-7007-3-17
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-841603595510.1186/1471-2407-5-84Research ArticleHIF1-alpha overexpression indicates a good prognosis in early stage squamous cell carcinomas of the oral floor Fillies Thomas [email protected] Richard [email protected] Diest Paul J [email protected] Burkhard [email protected] Ulrich [email protected] Horst [email protected] Department of Cranio-Maxillofacial Surgery, University of Münster, Waldeyerstrasse 30, 48129 Münster, Germany2 Department of Oral and Maxillofacial Surgery, Central German Armed Forces Hospital, Rübenacher Str. 170, 56072 Koblenz, Germany3 Department of Pathology, University Medical Center Utrecht, Utrecht, The Netherlands4 Institute of Clinical Chemistry and Laboratory Medicine, University of Muenster, Albert-Schweizer-Str. 29, 48129 Münster, Germany5 Institute of Pathology, University of Muenster, Domagkstraße 17, 48149 Münster, Germany2005 21 7 2005 5 84 84 22 3 2005 21 7 2005 Copyright © 2005 Fillies et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Hypoxia-inducible factor 1 (HIF-1) is a transcription factor, which plays a central role in biologic processes under hypoxic conditions, especially concerning tumour angiogenesis. HIF-1α is the relevant, oxygen-dependent subunit and its overexpression has been associated with a poor prognosis in a variety of malignant tumours. Therefore, HIF-1α expression in early stage oral carcinomas was evaluated in relation to established clinico-pathological features in order to determine its value as a prognostic marker.
Methods
85 patients with histologically proven surgically treated T1/2 squamous cell carcinoma (SCC) of the oral floor were eligible for the study. Tumor specimens were investigated by means of tissue micro arrays (TMAs) and immunohistochemistry for the expression of HIF-1. Correlations between clinical features and the expression of HIF-1 were evaluated by Kaplan-Meier curves, log-rank tests and multivariate Cox regression analysis.
Results
HIF-1α was frequently overexpressed in a probably non-hypoxia related fashion. The expression of HIF-1α was related with a significantly improved 5-year survival rate (p < 0.01) and a significantly increased disease free period (p = 0.01) independent from nodal status and tumour size. In primary node negative T1/T2 SCC of the oral floor, absence of HIF-1α expression specified a subgroup of high-risk patients (p < 0.05).
Conclusion
HIF-1α overexpression is an indicator of favourable prognosis in T1 and T2 SCC of the oral floor. Node negative patients lacking HIF-1α expression may therefore be considered for adjuvant radiotherapy.
==== Body
Background
Numerous molecular-biological and clinico-pathological studies have increased the knowledge about the hypoxia inducible regulation system in tumour biology. Protection against hypoxia in solid tumours is an important step in tumour development and progression. One system in hypoxia protection of tumour cells is represented by the hypoxia-inducible factor 1 (HIF-1) system which plays a crucial role in biologic processes under hypoxic conditions, especially in angiogenesis and carcinogenesis. HIF-1 is a heterodimeric protein consisting of an alpha-and beta subunit [1], in which the HIF-1α subunit mediates HIF-1 function as a transcription factor in response to cellular hypoxia. Being stabilized under decreased tissue oxygen concentration, it works as a cellular oxygen-sensing system, and induces the activation of key regulations systems through more than 40 proteins, including members of the glucose transporter (GLUT) and carbonic anhydrase (CA) family in the respective tumour cells [2]. Alteration and overexpression of HIF-1α has been detected in a variety of solid tumours, including breast, lung, ovarian and oral cancer with varying (diffuse and perinecrotic) staining patterns [3-7]. The expression of HIF-1α turned out to be of prognostic relevance in different tumours reviewed by Semenza [2]. The prognostic relevance of HIF-1α in tumours derived from squamous epithelium is however controversial.
Independently of improved operation-and adjuvant therapy strategies, the cure rate of oral squamous carcinoma has not changed significantly over the last decades. It still remains at approximately 50% over-all survival [8,9]. Better understanding of the biology of squamous cell carcinoma (SCC) of the oral cavity could identify new prognostic and predictive factors allowing tailoring of surgical and adjuvant therapy
In view of its important role in various cancers, we evaluated HIF-1α expression and related proteins such as GLUT 1 and CA IX in SCC of the oral cavity, which has not well been documented. The results were correlated with clinico-pathological features and prognosis. Our results demonstrate that non-hypoxia driven HIF-1α-expression is a frequent finding in SCC of the oral cavity and is associated with a favourable prognosis.
Methods
Patients and tumour material
Patients with histologically proven squamous cell carcinoma of the oral floor treated surgically were eligible for the study. Surgical treatment included radical tumour resection of the whole tumour with a free histopathological margin of at least 4 mm from the tumour borders. Selective neck dissection of Level I, II, III and V was performed in case of suspect results in preoperative tumour staging by computerized tomography or sonography examination or in case of tumour size over 2 cm. Bilateral selective neck dissection was performed when the tumour was extending over the midline (according to the recommendation of Robins et al. 2002)[10]. Radiotherapy was given when lymph node metastases were detected histologically. The series comprised 85 patients (71 men and 14 women) with a median age of 57 years (range 33–87). All tumours were classified postsurgically according to the TNM system (2002) [11]. Patients were clinically evaluated in our routine follow-up.
Immunohistochemistry
For tissue microarray (TMA) construction, two punch biopsies with a diameter of 0.6 mm were taken from the centre and the periphery (tumour stromal interface) of each tumour and transferred into the new acceptor block. TMA construction was performed by using a special tissue microarray instrument (Beecher Instruments, New Jersey, USA), according to standard protocols [12,13].
Immunohistochemistry for HIF-1 alpha, its downstream genes CAIX and Glut-1, and Ki67 was performed on 4-μm thick slides from the TMA. After deparaffinization and rehydration, endogenous peroxidase activity was blocked for 30 minutes in methanol containing 0.3% hydrogen peroxide. After antigen retrieval, a cooling-off period of 20 minutes preceded the incubation of the primary antibody (anti-HIF-1α; 1/500 dilution; BD Transduction Laboratories, Lexington, Kentucky, USA). Thereafter, the catalyzed signal amplification system (DAKO, Glostrup, Denmark) was used for HIF-1 alpha staining according to the manufacturer's instructions. The antibodies were detected by a standard avidin-biotin complex method with a biotinylated rabbit anti-mouse antibody (DAKO) and an avidin-biotin complex (DAKO), and developed with diaminobenzidine. Before the slides were mounted, all sections were counterstained for 45 seconds with hematoxylin and dehydrated in alcohol and xylene. Appropriate negative (obtained by omission of the primary antibody) and positive controls were used throughout.
For CA IX staining, sections were incubated without antigen retrieval with mouse anti-CA IX (MN 75; 1/50 dilution) for 30 minutes at 20°C and subsequently developed with an avidin-biotin-peroxidase complex method (Envision system peroxidase mouse; Dako). All sections were developed using diaminobenzidine, and subsequently counterstained with haematoxylin.
For GLUT-1 staining, sections were incubated without antigen retrieval with a rabbit polyclonal anti-GLUT-1 antibody (clone A 3536; Dako) and subsequently developed with a standard avidin-biotin-peroxidase complex method (biotinlyated goat antirabbit antibody; Dako; streptavidin peroxidase system; Dako) using an autostainer (Autostainer 480-2D; LabVision, Freemont, California, USA).
For Ki-67 staining (anti Ki-67, Mouse MAb, Immunotech SA, Marseille, France), sections were incubated with mouse anti-Ki-67 (MN 75; 1/40 dilution) over night at 4°C and subsequently developed with an avidin-biotin-peroxidase complex method (Envision system peroxidase mouse; Dako).
For Cyclin D1 staining (anti cyclin D1, Mouse MAb, Santa Cruz Biotechnology, Santa Cruz, USA), sections were incubated with mouse anti-cyclin D1 (1/400 dilution) for 20 minutes at 20°C and subsequently developed with an avidin-biotin-peroxidase complex method (Envision system peroxidase mouse; Dako).
The expression of HIF-1α was determined independently by two pathologists (PJ.v.D., H.B.). Both pathologists determined the percentage of positive cell nuclei in each core. The mean percentage value of the two cores representing one tumour were used for further evaluation. Three different cut offs for overexpression were set at 1%, 5% and 10% of positively stained nuclei.
Statistical analysis
TNM-stage, histological differentiation and expression of HIF-1α were related to the duration of the progression-free and the overall survival. The measurement of time started from the date of surgery to the date of histologically proven recurrent or metastatic carcinoma or disease related death, respectively. Patients who died from intercurrent diseases were censored at the date of death. Patients lost to follow-up were censored at the date of the last examination. Progression-free survival curves and the overall survival curves were constructed according to Kaplan and Meier [14]. The log-rank test was used to assess differences between groups and the multivariate survival analysis was performed with Cox regression [15].
Correlations between clinicopathologic features and expression of HIF-1α were evaluated by chi-square test. A p-value < 0.05 was considered to be significant.
Results
Clinical and tumour details of the 85 patients with oral floor squamous cell cancer are shown in Table 1.
HIF-1α expression was confined to the nuclei of neoplastic cells (according to Beasley et al. [4] and Bos et al. [16]). The expression levels did not differ much for the vast majority of the cases between the two tumour cores representing one case, indicating a rather homogeneous expression throughout the tumours. The difference between the evaluations of %HIF-1α expressing cells by both pathologists was within the range of 10%.
Detectable levels of HIF-1α (HIF-1α ≥ 1%) were found within the tumour cells in 63,5% (54/85) of the oral SCCs. Low levels of nuclear staining (HIF-1α ≥ 1% and < 5%) were found in 11,6 % (9/85) of the tumour specimen with detectable HIF-1α-expression, high levels of HIF 1-α expression (≥ 5%) were found in 52,9% (45/85) of the positive tumour specimens and very high levels (≥10%) were found in 35,9% (30/85) of the cases. HIF-1α expression was usually diffuse and tended to be more prominent towards the centre of tumour fields, sometimes associated with a better differentiation, indicated by the presence of keratinized tumour parts. In less than 5% of the cases necrotic areas were seen, and expression of HIF-1α could also be detected in perinecrotic vital tumour cells. HIF-1α-expression could be detected in the tumour stroma in only 5 cases. Cytoplasmatic expression was in 7 of 45 cases, all associated with nuclear staining levels of HIF-1α ≥ 5 %. 89.5 % (76/85) of the specimens showed membrane expression of Glut1 and 26% (12/85) of CA IX of neoplastic cells. No correlation could be demonstrated between HIF-1α, CA IX or Glut I, respectively.
No correlation was found between tumour size, tumour differentiation, lymph node status and expression of HIF-1α (≥1%, ≥5%, ≥10%). Also no correlation between the three expression levels of HIF-1α, cyclin D1 and Ki-67 could be shown in Chi-square test. Kaplan-Meier analysis of HIF-1α expression ≥1% and ≥10% showed no significantly association with disease free (p > 0.05) and overall survival (p > 05).
Kaplan-Meier analysis of carcinomas with low HIF 1-α expression (HIF-1α ≤ 5%) showed a significantly poorer disease free (p < 0.01) and overall survival (p < 0.01) (Figs. 1 and 2). No significant association between survival and the expression of CA IX or Glut-1 were found.
In the subgroup node negative T1/T2 tumours of the oral floor low HIF-1α expression was also correlated with a poor overall survival (p < 0.01) (Fig. 3). In multivariate Cox analysis we compared the overall survival and the disease free survival according to clinical usual prognostic factors tumour size and nodal status and also to cyclin D1 and Ki 67 with the HIF-1α expression. Nodal status, tumour size and HIF-1α expression were identified by Cox regression as independent predictors of overall survival (Table 2). Tumour size and HIF-1 expression were also predictors of tumour free survival in multivariate regression (Table 3).
Discussion
Postsurgical individual therapy decision making in SCC of the oral cavity is hampered by the lack of reliable prognostic markers. Whereas patients with histologically proven lymph node metastases are subjected to radiotherapy, this situation is much more complicated in primary lymph-node negative SCC. Postsurgical radiotherapy in these patients may often be over-treatment, as only a subgroup of poor prognosis lymph-node negative carcinomas would benefit from this type of adjuvant therapy. Nevertheless, at the present state of knowledge no reliable marker exists that identifies these patients.
In recent years it could be shown that the expression of HIF-1α might be a potential candidate for the prognostic assessment for a variety of malignancies. The HIF-1 system represents one of the central cellular anti-hypoxia systems. From a clinical point of view, high levels of HIF 1-α expression seem to predict a poor prognosis for various cancers [17-19].
In our study we aimed to evaluate the prognostic value of HIF 1-α in squamous cell carcinomas of the oral floor with focus on the clinically important subgroup of node negative cases. Up to date only a few data exist about the expression of HIF 1-α in correlation to prognosis in oral SCC, especially after surgical treatment without adjuvant radiotherapy. Unexpectedly, we were able to show that HIF-1α overexpression was related to a significantly poorer 5-year disease free and overall survival. In the clinically most relevant primary node negative T1/T2 tumours, loss of HIF-1α expression identified a subgroup of high-risk patients. A correlation with other established markers could not be defined. We are aware that the use of tissue microarrays might bias immunohistochemical results, especially in SCC's. However, it has to be stressed that the two punches representing one tumour were taken from the centre and the periphery (tumour stromal interface) as previously described by our group for other tumour entities [20]. We have not seen major, statistically relevant differences between the two respective punches. These results are in line with results published by Beasley et al. [4]. Using full sections, this group was able to demonstrate a rather homogeneous distribution of HIF-1 positive cells in SCC's. In detail, no major difference could be found between perinecrotic regions and areas at the tumour-stromal interface. We are therefore strongly convinced that our results are not biased by the use of two tumour punches representing the whole tumour. As we have pointed out ourselves before [21], the TMA technique may have limitations for proteins showing focal expression such as HIF-1α when it is expressed next to small necrotic areas [22]. Nevertheless, even HIF-1α immunohistochemistry on TMAs has provided relevant data in breast cancer [23]. However when there is more diffuse expression of HIF-1α as in oral SCC in the present study, the small tumour cores forming the TMA will more likely be representative.
The literature gives no uniform recommendation for a cut off point of HIF-1α. The used cut off value of HIF-1α expression varies in the literature. Beasley et al. [4] used a cut off value for HIF-1α ≥1% in oral cancer. Other authors used cut off values for HIF-1α between 1% and 5% (Bos R et al. [6] : HIF-1α ≥ 5%.; Bos R et al. [16] : HIF-1α ≥1%). With regard to the variety of the used levels, we choose three different cut off values for HIF-1α expression ≥1%, ≥5% and ≥10%. The 5% threshold of HIF-1α expression discriminated in our investigation two different populations with significant statistical differences in survival prognosis. In this view we have chosen 5% as the cut off value for the main statistical analysis in our study. In regard of the existing literature we also determined the cytoplasmic HIF-1α staining. However, in accordance to Beasley et al. [4] a cytoplasmic staining was a rather rare event and did not influence the statistical results.
In contrast to some adenocarcinomas (e.g. breast cancer [6]), our results indicate that HIF-1α overexpression is related to good prognosis.
The frequent diffuse type of HIF-1α overexpression is in contrast with most adenocarcinomas where usually perinecrotic, hypoxia induced HIF-1α expression is seen [6,24]. In breast cancer, diffuse type of HIF-1α overexpression is probably non-functional [25]. The diffuse pattern of HIF-1α staining is not hypoxia related but is due to alterations in oncogene or tumour suppressor genes (review Semenza [2]). The frequent diffuse HIF-1α expression in SCC of the oral cavity is probably also not hypoxia related. This hypothesis is substantiated by the lack of CAIX and Glut-1 expression, 2 important downstream targets of HIF-1α in the present study. This indicates that the expression of HIF-1α in oral SCC is rather oncogene/tumour suppressor gene related. In view of the observed expression of HIF-1α expression in the higher layers of normal squamous epithelium and the relation to keratinisation in SCC, HIF-1α may have a physiological role in differentiation in cases with diffuse expression. Thereby, HIF-1α expression may be an epiphenomenon rather than constitute a carcinogenetic event in these cases.
Only a small minority of SCC revealed necrotic tumour areas, frequently associated with HIF-1α expression and expression of CAIX and Glut-1, pointing to a hypoxia induced event. These cases tended to be of poor prognosis
Also in adenocarcinomas of the breast, diffuse HIF-1α expression was associated with a better survival than hypoxia induced perinecrotic HIF-1α expression [25].
Our results confirm those by Beasley et al. [4] who also reported a better clinical outcome of HIF-1α overexpressing head&neck SCC, but are at variance with another study on that reported the opposite [26]. Significant variations in the application of postsurgical radiotherapy may however at least in part explain the divergent results.
A small number of cases revealed HIF-1α-expression in fibroblastic cells. Recently, it could be shown that the stromal expression of CA IX indicated an improved patient prognosis in invasive breast cancer. The number oral SCC cases showing stromal HIF-1α expression was too small to evaluate its clinical significance. In view of the potential functional role of HIF-1α expression in the stroma of SCC this deserves to be further studied.
Conclusion
In summary, we were able to show frequent, probably non-hypoxia related expression of HIF-1-α in oral floor SCC that is related with improved prognosis. Lymph node negative patients lacking HIF-1α may be considered for adjuvant radiotherapy when these results can be confirmed in independent studies.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TF: Project planning, data analysis and writing of the manuscript
RW: Project planning and critical appraisal of the manuscript
PJD: Data analysis and critical appraisal of the manuscript
BB: Project planning, critical appraisal of the manuscript
UJ: critical appraisal of the manuscript
HB: Project planning, data analysis and writing of the manuscript
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank Mrs. Petra van der Groep for performing the immunohistochemical staining.
Figures and Tables
Figure 1 Overall survival of patient with oral squamous cell carcinoma in dependence on HIF-1 expression calculated by the Kaplan-Meier method.
Figure 2 Tumor-free survival of patient with oral squamous cell carcinoma in dependence on HIF-1 expression calculated by the Kaplan-Meier method.
Figure 3 Overall survival of nodal negative patient with oral squamous cell carcinoma in dependence on HIF-1 expression calculated by the Kaplan-Meier method.
Table 1 Tumour patient cases in this study
Age (mean) 57 years (range 33 – 87 years)
Sex 83.5 % men, 16,5 % women
Follow-up (mean) 51.8 months (range 4 – 154 months)
T stadium 50 T1 tumours and 35 T2 tumours
N stadium 60 lymph node negative-and 25 lymph node positive carcinoma without extracapsular spread
Grading 17 well differentiated tumours (G1)
59 moderately differentiated tumours (G2)
9 weakly differentiated tumours (G3)
Table 2 Multivariate analysis of survival in relation to the HIF-1α expression and other prognostic factors
Prognosticator P valuea RRb Confidence intervalc
Tumor size 0.005 3.4 1.4–7.8
Nodal status 0.003 2.3 1.3–4.2
HIF-1α 0.001 0.2 0.1–0.5
Cyclin D1 0.398 1.3 0.7–2.4
Ki-67 0.602 0.9 0.5–1.5
a The partially nonparametric regression model of Cox (1972) was used to evaluate the predictive power of vatious combinations of prognosticators in a multivariate manner.
b RR, relative risk
c 95% confidence interval
Table 3 Multivariate analysis of disease free survival in relation to the HIF-1α expression and other prognostic factors
Prognosticator P valuea RRb Confidence intervalc
Tumor size 0.11 2.1 0.8–5.1
Nodal status 0.93 1.0 0.5–2,3
HIF-1α 0.01 0.3 0.1–0.7
Cyclin D1 0.43 1.4 0.6–3,0
Ki-67 0.99 1.0 0.6–1.8
a The partially nonparametric regression model of Cox (1972) was used to evaluate the predictive power of vatious combinations of prognosticators in a multivariate manner.
b RR, relative risk
c 95% confidence interval
==== Refs
Wang GL Jiang BH Rue EA Semenza GL Hypoxia-inducible factor-1 is a basic-helix-loop-helix-PAS heterodimer regulated by cellular O2 tension Proc Natl Acad Sci USA 1995 92 5510 5514 7539918
Semenza GL Targeting HIF-1 for cancer therapy Nat Rev Cancer 2003 3 721 32 13130303 10.1038/nrc1187
Lee CH Lee MK Kang CD Kim YD Park do Y Kim JY Sol MY Suh KS Differential Expression of Hypoxia Inducible Factor-1 alpha and Tumor Cell Proliferation Between Squamous Cell Carcinomas and Adenocarcinomas Among Operable Non-Small Cell Lung Carcinomas J Korean Med Sci 2003 18 196 203 12692416
Beasley NJ Leek R Alam M Turley H Cox GJ Gatter K Millard P Fuggle S Harris AL Hypoxia-inducible factors HIF-1alpha and HIF-2alpha in head and neck cancer: relationship to tumor biology and treatment outcome in surgically resected patients Cancer Res 2002 62 2493 2497 11980639
Birner P Schindl M Obermair A Plank C Breitenecker G Oberhuber G Overexpression of hypoxia-inducible factor 1alpha is a marker for an unfavorable prognosis in early-stage invasive cervical cancer Cancer Res 2000 60 4693 4696 10987269
Bos R Van Der Groep P Greijer AE Shvarts A Meijer S Pinedo HM Semenza GL Van Diest PJ Van Der Wall E Levels of hypoxia-inducible factor-1alpha independently predict prognosis in patients with lymph node negative breast carcinoma Cancer 2003 97 1573 1581 12627523 10.1002/cncr.11246
Nakayama K Kanzaki A Hata K Katabuchi H Okamura H Miyazaki K Fukumoto M Takebayashi Y Hypoxia-inducible factor 1 alpha (HIF-1 alpha) gene expression in human ovarian carcinoma Cancer Lett 2002 176 215 223 11804750 10.1016/S0304-3835(01)00762-5
Casiglia J Woo SB A comprehensive review of oral cancer Gen Dent 2001 49 72 82 12004680
Hemprich A Breier Th Kovács G Long-term results of radical therapy of squamous cell cancers of the mouth with special reference to late recurrence and second malignancies – a catamnestic study of 20 years Fortschr Kiefer Gesichtschir 1992 37 43 5 German 1639309
Robbins KT Clayman G Levine PA Medina J Sessions R Shaha A Som P Wolf GT American Head and Neck Society; American Academy of Otolaryngology – Head and Neck Surgery Neck dissection classification update: revisions proposed by the American Head and Neck Society and the American Academy of Otolaryngology-Head and Neck Surgery Arch Otolaryngol Head Neck Surg 2002 128 751 8 12117328
Wittekind C Greene FL Hutter RVP Klimpfinger M Sobin LH TNM Atlas 2005 5 Springer Verlag Heidelberg
Kononen J Bubendorf L Kallioniemi A Barlund M Schraml P Leighton S Torhorst J Mihatsch MJ Sauter G Kallioniemi OP Tissue microarrays for high-throughput molecular profiling of tumour specimens Nat Med 1998 4 844 847 9662379 10.1038/nm0798-844
Packeisen J Buerger H Krech R Boecker W Tissue microarrays: a new approach for quality control in immunohistochemistry J Clin Pathol 2002 55 613 5 12147657 10.1136/jcp.55.8.613
Kaplan EL Meier P Nonparametric estimation from incomplete observations J Am Stat Ass 1958 53 457 481
Cox DR Regression models and life-tables JR Stat Soc [B] 1972 34 187 202
Bos R Zhong H Hanrahan CF Mommers EC Semenza GL Pinedo HM Abeloff MD Simons JW van Diest PJ van der Wall E Levels of hypoxia-inducible factor-1 alpha during breast carcinogenesis J Natl Cancer Inst 2001 93 309 14 11181778 10.1093/jnci/93.4.309
Takahashi R Tanaka S Hiyama T Ito M Kitadai Y Sumii M Haruma K Chayama K Hypoxia-inducible factor-1alpha expression and angiogenesis in gastrointestinal stromal tumor of the stomach Oncol Rep 2003 10 797 802 12792726
Kuwai T Kitadai Y Tanaka S Onogawa S Matsutani N Kaio E Ito M Chayam K Expression of hypoxia-inducible factor-1alpha is associated with tumor vascularization in human colorectal carcinoma Int J Cancer 2003 105 176 81 12673675 10.1002/ijc.11068
Giatromanolaki A Koukourakis MI Sivridis E Turley H Talks K Pezzella F Gatter KC Harris AL Relation of hypoxia inducible factor 1 alpha and 2 alpha in operable non-small cell lung cancer to angiogenic/molecular profile of tumours and survival Br J Cancer 2001 85 881 90 11556841 10.1054/bjoc.2001.2018
Brandt B Roetger A Bidart JM Packeisen J Schier K Mikesch JH Kemming D Boecker W Yu D Buerger H Early placenta insulin-like growth factor (pro-EPIL) is overexpressed and secreted by c-erbB-2-positive cells with high invasion potential Cancer Res 2002 62 1020 4 11861376
van Diest PJ Vleugel MM van der Groep P van der Wall E VEGF-D and HIF-1{alpha} in breast cancer J Clin Pathol 2005 58 335
van Diest PJ Vleugel MM van der Wall E Currie MJ Hanrahan V Gunningham SP Morrin HR Frampton C Han C Robinson BA Fox SB Expression of HIF-1{alpha} in human tumours J Clin Pathol 2005 58 335 6
Woelfle U Cloos J Sauter G Riethdorf L Janicke F van Diest P Brakenhoff R Pantel K Molecular signature associated with bone marrow micrometastasis in human breast cancer Cancer Res 2003 63 5679 84 14522883
Bos R Zhong H Hanrahan CF Mommers EC Semenza GL Pinedo HM Abeloff MD Simons JW van Diest PJ van der Wall E Levels of hypoxia-inducible factor-1 alpha during breast carcinogenesis J Natl Cancer Inst 2001 93 309 14 11181778 10.1093/jnci/93.4.309
Vleugel MM Greijer AE Shvarts A van der Groep P van Berkel M Aarbodem Y van Tinteren H Harris AL van Diest PJ van der Wall E Differential prognostic impact of hypoxia induced and diffuse HIF-1alpha expression in invasive breast cancer J Clin Pathol 2005 58 172 7 15677538 10.1136/jcp.2004.019885
Aebersold DM Burri P Beer KT Laissue J Djonov V Greiner RH Semenza GL Expression of hypoxia-inducible factor-1alpha: a novel predictive and prognostic parameter in the radiotherapy of oropharyngeal cancer Cancer Res 2001 61 2911 2916 11306467
|
16035955
|
PMC1190162
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Jul 21; 5:84
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-84
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-881604864610.1186/1471-2407-5-88Case ReportCalcitonin-producing well-differentiated neuroendocrine carcinoma (carcinoid tumor) of the urinary bladder: case report Mascolo Massimo [email protected] Vincenzo [email protected] Chiara [email protected] Giorgio [email protected] Rosa Gaetano [email protected] Luigi [email protected] Department of Biomorphological and Functional Sciences, Pathology Section, University of Naples "Federico II", via S. Pansini 5, 80131, Naples, Italy2 Department of Urology, University of Naples "Federico II", via S. Pansini 5, 80131, Naples, Italy2005 27 7 2005 5 88 88 7 4 2005 27 7 2005 Copyright © 2005 Mascolo et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The occurrence of calcitonin-secreting primary carcinoid tumor of the urinary bladder is extremely rare.
Case presentation
The case of a 68-year-old male with carcinoid tumor arising in the urinary bladder is presented. Transurethral resection of a polypoid small tumor 0.4 cm in diameter was performed. Immunohistochemical study using neuroendocrine markers allowed a straightforward diagnosis of a low-grade neuroendocrine carcinoma (carcinoid tumor) of the urinary bladder. Immunohistochemistry demonstrated calcitonin immunoreactivity in the most of the tumor cells.
Conclusion
This tumor shows specific clinical, macroscopical and histological features and must be considered in the differential diagnosis of bladder neoplasms.
==== Body
Background
The occurrence of primary carcinoid tumor of the urinary bladder is extremely rare. Two cases of these neoplasms have been reported recently by Martignoni and Eble [1]. According to a critical review of the literature performed by these authors, only four [2-5] of twelve previously reported cases showed convincing evidence od neuroendocrine differentiation and might therefore be regarded as pure carcinoid tumors of the urinary bladder. We report a case of a primary calcitonin-producing carcinoid tumor of the urinary bladder.
Case presentation
A 68-year-old man was admitted in January 2004 for a macroscopic hematuria. The patient had no relevant past medical or family history, and laboratory data were all within normal limits. A cystoscopic examination was performed and a transurethral resection of a tiny, sessile polypoid lesion of the bladder neck was carried out. Clinical evaluation included ultrasonography and computerized tomography (CT) scan. The patients was seen at last follow-up 14 months after diagnosis and he was completely negative.
Grossly, the tumor was a well-circumscribed, small ovoid nodule of 0.4 cm in diameter. Histologically a polypoid tumor appeared to be covered by slightly attenuated normal urothelium. The tumor filled-up the lamina propria showing a predominantly glandular arrangement (Fig 1A), with scanty anastomosing cords of cuboidal cells with fairly regular nuclei. Foci of von Brunn nests were seen at the periphery near the urothelium (Fig 1B). Mitoses and necrosis were not seen. A panel of immunostains, including antibodies against keratin 7, βHCG, chromogranin, synaptophysin, S-100 protein, NSE, serotonin, calcitonin, TTF-1, progesteron, and p53 protein was applied to representative sections of the tumor using the avidin-biotin complex technique (tab 1).
Immunohistochemically the tumor strongly reacted with synaptophysin (Fig 1C), chromogranin (Fig 1D), calcitonin (Fig 1E), keratin 7 (Fig 1F), and NSE. βHCG was positive in about 10% of the neoplastic cells. p53 protein, progesteron, S-100 protein, serotonin and TTF-1 were negative.
Conclusion
The urinary bladder can be the site of tumors exhibiting various degrees of endocrine differentiation. It has become the most common site of extrapulmonary small cell undifferentiated carcinoma [6]. The spectrum of endocrine tumors of the urinary bladder has widened considerably in the past concerning both its morphological feature and its clinical behaviour. The better differentiated members of this group, carcinoid tumors, have an organoid arrangement, with trabecular and glandular patterns of growth, and its origin in the urinary bladder is extremely rare. To the best of our knowledge only 6 cases of primary carcinoid tumor of the bladder have been reported [1-5].
The present case shows clinically and microscopically the same features showed by two cases of carcinoid of urinary bladder reported recently by Martignoni and Eble [1]. Their review of the literature disclosed 4 previously reported pure carcinoid tumors of the urinary bladder. These 6 cases and the present case show a striking clinical and pathological overlapping, as shown in tab 2.
All 6 cases presented with hematuria, most of the patients aged of seven decade. All tumors were small, polypoid nodules at cystoscopy and most of them were localized in the neck of the bladder or trigone. All previously reported 6 tumors and the present case show a glandular architecture. Previous study, and our results have demonstrated the neuroendocrine differentiation. Strikingly the present case showed, immunohistochemically, a strongly reactivity with calcitonin. To the best of our knowledge this is the first case of a well-differentiated neuroendocrine carcinoma of the urinary bladder reported positive with this neuropeptide.
Some tumors could simulate carcinoid of the bladder, in particular the nested variant of transitional cell carcinoma; differences in the cytologic features and immunohistochemistry should establish the diagnosis. The distinction between carcinoid tumor and inverted papilloma with glandular differentiation can be very difficult, however strong staining for neuroendocrine markers support the diagnosis of carcinoid tumor. Like the classic carcinoid tumors of the appendix, small bowel, or lung, all of the carcinoid tumors of the bladder are histologically and clinically not aggressive.
In conclusion, we here add another case of carcinoid tumor of the urinary bladder to the existing literature, this tumor shows specific clinical, macroscopical and histological features and must be considered in the differential diagnosis of bladder neoplasms.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MM participated to write the manuscript
VA operated the patient and participated to draft the manuscript
CM carried out imunohistochemical study
GN operated the patient and participated to draft the manuscript
GDR participated in the design of the study
LI participated in the design of the study
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Figure 1 Composite figure of carcinoid tumor of urinary bladder. A) a glandular arrangement of tumor cells is evident. B) normal urothelium lined the tumor. A von Brunn nest entrapped by the tumor is seen. C-D-E) diffusely positive synaptophysin, chromogranin, and calcitonin immunostaining of tumor cells, respectively. F) urothelium and tumor cells strongly reacted with keratin 7.
Table 1 Primary antibodies used for immunophenotyping
Antibody Manufacturer Dilution Method
Keratin 7 DAKO, Carpinteria, CA, USA 1:200 ABC
Chorionic Gonadotropin DAKO, Carpinteria, CA, USA 1:200 ABC
Chromogranin A DAKO, Carpinteria, CA, USA 1:50 ABC
Synaptophysin DAKO, Carpinteria, CA, USA 1:20 ABC
S-100 DAKO, Carpinteria, CA, USA 1:200 ABC
NSE DAKO, Carpinteria, CA, USA 1:8000 ABC
Serotonin DAKO, Carpinteria, CA, USA 1:100 ABC
Calcitonin Santa Cruz Biotechnology Inc, Santa Cruz, CA 1:100 ABC
Thyroid transcriptional factor DAKO, Carpinteria, CA, USA 1:100 ABC
Progesteron DAKO, Carpinteria, CA, USA Ready to use ABC
p-53 protein DAKO, Carpinteria, CA, USA 1:50 ABC
Table 2 Clinicopathologic profiles of patients with carcinoid tumor of the urinary bladder.
Author/Year/Reference Age/Sex Size (cm) Symptoms Cystoscopy Location Microscopy
Colby 1980 (2) 30/M 0.3 Hematuria Polypoid Neck Glandular
Walker et al 1992 (4) 62)F 1.2 Hematuria Polypoid Trigone Glandular
Stanfield et al 1994 (5) 54/F 0.9 Hematuria Polypoid Neck Glandular
Burgess et al 2000 (3) 61/F 0.3 Hematuria Polypoid Trigone Glandular
Martignoni et al 2003 (1) 69/M 0.3 Hematuria Polypoid Neck Glandular
Martignoni et al 2003 (1) 47/M 0.7 Hematuria Polypoid Neck Glandular
Insabato et al (present study) 68/M 0.4 Hematuria Polypoid Neck Glandular
==== Refs
Martignoni G Eble JN Carcinoid tumors of the urinary bladder. Immunohistochemical study of 2 cases and review of the literature Arch Pathol Lab Med 2003 127 22 24 12562289
Colby TV Carcinoid tumor of the bladder Arch Pathol Lab Med 1980 104 199 200 6892681
Burgess NA Lewis DC Matthews PN Primary carcinoid of the bladder Br J Urol 1992 69 213 214 1537038
Walker BF Someren A Kennedy JC Nicholas EM Primary carcinoid tumor of the urinary bladder Arch Pathol Lab Med 1992 116 1217 1220 1444752
Stanfield BC Grimes MM Kay S Primary carcinoid of the bladder arising beneath an inverterd papilloma Arch Pathol Lab Med 1994 118 666 667 8204019
Eble JN Young RH Carcinoma of the urinary bladder: a review of its diverse morphology Semin Diagn Pathol 1997 14 98 108 9179971
|
16048646
|
PMC1190163
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Jul 27; 5:88
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-88
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-911604865610.1186/1471-2407-5-91Research ArticleIntensity modulated radiotherapy for high risk prostate cancer based on sentinel node SPECT imaging for target volume definition Ganswindt Ute [email protected] Frank [email protected] Stefan [email protected] Kai [email protected] Stefan [email protected] Ilse [email protected] Mattias [email protected] Markus [email protected] Aristotelis [email protected] Arnulf [email protected] Roland [email protected] Wilfried [email protected] Michael [email protected] Claus [email protected] Department of Radiation Oncology, University of Tübingen, Tübingen, Germany2 Department of Radiation Oncology, Biomedical Physics, University of Tübingen, Tübingen, Germany3 Department of Urology, University of Tübingen, Tübingen, Germany4 Department of Nuclear Medicine, University of Tübingen, Tübingen, Germany5 Department of Radiation Oncology, University of Düsseldorf, Düsseldorf, Germany2005 28 7 2005 5 91 91 14 4 2005 28 7 2005 Copyright © 2005 Ganswindt et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The RTOG 94-13 trial has provided evidence that patients with high risk prostate cancer benefit from an additional radiotherapy to the pelvic nodes combined with concomitant hormonal ablation. Since lymphatic drainage of the prostate is highly variable, the optimal target volume definition for the pelvic lymph nodes is problematic. To overcome this limitation, we tested the feasibility of an intensity modulated radiation therapy (IMRT) protocol, taking under consideration the individual pelvic sentinel node drainage pattern by SPECT functional imaging.
Methods
Patients with high risk prostate cancer were included. Sentinel nodes (SN) were localised 1.5–3 hours after injection of 250 MBq 99mTc-Nanocoll using a double-headed gamma camera with an integrated X-Ray device. All sentinel node localisations were included into the pelvic clinical target volume (CTV). Dose prescriptions were 50.4 Gy (5 × 1.8 Gy / week) to the pelvis and 70.0 Gy (5 × 2.0 Gy / week) to the prostate including the base of seminal vesicles or whole seminal vesicles. Patients were treated with IMRT. Furthermore a theoretical comparison between IMRT and a three-dimensional conformal technique was performed.
Results
Since 08/2003 6 patients were treated with this protocol. All patients had detectable sentinel lymph nodes (total 29). 4 of 6 patients showed sentinel node localisations (total 10), that would not have been treated adequately with CT-based planning ('geographical miss') only. The most common localisation for a probable geographical miss was the perirectal area. The comparison between dose-volume-histograms of IMRT- and conventional CT-planning demonstrated clear superiority of IMRT when all sentinel lymph nodes were included. IMRT allowed a significantly better sparing of normal tissue and reduced volumes of small bowel, large bowel and rectum irradiated with critical doses. No gastrointestinal or genitourinary acute toxicity Grade 3 or 4 (RTOG) occurred.
Conclusion
IMRT based on sentinel lymph node identification is feasible and reduces the probability of a geographical miss. Furthermore, IMRT allows a pronounced sparing of normal tissue irradiation. Thus, the chosen approach will help to increase the curative potential of radiotherapy in high risk prostate cancer patients.
==== Body
Background
Treatment strategies for localised prostate cancer have to address issues of local control, prevention of distant spread and treatment of microscopic locoregional spread. It has been shown that local control is critically interrelated with the prevention of distant spread, since late metastases occur in patients with local failure [1]. In order to increase the efficacy of radiotherapy, many recent trials suggest the strategy of dose escalation [2-19].
Particularly the availability of three-dimensional conformal radiation treatment (3D-CRT) and further development of intensity modulated radiation treatment (IMRT) allow the application of higher radiation doses to the prostate without increased toxicities to normal tissues. The available data suggest that prostate cancer especially in patients with intermediate risk profile benefit from dose escalation (reviewed in [20]). On the other hand, patients with a high risk profile do not show such a steep dose response relationship because of the risks of loco- regional and ultimately systemic spread.
A further approach to increase local control rates and the rates of biochemical disease control is the combination of radiation with anti-hormonal treatment. Several recently published results of larger trials [21-23] could demonstrate a clear benefit from a combined treatment in patients with locally advanced prostate cancer. Most of these trials have employed radiation portals covering the locoregional lymphatic drainage in addition to the prostate using a standardised anatomical definition.
Of special importance in this regard is the four-armed RTOG 94-13 trial comparing neoadjuvant vs. adjuvant anti-hormonal treatment and radiotherapy of the whole pelvis vs. the prostate only. Patients with an estimated risk of pelvic lymph node involvement > 15% had a significant benefit when treated with neoadjuvant and concurrent hormonal ablation in combination with radiotherapy to the prostate (70.2 Gy) and the pelvic lymph nodes (50.4 Gy). The most significant difference in 4-year progression free survival rates was observed between the groups of whole pelvis (61%) and prostate only (45%) irradiated patients when treated with neoadjuvant and concurrent hormonal ablation [23].
Thus, besides dose escalation strategies or combination with anti-hormonal treatments, adjuvant coverage of the pelvic lymph nodes may be of importance for an optimised tumour control.
Currently, the definition of an adjuvant pelvic target volume is derived from non-individualised anatomical lymphatic drainage patterns. Radiation planning and therapy is mostly performed as 3D-CT based treatment along the ICRU 50 guidelines.
However, the lymphatic drainage from the prostate is highly variable. In this regard, data on sentinel lymph nodes analysis in prostate cancer are of key importance. To optimise lymphadenectomy techniques, Wawroschek et al. [24,25] and Weckermann et al. [26] tested the sentinel lymph node (SN) identification with prostate lymphoscintigraphy and radio-guided surgery in large cohort. The results demonstrated that sentinel node identification in prostate cancer is feasible with a sensitivity between 93 and 96% and enables higher detection rates of micrometastases. However, the authors noted a strong individual variability of the lymphatic drainage in their patient cohort. Interestingly, only 44.2% of detectable cases with positive nodes were found in the obturator fossa, the most common area for a limited lymphadenectomy. These findings correspond to surgical data of 103 patients after extended lymphadenectomy [27]. In the latter, the most common areas of lymph node metastases were the external and internal iliac, followed by obturator, common iliac and presacral regions.
Considering these results, the definition of standardised target volumes covering the pelvic lymph nodes is associated with several problems. Trying to cover all possible areas of lymphatic drainage increases the probability of toxicities. In contrast, by limiting the target volume to the most commonly involved lymph node areas increases the risk of incomplete target volume coverage.
To overcome the limitations described above, we started an IMRT trial for high risk prostate cancer with a target volume definition based on the average distribution of positive lymph nodes in prostate cancer patients complemented by sentinel node functional imaging with SPECT.
Methods
Patients
Patients with histologically proven high risk prostate cancer, but cN0 cM0 stage, were included after providing informed consent. High risk profile was defined as T3 or T4 stage (all) or PSA > 20 ng/mL (all) or PSA 10 – 20 ng/mL with Gleason Scores = 7. Besides PSA level, rectal-digital examination and biopsy, the pre-therapeutic staging included a computed tomographic and/or magnetic resonance imaging of the abdomen and pelvis, an ultrasound of the prostate, a total body bone scan and X-ray of the chest. The clinical TNM stage was determined based on these staging results.
Localisation of sentinel nodes
To permit a 3D-localisation of the sentinel nodes, transmission and emission scans were acquired using a double-headed gamma camera with an integrated X-Ray device (Millennium VG & Hawkeye®, GE Medical Systems Europe, Buc Cedex, France) 1.5–3 hours after transrectal intraprostatic injection of 250 MBq 99mTc-Nanocoll. A single ultrasound-guided central application was performed per prostate lobe. After the anatomical-functional image fusion, the scans were analysed with respect to 3D-localisation and number of sentinel nodes. The sentinel nodes were three-dimensionally located by an experienced member of the nuclear medicine department (K.E., R.B.). For a systematic topographic evaluation of lymph node localisation, we used the Cross-sectional Nodal Atlas published by Martinez-Monge 1999 [28](figure 1).
Treatment planning
Intensity modulated radiotherapy planning was based on three CT scans of the pelvis in supine (2 patients) and prone (4 patients, using a belly board) position at 3 mm slice spacing from two cm below the ischeal tuberosities to the L 4/5 interspace. For a comfortable bladder filling, patients were instructed to avoid urination during the last 45 minutes before the procedure. The CT datasets were transferred to the Pinnacle® treatment planning system (Philips Medical Systems, DA Best, Netherlands) for segmentation. Clinical target volume (CTV) and organs at risk (OAR) were outlined in each CT image. OAR included rectum, colon/sigmoid, small bowel, bladder and hips. Image fusion was done by adjusting the bone structures of the pelvis. Finally, a single enclosing contour was derived from the three outlined CTV/OAR. For the resulting planning target volume (PTV) the enclosing CTV contour was expanded with a 3D safety margin of 7 mm.
Since two different risk areas are treated simultaneously, the term first- and second-order CTV will be used in the following. The first-order CTV included the prostate and the base of seminal vesicles. In cases of T stage ≥ 3 or an estimated risk for an involvement of the seminal vesicles > 15% (Partin nomograms) [29], seminal vesicles were included into the boost volume.
The second-order CTV included the first order CTV and routinely the obturator and hypogastric lymph nodes (following the nomenclature introduced by Martinez-Monge: Periprostatic and seminal vesicle lymphatic plexus, the parts of the perivesical and the perirectal lymphatic plexus nearby the prostate, the ventrocranial parts of the internal pudendal and inferior rectal nodes), the internal and external iliac lymph nodes (from the bifurcation of the common iliac lymph artery at the level of the upper sacroiliac joints, to the crossing point of the external iliac artery and the inguinal ligament) and the sacral nodes anterior the first and second sacral segment. In addition to this regular second-order CTV all detected sentinel node localisations were outlined separately and involved into the IMRT target volume definition.
Dose prescriptions were 50.4 Gy (5 × 1.8 Gy / week) to the pelvic nodes and 70.0 Gy (5 × 2.0 Gy / week) to the prostate and the base of seminal vesicles or whole seminal vesicles by an integrated boost planning. After forward decision regarding gantry angles, an inverse treatment planning was carried out by the Hyperion® software package (University of Tübingen, Germany) [30]. Based on a class solution, six IMRT fields were used for the initial 28 fractions and 5 fields for the ensuing 7 fractions. The following normal tissue constraints were chosen: Rectum, 58–65 Gy serial dose constraint; Bladder, 50–56 Gy serial dose constraint; Small Bowel, 20–22 Gy mean dose constraint and 50–54 Gy maximum dose constraint. The dosimetrics were calculated by Monte Carlo dose calculation.
Plans were produced for a 15 MV linear accelerator (Elekta Oncology Systems®, Crawley, UK) for delivery with a multileaf collimator system using a step and shoot technique. The IMRT plans were compared with 3D-CRT plans (four-field-technique, gantry angles 0°, 90°, 180°, 270°) using the Pinnacle® treatment planning system based on the identical CTV. Dose volume histograms (DVH) were calculated for each plan. The PTV dose ranges were calculated with the minimum PTV dose defined as dose received by ≥ 99% of the PTV and the maximum dose defined as the dose received by = 1% of the PTV. The volume percentages of small bowel, large bowel (sigmoid/colon), rectum and bladder irradiated with 63 Gy, 56 Gy, 35 Gy, 14 Gy were recorded for each technique. The DVH comparisons were done for each patient using the summarised plan.
Verification
Prior to each radiation fraction, the isocenter was verified by portal imaging from 0° and 90° gantry angle. An error of more than 2 mm was corrected. The administered monitor units needed for one daily verification were included in the initial dose calculation.
Analysis of toxicity
Acute gastrointestinal, genitourinary and skin toxicities (RTOG criteria) were documented by standard forms weekly during RT and at least after three months.
Results
Patients
Since August 2003, six patients with T1c-3b high risk prostate cancers (figure 2) were evaluated. No patient had undergone a staging lymphadenectomy of the pelvic lymph nodes before. All patients had normal pre-therapeutic diagnostic findings with a clinical N0 M0 stage. All patients received neoadjuvant and concurrent hormonal ablation, which was recommended to continue adjuvantly for three years. In all cases neoadjuvant treatment had been started before the first visit in hospital.
Localisation of sentinel nodes
The transrectal intraprostatic injection of 99mTc-Nanocoll was performed in all patients without any complications. After radioisotope injection and image fusion, each patient had detectable sentinel lymph nodes. The number per patient ranged from two to nine detected lymph nodes. Altogether, a total of twenty-nine lymph nodes could be identified for all six patients (figure 3). The most common localisations of identified lymph nodes were external iliac (9), followed by internal iliac (6) and perirectal lymphatic plexus (6). Further localisations were common iliac (2), sacral (2), internal pudendal (1), seminal vesicle lymphatic plexus (1), superior rectal (1) and left paraaortic (1).
IMRT treatment planning
The first two patients were planned and treated in supine position. Since prone position in combination with a belly board allowed easier sparing of small bowel [31], we changed to the prone position for the following four patients. All detected sentinel lymph nodes could be included into the second-order CTV. The IMRT plans carried out by Hyperion® produced an average of 55 segments (total/fraction) and 495 Monitor Units (total/fraction) for the first 28 fractions and an average of 35 segments (total/fraction) and 346 Monitor Units (total/fraction) for the ensuing 7 fractions. Thus, fast delivery and practical treatment times were possible. A class solution delivered stable and comparable results for all patients. Detailed characteristics of IMRT technique are shown in figure 4.
The minimum PTV dose ranged from 40 to 48.1 Gy for the second-order PTV, including the pelvic nodes. The maximum PTV dose ranged from 72.4 to 74.2 Gy for the prostate gland (figure 5). The critical dose of 60 Gy to the whole bladder volume was not reached in any patient.
A critical dose to the small bowel of 50 Gy was reached in two patients in very small volumes (0.5% and 1%, respectively). Large bowel doses ≥ 50 Gy were seen in very limited volumes (0%, 2.7%, 2.7%, 4.1%, 6.6% and 13.6%). The single event of a higher dose volume of 13.6% was seen in one patient (No. 5) with an adherent sigmoid loop in close vicinity to the prostate. Rectum volumes irradiated with more than 56 Gy ranged from 16.3% to 52.1%, with more than 63 Gy from 9.5% to 36.6%. The detailed analysis of irradiated organs at risk volume percentages is given in table 6.
Influence of sentinel node localisation on target volume definition
4 of 6 patients had sentinel node localisations (total 10 of 29 sentinel lymph nodes) that would not have been treated adequately with conventional target volume definition ('geographical miss'). Thus, the target volume definition was modified by sentinel lymph node information in 4 of 6 patients. The results of numbers and localisations of sentinel lymph nodes are shown in table 3, grey shaded fields show the number and localisation of geographical misses.
The lymph node area associated with the highest probability of a 'geographical miss' was the perirectal lymph node group with 6 identified sentinel nodes partly localised nearby the dorsal circumference of the rectal wall. Single sentinel nodes were found in the internal pudendal, sacral, superior rectal and left paraaortic region. None of those was covered by standardised planning target volumes. The identified sentinel lymph node in the left paraaortic area was localised at the L4/5 interspace. Because the CT scans showed a lymph node of 11 mm diameter at this localisation, we decided to include it.
DVH comparison between IMRT and 3D-CRT treatment planning
The patients were treated using IMRT technique (figure 4). Additionally, a 3D-conformal treatment planning with a four-field-technique (gantry angles 0°, 90°, 180°, 270°) was produced for each patient based on the identical CT datasets with the identical outlined CTV and OAR.
The comparison of minimum and maximum PTV doses between 3D-CRT and IMRT is shown in figure 5. IMRT minimum PTV dose ranged from 40 to 48.1 Gy, 3D-CRT minimum PTV dose from 47 to 49.3 Gy for the 2nd order PTV. IMRT minimum PTV dose ranged from 62.4 to 67 Gy, 3D-CRT minimum PTV dose from 67.4 to 68.9 Gy for the 1st order PTV. IMRT maximum PTV dose ranged from 72.4 to 74.2 Gy, 3D-CRT maximum PTV dose from 71.5 to 72.6 Gy for the 1st order PTV.
The DVH comparisons between IMRT and 3D-CRT in regard to small bowel, large bowel and rectum are shown in figure 6. Small bowel volumes irradiated with 14Gy and 35 Gy increased in four patients (in part very slightly) by using IMRT and decreased clearly in two patients with a volume reduction in more than 10%. With IMRT the critical dose to the small bowel of 50 Gy was seen in two patients with volumes of 0.5% and 1%, respectively, compared to 3D-CRT with five patients with volumes between 0 and 25.3%. Concerning the large bowel, IMRT similarly caused increased low dose irradiated volumes, but significant smaller irradiated volumes at the critical dose of 50 Gy. The most important difference was seen in the irradiated rectum volumes. All patients showed a clear benefit from IMRT with a clear decrease in volumes irradiated with 56 Gy and 63 Gy. As an example a dose distribution of a single patient are shown in figure 7.
Acute toxicities
All patients completed the radiotherapy course after seven weeks with a dose of 70.0 Gy without treatment interruption. During radiotherapy interval, genitourinary toxicity RTOG Grade 1 occurred in all six patients, four patients increased to a genitourinary toxicity RTOG Grade 2. Gastrointestinal toxicity RTOG Grade 1 was seen in all six patients, two patients showed a gastrointestinal toxicity RTOG Grade 2. No gastrointestinal or genitourinary acute toxicity Grade 3 or 4 was noted. There were no toxicities seen concerning the skin. All patients reported erectile dysfunction, since neoadjuvant hormonal ablation had been started. After three months, two patients reported slight genitourinary symptoms (RTOG Grade 1), one of them had a slightly increased defecation frequency corresponding to a gastrointestinal toxicity RTOG Grade 1. The other four patients did not show any toxicity after three months.
Discussion
Although being still controversial, several data suggest that an adjuvant coverage of the pelvic lymph nodes increases the likelihood of tumour control in patients with high risk prostate cancer. Whereas several older trials suggested that larger irradiation portals may not increase the treatment efficacy [32-41], recent randomised and nonrandomised studies could demonstrate a benefit of elective pelvic lymph node irradiation [21-23,42-47].
Of special value is the randomised four-armed RTOG 94-13 trial, since it was designed to provide a final answer regarding the role of adjuvant lymph node coverage. It showed that high risk prostate cancer patients benefit from radiotherapy of the pelvic lymph nodes combined with neoadjuvant and concurrent hormonal ablation. A significant benefit was documented in terms of advanced 4-year progression free survival rates [23], but not for overall survival. Interestingly, there was no significant difference in toxicities between patients treated on any of the four arms that may be caused by improved treatment techniques in comparison to former studies. The role of pelvic node coverage is also underlined by the fact that the EORTC trial [21] randomised high risk and lymph node positive patients to receive concurrent and long term adjuvant hormonal ablation treatment or radiotherapy alone, which included 50 Gy to the pelvis and a 20 Gy boost to the prostate. A substantial benefit in terms of local control, disease specific survival and overall survival was noted for the group receiving anti-hormonal treatment. Regarding 5-year clinical disease-free survival rates of 74% and 40% (no hormonal ablation) respectively, it has to be assumed that combined treatment with anti-hormonal therapy and radiation of the pelvic lymph nodes improves outcome of high risk and lymph node positive patients. Comparable results were shown by the RTOG 92-02 trial (including patients with cT2c-4, PSA < 150 ng/mL) combining radiotherapy of the pelvic nodes and the prostate with anti-hormonal short vs. long term treatment. An overall survival advantage was seen in a subset analysis of patients with Gleason Scores 8 to 10 (81% vs. 71%) [22].
Regarding efficacy, available data make it highly likely that the subgroup of patients with high risk of pelvic lymph node involvement will benefit from this treatment. Thus, it can be assumed that any further improvement in staging and irradiation techniques will help to optimise outcome. The sentinel concept was mainly developed for an optimised surgical lymph node identification and dissection in patients with malignant melanoma [48]. This concept can also be perfectly applied to breast cancer [49], head & neck cancer [50] and probably also to prostate cancer [24,25]. Key work was done by Wawroschek and Harzmann providing evidence that 99mTc-Nanocoll based sentinel node identification is highly sensitive and specific. Similar results were obtained by us in a smaller cohort [51]. Therefore, sentinel image data may be used for radiation treatment planning.
The discussion on a potentially increased toxicity of larger irradiation portals was mainly supported by results from older studies, which were not based on modern CT planning. Thus, the favourable results from the RTOG 94-13 trial may in part be explained by optimised treatment techniques. Meanwhile, few clinical studies suggest superiority of IMRT when used for radiotherapy of the prostate alone [7]. However, even less data are available concerning an inclusion of the pelvic lymph nodes with IMRT. Nutting et al. [52] performed a theoretical trial in 10 patients and compared different IMRT treatment techniques among each other and to conventional 3D-CRT. The data provided in this study show eloquently that IMRT based irradiation protocols allow clear reduction of limiting normal tissue exposure. In a second paper from the same group, quality assurance aspects of clinical introduction comparing a step and shoot versus dynamic arch application were described. The authors concluded that the technique is feasible and neither of both application modi was superior [53]. Similarly, Mundt analysed the applicability of an IMRT based protocol for the treatment of pelvic lymph nodes in patients with gynaecological malignancies [54]. The authors conclude that IMRT based irradiation is associated with a diminished normal tissue toxicity profile and an optimised target coverage. However, in contrast to our work, both authors used anatomically based target volume definitions rather than an individualised sentinel lymph nodes based planning approach.
Our approach to combine sentinel data with IMRT addresses crucial aspects of target coverage and also toxicity. To reduce the likelihood of a 'geographical miss' we purposed the complete individualised coverage of lymphatic drainage using functional imaging. Possibly, individualised target volumes may cause increased toxicities. Not to mention, we observed a notable number of sentinel nodes nearby the dorsal rectal wall and their inclusion into the target volumes causes increased toxicities when a conventional four-field-technique is used (figure 8). Whereas the focus of 3D treatment planning is on target coverage, IMRT target coverage is compromised by the need of normal tissue sparing. Our own data hint at an optimal target volume coverage and showed a clear benefit in OAR sparing by using IMRT with low overall acute toxicities. Thus, IMRT is necessary to decrease toxicities resulting from individualised target volume definition.
Conclusion
In order to optimise the balance between increased probability of side effects and improved target volume coverage, we tested a sentinel node based IMRT approach. Based on a limited number of patients the following conclusions can be drawn: Sentinel functional imaging optimised target volume definition for IMRT is clearly feasible with no obvious increase in acute toxicity and suitable for larger numbers of patients. The probability of a geographical miss in target volume definition might be reduced by an individualised target volume definition. However, larger series are necessary to verify the expected benefit in regard to efficacy and toxicities.
Abbreviations
3D-CRT Three-dimensional conformal radiotherapy
CT Computed tomography
CTV Clinical target volume
DVH Dose volume histogram
Gy Gray
ICRU International Commission on Radiation Units and Measurements
IMRT Intensity modulated radiotherapy
MBq mega Becquerel
MV mega Volt
ng/mL nanogram/millilitre
OAR Organ at risk
PSA Prostate specific antigen
PTV Planning target volume
RTOG Radiation Therapy Oncology Group
SN Sentinel node
SPECT Single Photon Emission Computed Tomography)
TNM TNM stage classification (tumour/nodal/metastases stage)
Tc Technetium
vs. versus
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CB & UG planned, coordinated and conducted the study. KE, WB, FP took part in designing the study. UG, CB, MA, MBi analyzed the data. UG & CB prepared the manuscript. Medical care was covered by FP, SC, KE, SG, IH, AA, AS, RB, WB, MBa. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Figure 1 Terms and definitions of nodal groups according to Martinez-Monge et al. (1999). The obturator and hypogastric lymph nodes include periprostatic and seminal vesicle lymphatic plexus, parts of perivesical and the perirectal lymphatic plexus nearby the prostate, the ventrocranial parts of the internal pudendal and inferior rectal nodes
Figure 2 Patients characteristics.
Figure 3 Number and localisation of sentinel nodes. Localisation of lymph node areas not covered by a standard target volume definition plan are shaded grey (geographical misses).
Figure 4 IMRT planning results: 28 fractions including pelvic nodes with integrated boost (upper table) and 7 fractions without pelvic nodes (lower table). MU total / fraction adds up MU's per field and fraction (rounded), including additionally 3 MU's necessary for daily verification (not included in 'MU's per field and fraction)
Figure 5 PTV dose ranges 3D-CRT vs. IMRT. Minimum PTV dose defined as the dose received by = 99% of the PTV, maximum PTV dose defined as the dose received by = 1% of the PTV. * Dose contribution by the boost RT.
Figure 6 3D-CRT vs. IMRT. – DVH comparison of small bowel, large bowel and rectum. V63Gy, V56Gy, V50Gy, V35Gy, V14Gy denote the volume of a given organ (in %) irradiated with 63 Gy, 56 Gy, 50 Gy, 35 Gy, 14 Gy.
Figure 7 Planning example: Upper left : Sentinel lymph nodes left internal iliac (2), right internal iliac (1), perirectal on an axial SPECT image (prone position). Upper middle: IMRT planning: sentinel node localisations and second order PTV are outlined (prone position). The 95% isodose curve is shaped green. Upper right : 3D-CRT planning (four-field-box): sentinel node localisations and second order PTV are outlined (prone position). The 95% isodose curve is shaped green. Lower left: Dose-volume-histogram IMRT. Lower right: Dose-volume-histogram 3D-CRT
Figure 8 Axial CT slices of three example cases.
==== Refs
Coen JJ Zietman AL Thakral H Shipley WU Radical radiation for localized prostate cancer: local persistence of disease results in a late wave of metastases J Clin Oncol 2002 20 3199 3205 12149291 10.1200/JCO.2002.01.086
Hanks GE Hanlon AL Schultheiss TE Pinover WH Movsas B Epstein BE Hunt MA Dose escalation with 3D conformal treatment: five year outcomes, treatment optimization, and future directions Int J Radiat Oncol Biol Phys 1998 41 501 510 9635695 10.1016/S0360-3016(98)00089-3
Hanks GE Hanlon AL Pinover WH Horwitz EM Schultheiss TE Survival advantage for prostate cancer patients treated with high-dose three-dimensional conformal radiotherapy Cancer J Sci Am 1999 5 152 158 10367171
Hanks GE Hanlon AL Epstein B Horwitz EM Dose response in prostate cancer with 8-12 years' follow-up Int J Radiat Oncol Biol Phys 2002 54 427 435 12243818 10.1016/S0360-3016(02)02954-1
Zelefsky MJ Leibel SA Gaudin PB Kutcher GJ Fleshner NE Venkatramen ES Reuter VE Fair WR Ling CC Fuks Z Dose escalation with three-dimensional conformal radiation therapy affects the outcome in prostate cancer Int J Radiat Oncol Biol Phys 1998 41 491 500 9635694 10.1016/S0360-3016(98)00091-1
Zelefsky MJ Fuks Z Hunt M Lee HJ Lombardi D Ling CC Reuter VE Venkatraman ES Leibel SA High dose radiation delivered by intensity modulated conformal radiotherapy improves the outcome of localized prostate cancer J Urol 2001 166 876 881 11490237 10.1097/00005392-200109000-00017
Zelefsky MJ Fuks Z Hunt M Yamada Y Marion C Ling CC Amols H Venkatraman ES Leibel SA High-dose intensity modulated radiation therapy for prostate cancer: early toxicity and biochemical outcome in 772 patients Int J Radiat Oncol Biol Phys 2002 53 1111 1116 12128109 10.1016/S0360-3016(02)02857-2
Symon Z Griffith KA McLaughlin PW Sullivan M Sandler HM Dose escalation for localized prostate cancer: substantial benefit observed with 3D conformal therapy Int J Radiat Oncol Biol Phys 2003 57 384 390 12957249 10.1016/S0360-3016(03)00569-8
Pollack A Zagars GK Smith LG Lee JJ von Eschenbach AC Antolak JA Starkschall G Rosen I Preliminary results of a randomized radiotherapy dose-escalation study comparing 70 Gy with 78 Gy for prostate cancer J Clin Oncol 2000 18 3904 3911 11099319
Pollack A Zagars GK Starkschall G Antolak JA Lee JJ Huang E von Eschenbach AC Kuban DA Rosen I Prostate cancer radiation dose response: results of the M. D. Anderson phase III randomized trial Int J Radiat Oncol Biol Phys 2002 53 1097 1105 12128107 10.1016/S0360-3016(02)02829-8
Valicenti R Lu J Pilepich M Asbell S Grignon D Survival advantage from higher-dose radiation therapy for clinically localized prostate cancer treated on the Radiation Therapy Oncology Group trials J Clin Oncol 2000 18 2740 2746 10894874
Pollack A Smith LG von Eschenbach AC External beam radiotherapy dose response characteristics of 1127 men with prostate cancer treated in the PSA era Int J Radiat Oncol Biol Phys 2000 48 507 512 10974469 10.1016/S0360-3016(00)00620-9
Hanks GE Hanlon AL Pinover WH Horwitz EM Price RA Schultheiss T Dose selection for prostate cancer patients based on dose comparison and dose response studies Int J Radiat Oncol Biol Phys 2000 46 823 832 10705002 10.1016/S0360-3016(99)00498-8
Kuban DA Thames HD Levy LB Horwitz EM Kupelian PA Martinez AA Michalski JM Pisansky TM Sandler HM Shipley WU Zelefsky MJ Zietman AL Long-term multi-institutional analysis of stage T1-T2 prostate cancer treated with radiotherapy in the PSA era Int J Radiat Oncol Biol Phys 2003 57 915 928 14575822 10.1016/S0360-3016(03)00632-1
Pinover WH Hanlon AL Horwitz EM Hanks GE Defining the appropriate radiation dose for pretreatment PSA < or = 10 ng/mL prostate cancer Int J Radiat Oncol Biol Phys 2000 47 649 654 10837947 10.1016/S0360-3016(00)00465-X
Kupelian PA Buchsbaum JC Reddy CA Klein EA Radiation dose response in patients with favorable localized prostate cancer (Stage T1-T2, biopsy Gleason < or = 6, and pretreatment prostate-specific antigen < or = 10) Int J Radiat Oncol Biol Phys 2001 50 621 625 11395228 10.1016/S0360-3016(01)01466-3
Fiveash JB Hanks G Roach M Wang S Vigneault E McLaughlin PW Sandler HM 3D conformal radiation therapy (3DCRT) for high grade prostate cancer: a multi-institutional review Int J Radiat Oncol Biol Phys 2000 47 335 342 10802357 10.1016/S0360-3016(00)00441-7
Hurwitz MD Schnieder L Manola J Beard CJ Kaplan ID D'Amico AV Lack of radiation dose response for patients with low-risk clinically localized prostate cancer: a retrospective analysis Int J Radiat Oncol Biol Phys 2002 53 1106 1110 12128108 10.1016/S0360-3016(02)02885-7
Jani AB Hand CM Pelizzari CA Roeske JC Krauz L Vijayakumar S Biological-effective versus conventional dose volume histograms correlated with late genitourinary and gastrointestinal toxicity after external beam radiotherapy for prostate cancer: a matched pair analysis BMC Cancer 2003 3 16 12744725 10.1186/1471-2407-3-16
Ganswindt U Paulsen F Anastasiadis AG Stenzl A Bamberg M Belka C 70 Gy or more: which dose for which prostate cancer? J Cancer Res Clin Oncol 2005
Bolla M Collette L Blank L Warde P Dubois JB Mirimanoff RO Storme G Bernier J Kuten A Sternberg C Mattelaer J Lopez Torecilla J Pfeffer JR Lino Cutajar C Zurlo A Pierart M Long-term results with immediate androgen suppression and external irradiation in patients with locally advanced prostate cancer (an EORTC study): a phase III randomised trial Lancet 2002 360 103 106 12126818 10.1016/S0140-6736(02)09408-4
Hanks GE Pajak TF Porter A Grignon D Brereton H Venkatesan V Horwitz EM Lawton C Rosenthal SA Sandler HM Shipley WU Phase III trial of long-term adjuvant androgen deprivation after neoadjuvant hormonal cytoreduction and radiotherapy in locally advanced carcinoma of the prostate: the Radiation Therapy Oncology Group Protocol 92-02 J Clin Oncol 2003 21 3972 3978 14581419 10.1200/JCO.2003.11.023
Roach M DeSilvio M Lawton C Uhl V Machtay M Seider MJ Rotman M Jones C Asbell SO Valicenti RK Han S Thomas CRJ Shipley WS Phase III Trial Comparing Whole-Pelvic Versus Prostate-Only Radiotherapy and Neoadjuvant Versus Adjuvant Combined Androgen Suppression: Radiation Therapy Oncology Group 9413 J Clin Oncol 2003 21 1904 1911 12743142 10.1200/JCO.2003.05.004
Wawroschek F Wagner T Hamm M Weckermann D Vogt H Markl B Gordijn R Harzmann R The influence of serial sections, immunohistochemistry, and extension of pelvic lymph node dissection on the lymph node status in clinically localized prostate cancer Eur Urol 2003 43 132 6; discussion 137. 12565770 10.1016/S0302-2838(02)00450-5
Wawroschek F Vogt H Wengenmair H Weckermann D Hamm M Keil M Graf G Heidenreich P Harzmann R Prostate lymphoscintigraphy and radio-guided surgery for sentinel lymph node identification in prostate cancer. Technique and results of the first 350 cases Urol Int 2003 70 303 310 12740496 10.1159/000070140
Weckermann D Wawroschek F Harzmann R Is there a need for pelvic lymph node dissection in low risk prostate cancer patients prior to definitive local therapy? Eur Urol 2005 47 45 50; discussion 50-1 15582248 10.1016/j.eururo.2004.07.017
Heidenreich A Varga Z Von Knobloch R Extended pelvic lymphadenectomy in patients undergoing radical prostatectomy: high incidence of lymph node metastasis J Urol 2002 167 1681 1686 11912387 10.1097/00005392-200204000-00023
Martinez-Monge R Fernandes PS Gupta N Gahbauer R Cross-sectional nodal atlas: a tool for the definition of clinical target volumes in three-dimensional radiation therapy planning Radiology 1999 211 815 828 10352611
Partin AW Mangold LA Lamm DM Walsh PC Epstein JI Pearson JD Contemporary update of prostate cancer staging nomograms (Partin Tables) for the new millennium Urology 2001 58 843 848 11744442 10.1016/S0090-4295(01)01441-8
Alber M Nusslin F Optimization of intensity modulated radiotherapy under constraints for static and dynamic MLC delivery Phys Med Biol 2001 46 3229 3239 11768502 10.1088/0031-9155/46/12/311
Adli M Mayr NA Kaiser HS Skwarchuk MW Meeks SL Mardirossian G Paulino AC Montebello JF Gaston RC Sorosky JI Buatti JM Does prone positioning reduce small bowel dose in pelvic radiation with intensity-modulated radiotherapy for gynecologic cancer? Int J Radiat Oncol Biol Phys 2003 57 230 238 12909238 10.1016/S0360-3016(03)00409-7
Aristizabal SA Steinbronn D Heusinkveld RS External beam radiotherapy in cancer of the prostate. The University of Arizona experience Radiother Oncol 1984 1 309 315 6505267
Asbell SO Krall JM Pilepich MV Baerwald H Sause WT Hanks GE Perez CA Elective pelvic irradiation in stage A2, B carcinoma of the prostate: analysis of RTOG 77-06 Int J Radiat Oncol Biol Phys 1988 15 1307 1316 3058656
Asbell SO Martz KL Shin KH Sause WT Doggett RL Perez CA Pilepich MV Impact of surgical staging in evaluating the radiotherapeutic outcome in RTOG #77-06, a phase III study for T1BN0M0 (A2) and T2N0M0 (B) prostate carcinoma Int J Radiat Oncol Biol Phys 1998 40 769 782 9531360 10.1016/S0360-3016(97)00926-7
Harisiadis L Veenema RJ Senyszyn JJ Puchner PJ Tretter P Romas NA Chang CH Lattimer JK Tannenbaum M Carcinoma of the prostate: treatment with external radiotherapy Cancer 1978 41 2131 2142 657085
Hill DR Crews QEJ Walsh PC Prostate carcinoma: radiation treatment of the primary and regional lymphatics Cancer 1974 34 156 160 4834173
Neglia WJ Hussey DH Johnson DE Megavoltage radiation therapy for carcinoma of the prostate Int J Radiat Oncol Biol Phys 1977 2 873 883 412817
Rosen E Cassady JR Connolly J Chaffey JT Radiotherapy for prostate carcinoma: the JCRT experience (1968-1978). II. Factors related to tumor control and complications Int J Radiat Oncol Biol Phys 1985 11 723 730 3980269
Rounsaville MC Green JP Vaeth JM Purdon RP Heltzel MM Prostatic carcinoma: limited field irradiation Int J Radiat Oncol Biol Phys 1987 13 1013 1020 3597143
Zagars GK von Eschenbach AC Johnson DE Oswald MJ Stage C adenocarcinoma of the prostate. An analysis of 551 patients treated with external beam radiation Cancer 1987 60 1489 1499 3113715
Zagars GK von Eschenbach AC Johnson DE Oswald MJ The role of radiation therapy in stages A2 and B adenocarcinoma of the prostate Int J Radiat Oncol Biol Phys 1988 14 701 709 3350725
Seaward SA Weinberg V Lewis P Leigh B Phillips TL Roach M Improved freedom from PSA failure with whole pelvic irradiation for high-risk prostate cancer Int J Radiat Oncol Biol Phys 1998 42 1055 1062 9869229 10.1016/S0360-3016(98)00282-X
Bagshaw MA Cox RS Ray GR Status of radiation treatment of prostate cancer at Stanford University NCI Monogr 1988 47 60 3173503
McGowan DG The value of extended field radiation therapy in carcinoma of the prostate Int J Radiat Oncol Biol Phys 1981 7 1333 1339 7319858
Perez CA Michalski J Brown KC Lockett MA Nonrandomized evaluation of pelvic lymph node irradiation in localized carcinoma of the prostate Int J Radiat Oncol Biol Phys 1996 36 573 584 8948341 10.1016/S0360-3016(96)00378-1
Ploysongsang S Aron BS Shehata WM Jazy FK Scott RM Ho PY Morand TM Comparison of whole pelvis versus small-field radiation therapy for carcinoma of prostate Urology 1986 27 10 16 3079933 10.1016/0090-4295(86)90197-4
Rangala N Cox JD Byhardt RW Wilson JF Greenberg M Lopes da Conceicao A Local control and survival after external irradiation for adenocarcinoma of the prostate Int J Radiat Oncol Biol Phys 1982 8 1909 1914 6818193
Hauschild A Rosien F Lischner S Surgical standards in the primary care of melanoma patients Onkologie 2003 26 218 222 12845205 10.1159/000071616
Veronesi U Paganelli G Viale G Luini A Zurrida S Galimberti V Intra M Veronesi P Robertson C Maisonneuve P Renne G De Cicco C De Lucia F Gennari R A randomized comparison of sentinel-node biopsy with routine axillary dissection in breast cancer N Engl J Med 2003 349 546 553 12904519 10.1056/NEJMoa012782
Balkissoon J Rasgon BM Schweitzer L Lymphatic mapping for staging of head and neck cancer Semin Oncol 2004 31 382 393 15190496
Corvin S Eichhorn K Wurm T Sturm W Bares R Stenzl A Radioisotope guided laparoscopic pelvic lymph node dissection - a novel technique for prostate cancer staging Eur Urol 2004 3 Supp 140
Nutting CM Convery DJ Cosgrove VP Rowbottom C Padhani AR Webb S Dearnaley DP Reduction of small and large bowel irradiation using an optimized intensity-modulated pelvic radiotherapy technique in patients with prostate cancer Int J Radiat Oncol Biol Phys 2000 48 649 656 11020560 10.1016/S0360-3016(00)00653-2
Adams EJ Convery DJ Cosgrove VP McNair HA Staffurth JN Vaarkamp J Nutting CM Warrington AP Webb S Balyckyi J Dearnaley DP Clinical implementation of dynamic and step-and-shoot IMRT to treat prostate cancer with high risk of pelvic lymph node involvement Radiother Oncol 2004 70 1 10 15036846 10.1016/j.radonc.2003.09.004
Mundt AJ Lujan AE Rotmensch J Waggoner SE Yamada SD Fleming G Roeske JC Intensity-modulated whole pelvic radiotherapy in women with gynecologic malignancies Int J Radiat Oncol Biol Phys 2002 52 1330 1337 11955746 10.1016/S0360-3016(01)02785-7
|
16048656
|
PMC1190164
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Jul 28; 5:91
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-91
|
oa_comm
|
==== Front
BMC CancerBMC Cancer1471-2407BioMed Central London 1471-2407-5-941608079010.1186/1471-2407-5-94Case ReportTrauma-associated growth of suspected dormant micrometastasis El Saghir Nagi S [email protected] Ihab I [email protected] Fady B [email protected] Mukbil H [email protected] Department of Internal Medicine, American University of Beirut Medical Center, Beirut, Lebanon2 Department of Radiation Therapy, American University of Beirut Medical Center, Beirut, Lebanon3 Department of Diagnostic Radiology, American University of Beirut Medical Center, Beirut, Lebanon2005 4 8 2005 5 94 94 5 1 2005 4 8 2005 Copyright © 2005 El Saghir et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Cancer patients may harbor micrometastases that remain dormant, clinically undetectable during a variable period of time. A traumatic event or surgery may trigger the balance towards tumor growth as a result of associated angiogenesis, cytokine and growth factors release.
Case presentation
We describe a patient with non-small lung cancer who had a rapid tumor growth and recurrence at a minor trauma site of his skull bone.
Conclusion
This case is an illustration of the phenomenon of tumor growth after trauma or surgery and its associated cellular mechanisms. This phenomenon deserves further investigation and study.
==== Body
Background
Cancer patients may harbor micrometastases which are responsible for recurrent disease. Micrometastases remain dormant as a result of a balance between tumor cell proliferation and an equivalent rate of cell death [1,2]. Surgical interventions may trigger tumor growth an effect associated with angiogenesis, cytokines and growth factors release [3]. We report a patient with non-small lung cancer who had a rapid tumor growth and recurrence at a minor trauma site of his parietal skull bone. We suggest that the phenomenon of tumor growth after trauma or surgery deserves further investigation and study.
Case presentation
A 43 year-old smoker was diagnosed with a non small-cell lung cancer (NSCLC) in December 2001. He had a left upper lobe mass with adenopathy in the aorto-pulmonary window. CT-guided FNA of the lung lesion revealed a poorly differentiated NSCLC. Metastatic work-up showed only a suspicious 4 × 2 cm right supra-acetabular lesion of which a core biopsy was negative for malignancy. His total body bone scan showed no abnormal uptake in the skull and CT of brain was negative for metastatic disease.
He was given chemotherapy and radiotherapy and had a very good response. In April 2003, the patient reported that he suffered a minor trauma to the right lateral side of his head when he was in the passenger front seat of his car that rolled over a road bump. He reported a new small swelling that grew rapidly over a one month period. On physical examination, the mass was rubbery and not freely mobile measuring 7 cm in largest dimension (Figure 1). Plain skull x-ray films showed a 2 cm irregular lytic lesion of the right parietal bone (Figure 2). There were no underlying bone fractures. T1 weighted MRI of brain before (Figure 3A) and after Gadolinium contrast enhancement (Figure 3B), showed a metastatic deposit involving the right parietal bone with a large extracranial soft tissue component and meningeal invasion. T2 weighted image demonstrated the above described findings with meningeal thickening. There was no evidence of metastatic brain disease. Bone window of a non-enhanced CT of brain for radiation therapy planning purposes, showed a large lytic lesion with a soft tissue component of the right parietal bone. Chest X-ray showed no change in the residual ill defined left upper lobe density that the patient had in 2002. We concluded that the patient had an isolated tumor recurrence in his skull and proceeded with radiation therapy. The lesion showed decrease in size by the end of radiation. Fifteen days later, the family reported that the patient died at home of massive hemoptysis.
Discussion
We report the case of a patient with NSCLC who had an unusual recurrence in his skull after a direct minor trauma. The traumatic event was followed by the development of a small swelling that grew rapidly and manifested itself as an unusual metastatic lesion in the parietal skull bone as demonstrated in the accompanying figures. This event may be explained by the presence of dormant cancer cells might have been stimulated by a new environment of stimulatory factors. Several authors have investigated similar issues and incriminated growth factors, cytokines and angiogenic mechanisms [1]. It is now well known that cancer patients may harbor micrometastases and dormant cancer cells. Metastases remain dormant and clinically undetectable during a variable period of time when tumor cell proliferation is balanced by an equivalent rate of cell death.
Establishment and growth of metastases are thought to be influenced by endogenous inhibitors of angiogenesis which keep metastases in a non-proliferating quiescent state characterized by normal proliferation, increased apoptosis, and insufficient neovascularization [2,4]. Folkman suggested that, for tumors to grow and develop metastatic potential, they must make an "angiogenic switch" through perturbing the local balance of proangiogenic and antiangiogenic factors [5]. Lesions in the nervous system induce rapid activation of glial cells and under certain conditions additional recruitment of granulocytes, T-cells and monocytes/macrophages from the blood stream triggered by upregulation of cell adhesions molecules, chemokines, and cytokines [6].
Cytokines and chemokines may act to promote tumors by several mechanisms that include: DNA damage, bypass of p53, angiogenesis, growth stimulation, enhanced survival, subversion of immunity and enhanced invasion [7]. Some chemokines (eg, IL-8) are proangiogenic whereas others such as IP-10 have antiangiogenic activity. Chemokines have direct actions on microvacsular endothelial cells. In addition, CC chemokines may inhibit or stimulate angiogenesis indirectly, via their influence on tumor-associated macrophages. Inflammatory macrophages produce transforming growth factor β1 (TGF-β1) that is itself angiogenic and induces production of vascular endothelial growth factor (VEGF) [8]. This activation causes temporarily dormant micrometatases to vascularize, and thus to enter a rapid growth phase. In many tumors (eg, NSCLC and pancreatic carcinoma) it is postulated that the balance between proangiogenic and antiangiogenic cytokines and chemokines, rather than absolute amounts, regulates tumor angiogenesis [1].
A traumatic event triggers several mechanisms of soft tissue and bone repair of which angiogenesis is part. Dormant cancer cells at the site of tissue trauma and thereby exposed to pro-inflammatory mediators, may be sufficiently stimulated to overcome dormancy. Lee et al. [9] studied the effect of trauma on the implantation of metastatic tumor in bone in mice. The results suggest that the healing wound is a privileged site for experimental metastasis, particularly in the early stages. It is likely that the proteins in the blood clotting cascade are involved in local tumor implantation [10].
Another mechanism whereby trauma may alter subsequent tumor growth is that cytokine genes are highly polymorphic and since polymorphisms are frequently in regions of DNA that regulate transcription or post-transcriptional events, the cytokines may be cancer-modified genes [11].
Baum suggested that the act of surgery can provoke the outgrowth of dormant micrometatstasis [12,13]. Coffey et al. [3] reviewed mechanisms by which tumor excision may alter residual tumor growth and discussed the potential use of peri-operative chemotherapy, antiendotoxin agents, immunotherapy and biomodulation with use of dendritic-cell vaccines. Retsky et al. [14] hypothesized that induced angiogenesis after surgery in premenopausal node-positive breast cancer patients is a major underlying reason why adjuvant chemotherapy works particularly well for those patients. Our present case is an illustration of cellular mechanisms also associated with trauma that might either stimulate tumor growth of already present and dormant cells, or attract and trap circulating tumor cells.
Conclusion
Most traumatic and surgical events in cancer patients do not lead to tumor growth nor metastases, and swellings occurring after trauma do not require biopsy to prove their nature. However, we suggest that the phenomenon of tumor growth after trauma or surgery deserves further investigation and study. Antiangiogenic drugs might have a potential role for investigation in some circumstances of cancer patients in remission who require surgery, or are subject to a traumatic event.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
NSES, the corresponding author, was the main operator in charge of the case, the idea and the manuscript. IIEH assisted in the literature search and manuscript. FBG was involved in the patient's management and radiation therapy. MHH was involved in imaging and preparation of figures. All authors contributed to the preparation of the manuscript. All authors read and approved the final version of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Written consent for publication of the case was obtained from the patient's wife.
Figures and Tables
Figure 1 Photograph of the patient showing the described 7 cm mass at the site of the trauma at the right lateral side of his head.
Figure 2 Plain skull x-ray film showing an irregular lytic lesion of the right parietal lobe.
Figure 3 T1 weighted MRI of brain before (figure 3A) and after Gadolinium contrast enhancement (figure 3B>), showing a metastatic deposit involving the right frontal bone with a large extracranial soft tissue component and meningeal invasion.
==== Refs
O'Byrne KJ Dalgleish AG Browning MJ Steward WP Harris AL The relationship between angiogenesis and the immune response in carcinogenesis and the progression of malignant disease Eur J Cancer 2000 36 151 69 10741273 10.1016/S0959-8049(99)00241-5
Holmgren L O'Reilly MS Folkman J Dormancy of micrometastases: balanced proliferation and apoptosis in the presence of angiogenesis suppression Nat Med 1995 1 149 53 7585012 10.1038/nm0295-149
Coffey JC Wang JH Smith MJ Bouchier-Hayes D Cotter TG Redmond HP Excisional surgery for cancer cure: therapy at a cost Lancet Oncol 2003 4 760 8 14662433 10.1016/S1470-2045(03)01282-8
Kirsch M Schackert G Black PM Metastasis and angiogenesis Cancer Treat Res 2004 117 285 304 15015566
Folkman J Role of angiogenesis in tumor growth and metastasis Semin Oncol 2002 29 15 18 12516034 10.1053/sonc.2002.37263
Stoll G Jander S Schroeter M Detrimental and beneficial effects of injury-induced inflammation and cytokine expression in the nervous system Adv Exp Med Biol 2002 513 87 113 12575818
Balkwill F Mantovani A Inflammation and cancer: back to Virchow? Lancet 2001 357 539 45 11229684 10.1016/S0140-6736(00)04046-0
Opdenakker G Van Damme J Chemotactic factors, passive invasion and metastasis of cancer cells Immunol Today 1992 13 463 4 1298279 10.1016/0167-5699(92)90079-M
Lee JY Murphy SM Scanlon EF Effect of trauma on implantation of metastatic tumor in bone in mice J Surg Oncol 1994 56 178 84 8028350
Rosenberg GA Matrix metalloproteinases in brain injury J Neurotrauma 1995 12 833 42 8594211
Jaiswal M LaRusso NF Burgart LJ Gores GJ Inflammatory cytokines induce DNA damage and inhibit DNA repair in cholangiocarcinoma cells by a nitric oxide-dependent mechanism Cancer Res 2000 60 184 190 10646872
Baum M Does surgery disseminate or accelerate cancer? Lancet 1996 347 260 8551898 10.1016/S0140-6736(96)90433-X
Baum M Does the act of surgery provoke activation of "latent" metastases in early breast cancer? Breast Cancer Res 2004 6 160 1 15217487 10.1186/bcr902
Retsky M Bonadonna G Demicheli R Folkman J Hrushesky W Valagussa P Hypothesis: Induced angiogenesis after surgery in premenopausal node-positive breast cancer patients is a major underlying reason why adjuvant chemotherapy works particularly well for those patients Breast Cancer Res 2004 6 R372 4 15217504 10.1186/bcr804
|
16080790
|
PMC1190165
|
CC BY
|
2021-01-04 16:03:06
|
no
|
BMC Cancer. 2005 Aug 4; 5:94
|
utf-8
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-94
|
oa_comm
|
==== Front
BMC Cardiovasc DisordBMC Cardiovascular Disorders1471-2261BioMed Central London 1471-2261-5-221601882210.1186/1471-2261-5-22Research ArticleSirolimus increases tissue factor expression but not activity in cultured human vascular smooth muscle cells Zhu Shengsi [email protected] Hema [email protected] Thusitha [email protected] Xiu-Fen [email protected] Zhihong [email protected] Vascular Biology, Department of Medicine, Division of Physiology,University of Fribourg, Rue du Musée 5, CH-1700 Fribourg, Switzerland2 Cardiovascular Department, 1st Affiliated Hospital of Dalian Medical University, Dalian, P.R. China2005 15 7 2005 5 22 22 7 1 2005 15 7 2005 Copyright © 2005 Zhu et al; licensee BioMed Central Ltd.2005Zhu et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Sirolimus-eluting stents (CYPHER stents) demonstrated remarkable efficacy in reducing restenosis rates in patients with coronary artery disease. There is a concern of sub-acute and late stent thrombosis. Tissue factor (TF) is critical in thrombosis. This study investigated the effect of sirolimus on TF expression and activity in cultured human vascular smooth muscle cells (SMCs).
Methods
SMCs were cultured from human saphenous veins and aortas. Quiescent cells were stimulated with sirolimus (0.1 – 20 ng/ml) over 24 hours. Cellular TF expression and activity released into culture medium were measured. The effect of sirolimus on activation of mammalian target of rapamycin (mTOR) was measured by phosphorylation of the substrate p70s6k at T389, and activation of RhoA was measured by pull-down assay.
Results
Sirolimus increased TF protein level in cultured human SMCs in a concentration and time-dependent manner (about 2-fold, p < 0.01) reaching maximal effect at 5 ng/ml. The stimulation of TF expression by sirolimus was associated with inhibition of basal activity of mTOR. No effects of sirolimus on RhoA or p38mapk activation that are positive regulators of TF in vascular wall cells were observed. The stimulation of TF expression by sirolimus (20 ng/ml) was prevented by the HMG-CoA reductase inhibitor fluvastatin (1 μmol/L). However, no increase in TF activity released from SMC into culture medium was observed after sirolimus treatment.
Conclusion
Although sirolimus stimulates TF protein expression in human SMC associated with inhibition of mTOR, it does not enhance TF activity released from the cells, suggesting a relatively safe profile of CYPHER stents. The inhibition of TF expression by fluvastatin favors clinical use of statins in patients undergoing coronary stenting.
==== Body
Background
Since the first human study with sirolimus (rapamycin)-eluting stents (Cordis CYPHER™ stent) by Sousa [1], considerable promise of sirolimus-eluting stents for reducing restenosis rates and clinical parameters was subsequently demonstrated by several randomized clinical trials [2-7]. The mechanism of inhibition of restenosis by sirolimus has been suggested to be attributed to the blockade of smooth muscle cell (SMC) cycle progression from G1 to S phase via inhibition of the protein kinase, mammalian target of rapamycin (mTOR)[8].
Despite the promising results on restenosis rates, there is concern that drug-eluting stents may be associated with increased thrombosis rates. Although stent thrombosis associated with sirolimus-eluting stents has been reported in several clinical trials, it remains a rare event and is not higher in patients receiving bare metal stents [2-4,9,10]. Pooled analysis of clinical trials does not reveal a higher incidence of stent thrombosis, suggesting a relative safe profile of drug-eluting stents at least under the condition of anti-platelet regiment [11]. However, individual case reports generated some suspicion that drug-eluting stents may be prone to thrombosis [12]. In a report, four cases of late coronary thrombosis related to drug-eluting stents were presented, all of them occurred shortly after anti-platelet therapy was interrupted [12], and in two patients who received both a bare metal stent and a sirolimus-eluting stent, only the sirolimus-eluting stents were closed due to thrombosis, while the bare metal stents remained open in the same patients [12]. Based on the controversial reports and concerns, we analyzed whether sirolimus per se exerts some adverse effects related to thrombosis in vascular cells namely smooth muscle cells.
Tissue factor (TF) plays an important role in thrombosis and acute coronary syndromes [13]. It is the principle initiator of extrinsic coagulation pathway activating thrombin and generating fibrin leading to thrombus formation. Recent study suggests that aberrant TF expression in the vascular wall cells plays a crucial role in triggering intravascular thrombosis [14]. Under non-stimulated conditions, vascular wall cells i.e. endothelial cells and SMCs express negligible or low level of TF that can be up-regulated by cytokines and thrombin [15-17]. Several intracellular signal transduction mechanisms have been demonstrated to be involved in the regulation of TF expression. The small G-protein RhoA and the protein kinase p38mapk are positive regulators, whereas phosphatidylinositol 3-kinase (PI3-K) negatively regulates TF expression in vascular wall cells [17].
The HMG CoA reductase inhibitors or statins reduce cardiovascular events in patients with coronary heart disease [18]. The non-cholesterol lowering effects i.e. pleiotropic effects of statins seem to play important roles [19]. Experimental studies demonstrate that statins increase eNOS expression in endothelial cells, inhibit TF expression in SMC via inhibition of Rho/ROCK pathway [16,20]. Hence, the present study is aimed to investigate whether sirolimus could promote TF expression in human SMC, and whether this is associated with an increased TF activity. The effects of statin such as fluvastatin on TF expression in SMC were also investigated.
Methods
Materials
Sirolimus was purchased from Calbiochem (Lucerne, Switzerland); fluvastatin was kindly provided by Novartis (Basel, Switzerland); tumor necrosis factor-α (TNF-α) was purchased from R & D, France); monoclonal mouse anti-TF antibody and tissue factor activity kit were purchased from American Diagnostica Inc (Socochim, Lausanne, Switzerland); anti-tubulin and all the other chemicals for immunoblotting were purchased from Sigma (Buchs, Switzerland); anti-phospho p70s6k (T389) was from Cell Signaling Technology. Alkaline phosphatase (AP)-conjugated anti-mouse IgG and BCIP/NBT substrate for AP were from Interchim (Chemie Brunschwig AG, Basel, Switzerland).
SMC and endothelial cell culture
SMC were isolated and cultured from human saphenous veins [20] and human aortic SMC were kindly provided by Dr. Therese Resink (University of Basel, Switzerland). Endothelial cells from human umbilical veins were isolated as previously described [17].
TF expression
Cells were rendered quiescent for 24 hours in DMEM containing 0.2% BSA before they were treated with sirolimus (20 ng/ml, 24 hours), a concentration which fully inhibits mTOR/p70s6k pathway as previously shown [21]. To study the effect of fluvastatin on sirolimus-induced TF expression, the cells were pre-incubated with fluvastatin (1 μmol/L) for 60 minutes. Cell lysates were prepared as described [17]. 30 μg extracts were used for immunoblotting of TF expression [17]. Tubulin expression was used to ensure equal protein loading. Quantification was performed using NIH Image-J software. TF expression was expressed as percentage changes of the basal level.
TF activity in cell conditioned medium
2 × 10-5 cells/ml were seeded onto each dish for overnight attachment. Cells were then rendered quiescent in phenol-red free DMEM medium containing 0.2% BSA for 24 hours and then treated with sirolimus (20 ng/ml; 24 hours) as described above except that conditioned medium was collected and TF activity was measured as instructed by the manufacturer. Briefly, the same amount of conditioned medium (25 μl) was incubated in the presence of Factor VIIa and Factor X for 15 minutes in a 96-well plate, after which the substrate was added and further incubated for another 60 minutes before the reaction was stopped with glacial acetic acid and color reaction was measured with a microplate reader at 405 nm. TF activity is expressed in picomolar obtained from the standard curve.
RhoA activation
The activation of RhoA was assessed by pull-down assay in the cells stimulated with sirolimus (20 ng/ml) over one hour as described [17]. Briefly, SMCs were washed with ice-cold Tris-buffered saline and lysed in RIPA buffer (50 mmol/L Tris, pH 7.2, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 500 mmol/L NaCl, 10 mmol/L MgCl2, 10 μg/ml each of leupeptin and aprotinin, and 1 mmol/L PMSF). 200 μg of cell lysates were incubated with 10 μg of GST-TRBD beads at 4°C for 60 min. The beads were washed four times with buffer B (Tris-buffer containing 1% Triton X-100, 150 mmol/L NaCl, 10 mmol/L MgCl2, 10 μg/ml each of leupeptin and aprotinin, and 0.1 mmol/L PMSF). Bound RhoA proteins were then detected by immunoblotting using a monoclonal antibody against RhoA (Santa Cruz Biotechnology). The total amount of RhoA in cell lysates was used as a control for the cross-comparison of RhoA activity (level of GTP-bound RhoA).
mTOR activation
mTOR activation was examined by immunoblotting measuring p70s6k phosphorylation at T389 in quiescent cells with or without sirolimus treatment (20 ng/ml, 1 hour). 40 μg cell extracts were subjected to 8% SDS-PAGE and phosphorylated p70s6k was detected using anti-phospho-p70s6k (T389) antibody. Activation of p70s6k was calculated as ratio of phospho-p70s6k against tubulin.
Statistic analysis
All data were expressed as mean ± SEM and one way analysis of variance (ANOVA) with Bonferroni's post-test was used for statistical analysis. A two-tailed value of p ≤ 0.05 was considered statistically significant.
Results and discussion
Sirolimus-eluting stents have demonstrated remarkable clinical efficacy in reducing restenosis rates in the short-to-medium term [1-6]. There is a concern of subacute and late stent thrombosis [22-24] although individual clinical trials and pooled analysis of all the randomized trials showed no evidence of increase in stent thrombosis with drug-eluting stents as compared to the bare-metal stents in the short-to-medium term [2-4,9,10,25]. Some clinical experiences suggest that sirolimus-eluting stents may be prone to thrombosis [12,22,23,26,27]. In an earlier report [22], a case of late stent thrombosis associated with sirolimus-eluting stent was noticed 2 weeks after the patient stopped anti-platelet therapy. More recently, a clinical report presented four cases of late coronary thrombosis related to drug-eluting stents that occurred several months after coronary intervention [12]. It raises much concern by the observation that in two patients who received both a bare metal stent and a sirolimus-eluting stent, only the drug-eluting stents were closed due to late stent thrombosis which developed shortly after anti-platelet therapy was interrupted, whereas the bare metal stents in the same patients remained open [12]. This observation may indicate that sirolimus could exert pro-thrombotic effects, in particular, when anti-platelet therapy was discontinued. Adverse effects of sirolimus related to thrombosis have been documented in vitro experiments and also in vivo in an animal model [28-30]. Sirolimus has been reported to inhibit endothelium-dependent relaxations in porcine coronary arteries [28]. Inhibition of endothelialization by sirolimus has also been shown in human necropsy specimens and in animal models [31,32]. Moreover, stimulation or facilitation of platelet aggregation and secretion by sirolimus has also been demonstrated [29]. The pro-thrombotic effect of sirolimus was also demonstrated in a rat model of synthetic vascular grafts [30]. In our present study, we demonstrated that in human SMC, sirolimus increased TF protein level at a low concentration i.e. 0.1 ng/ml, which reached the maximal effect at 5 ng/ml (Fig. 1A). This concentration is in the range of clinical settings, since systemic level of sirolimus was reported to be in the range of 1~2 ng/ml within the first hours after CYPHER stent placement [33]. The concentration in the vascular wall is expected to be higher. The stimulation of TF expression by sirolimus (20 ng/ml) is also time-dependent (24 hours, 270% increase above control, Fig. 1B, n = 6, p < 0.01). Our result is in line with a recent observation by Guba et al., showing that sirolimus stimulates TF expression in human endothelial cells [34].
Figure 1 Sirolimus up-regulates tissue factor (TF) expression in SMC. (A). Sirolimus enhances TF expression in a concentration- and (B) time-dependent manner in human SMC. n = 6, * = p < 0.05 vs. control, ** = p < 0.01 vs. control.
Furthermore, we showed that the induction of TF by sirolimus (20 ng/ml, 24 hours) was fully inhibited by the HMG-CoA reductase inhibitor fluvastatin (1 μmol/L) in human saphenous vein (Fig. 2A) and aorta SMCs (Fig. 2B) (n = 4, p < 0.05). Fluvastatin alone, however, did not significantly affect the basal TF expression in the cells (Fig. 2A and Fig. 2B). Statins exert many effects on vascular cells via inhibition of RhoA [20,35]. It is, however, unlikely that this mechanism explains the inhibitory effect of TF expression by fluvastatin. Firstly, a significant basal activity of RhoA was present in SMC (Fig. 2C), which was, however, not further stimulated by sirolimus (20 ng/ml) over 60 minutes (Fig. 2C). Secondly, fluvastatin alone did not inhibit basal TF expression (Fig. 2A and 2B), suggesting that basal TF expression is not mediated by RhoA. Results of our previous studies and others demonstrated that besides RhoA, p38mapk is also a positive regulator of TF expression in vascular endothelial cells and SMCs [15-17]. Our present study showed that no basal activity of p38mapk could be detected, and sirolimus (20 ng/ml, over 60 minutes) did not activate p38mapk in SMC. These results suggest that sirolimus enhances TF expression not through p38mapk and RhoA. Although the exact mechanism of statin-induced inhibition of TF expression by sirolimus is still obscure under this condition, our results support the clinical benefit of statins in patients with coronary stenting.
Figure 2 Fluvastatin inhibits TF expression in SMC. In human saphenous vein SMC (HSVSMC, panel A, n = 4) as well as in human aortic SMC (HAoSMC, panel B, n = 4) sirolimus (20 ng/ml, 24 hours) up-regulated TF expression which was significantly inhibited by fluvastatin (1 μmol/L). Basal activity of RhoA was not influenced by sirolimus (20 ng/ml, n = 3, panel C), while the basal activity of mTOR was fully inhibited by sirolimus (20 ng/ml, 1 hour, n = 3, panel D). * = p < 0.01 vs. control, † = p < 0.05 vs. sirolimus.
It is well described that sirolimus is a natural immunosuppressant which interferes with cellular functions via blockade of the protein kinase, mTOR [8] which further activates its downstream effector p70s6k by phosphorylating T389 residue [36]. In our present study, we showed a significant basal activity of mTOR in SMC as measured by p70s6k phosphorylation at T389 (Fig. 2D). The activity of mTOR was abolished by sirolimus (20 ng/ml, 1 hour treatment, Fig. 2D). Whether this data suggest an inhibitory effect of mTOR on TF expression needs further investigation. Further results demonstrate that fluvastatin does not reverse the inhibition of mTOR i.e. phosphorylation of p70s6k at T389 by sirolimus, nor it had any effect on basal mTOR activity in the cells (Fig. 2D), suggesting that statin inhibits TF expression not through regulation of mTOR.
Despite increased TF protein expression by sirolimus in SMC, the activity of TF released from SMC into the culture medium was not enhanced by sirolimus (Fig. 3). For the validity of the method, a control experiment using endothelial cells was performed. Endothelial cells had much lower TF activity than SMC (p < 0.0001; n = 6–9), that was significantly enhanced by TNF-α (20 ng/ml; 5 hours, p < 0.05, n = 9, Fig. 3)
Figure 3 Sirolimus does not stimulate TF release and activity from SMC. Human SMC released much higher TF activity into the culture medium than endothelial cells (HUVECs) (p < 0.0001, n = 6–9). TF activity released from SMC was not further increased by sirolimus (20 ng/ml, 24 hours). The TF activity released from HUVECs was significantly stimulated by TNF-α (20 ng/ml; 5 hours, p < 0.05, n = 9).
Conclusion
Taken together, our results demonstrate that although sirolimus stimulates TF expression in human SMC, it does not enhance TF activity released from the cells. The results support the safe profile of CYPHER stents observed by clinical trials. The inhibition of TF expression by fluvastatin favors clinical use of statins in patients undergoing coronary stenting.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SZ and HV performed Western blot analysis for TF expression and activity and activation of mTOR. TG assisted us in human SMC culture and statistical analysis. XFM performed RhoA pull-down assay. ZY conceived and coordinated the study and drafted the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This study was supported by Swiss National Science Foundation (3100A0-105917/1) and Swiss University Conference (SUK) program. X-F. Ming was supported by Swiss Heart Foundation and Roche Research Foundation. S. Zhu and T. Gajanayake are recipients of Swiss Federal Scholarship.
==== Refs
Sousa JE Costa MA Abizaid AC Rensing BJ Abizaid AS Tanajura LF Kozuma K Van LG Sousa AG Falotico R Jaeger J Popma JJ Serruys PW Sustained suppression of neointimal proliferation by sirolimus-eluting stents: one-year angiographic and intravascular ultrasound follow-up Circulation 2001 104 2007 2011 11673337
Morice MC Serruys PW Sousa JE Fajadet J Ban HE Perin M Colombo A Schuler G Barragan P Guagliumi G Molnar F Falotico R A randomized comparison of a sirolimus-eluting stent with a standard stent for coronary revascularization N Engl J Med 2002 346 1773 1780 12050336 10.1056/NEJMoa012843
Moses JW Leon MB Popma JJ Fitzgerald PJ Holmes DR O'Shaughnessy C Caputo RP Kereiakes DJ Williams DO Teirstein PS Jaeger JL Kuntz RE Sirolimus-eluting stents versus standard stents in patients with stenosis in a native coronary artery N Engl J Med 2003 349 1315 1323 14523139 10.1056/NEJMoa035071
Schofer J Schluter M Gershlick AH Wijns W Garcia E Schampaert E Breithardt G Sirolimus-eluting stents for treatment of patients with long atherosclerotic lesions in small coronary arteries: double-blind, randomised controlled trial (E-SIRIUS) Lancet 2003 362 1093 1099 14550694 10.1016/S0140-6736(03)14462-5
Lemos PA Serruys PW van Domburg RT Saia F Arampatzis CA Hoye A Degertekin M Tanabe K Daemen J Liu TK McFadden E Sianos G Hofma SH Smits PC van der Giessen WJ de Feyter PJ Unrestricted utilization of sirolimus-eluting stents compared with conventional bare stent implantation in the "real world": the Rapamycin-Eluting Stent Evaluated At Rotterdam Cardiology Hospital (RESEARCH) registry Circulation 2004 109 190 195 14691037 10.1161/01.CIR.0000109138.84579.FA
Popma JJ Leon MB Moses JW Holmes DR JrCox N Fitzpatrick M Douglas J Lambert C Mooney M Yakubov S Kuntz RE Quantitative assessment of angiographic restenosis after sirolimus-eluting stent implantation in native coronary arteries Circulation 2004 110 3773 3780 15596568 10.1161/01.CIR.0000150331.14687.4B
Fajadet J Morice MC Bode C Barragan P Serruys PW Wijns W Constantini CR Guermonprez JL Eltchaninoff H Blanchard D Bartorelli A Laarman GJ Perin M Sousa JE Schuler G Molnar F Guagliumi G Colombo A Ban Hayashi E Wulfert E Maintenance of long-term clinical benefit with sirolimus-eluting coronary stents: three-year results of the RAVEL trial Circulation 2005 111 1040 1044 15723977 10.1161/01.CIR.0000156334.24955.B2
Poon M Badimon JJ Fuster V Overcoming restenosis with sirolimus: from alphabet soup to clinical reality Lancet 2002 359 619 622 11867133 10.1016/S0140-6736(02)07751-6
Babapulle MN Joseph L Belisle P Brophy JM Eisenberg MJ A hierarchical Bayesian meta-analysis of randomised clinical trials of drug-eluting stents Lancet 2004 364 583 591 15313358 10.1016/S0140-6736(04)16850-5
Ong AT Hoye A Aoki J van Mieghem CA Rodriguez Granillo GA Sonnenschein K Regar E McFadden EP Sianos G van der Giessen WJ de Jaegere PP de Feyter P van Domburg RT Serruys PW Thirty-day incidence and six-month clinical outcome of thrombotic stent occlusion after bare-metal, sirolimus, or paclitaxel stent implantation J Am Coll Cardiol 2005 45 947 953 15766834 10.1016/j.jacc.2004.09.079
Moreno R Fernandez C Hernandez R Alfonso F Angiolillo DJ Sabate M Escaned J Banuelos C Fernandez-Ortiz A Macaya C Drug-eluting stent thrombosis: results from a pooled analysis including 10 randomized studies J Am Coll Cardiol 2005 45 954 959 15766835 10.1016/j.jacc.2004.11.065
McFadden EP Stabile E Regar E Cheneau E Ong AT Kinnaird T Suddath WO Weissman NJ Torguson R Kent KM Pichard AD Satler LF Waksman R Serruys PW Late thrombosis in drug-eluting coronary stents after discontinuation of antiplatelet therapy Lancet 2004 364 1519 1521 15500897 10.1016/S0140-6736(04)17275-9
Moons AH Levi M Peters RJ Tissue factor and coronary artery disease Cardiovasc Res 2002 53 313 325 11827681 10.1016/S0008-6363(01)00452-7
Day SM Reeve JL Pedersen B Farris DM Myers DD Im M Wakefield TW Mackman N Fay WP Macrovascular thrombosis is driven by tissue factor derived primarily from the blood vessel wall Blood 2005 105 192 198 15339841 10.1182/blood-2004-06-2225
Archipoff G Beretz A Freyssinet JM Klein-Soyer C Brisson C Cazenave JP Heterogeneous regulation of constitutive thrombomodulin or inducible tissue-factor activities on the surface of human saphenous-vein endothelial cells in culture following stimulation by interleukin-1, tumour necrosis factor, thrombin or phorbol ester Biochem J 1991 273 679 684 1847620
Eto M Kozai T Cosentino F Joch H Luscher TF Statin prevents tissue factor expression in human endothelial cells: role of Rho/Rho-kinase and Akt pathways Circulation 2002 105 1756 1759 11956113 10.1161/01.CIR.0000015465.73933.3B
Viswambharan H Ming XF Zhu S Hubsch A Lerch P Vergeres G Rusconi S Yang Z Reconstituted high-density lipoprotein inhibits thrombin-induced endothelial tissue factor expression through inhibition of RhoA and stimulation of phosphatidylinositol 3-kinase but not Akt/endothelial nitric oxide synthase Circ Res 2004 94 918 925 14988229 10.1161/01.RES.0000124302.20396.B7
Grundy SM Cleeman JI Merz CN Brewer HB JrClark LT Hunninghake DB Pasternak RC Smith SC JrStone NJ Coordinating Committee of the National Cholesterol Education Program Implications of recent clinical trials for the National Cholesterol Education Program Adult Treatment Panel III Guidelines J Am Coll Cardiol 2004 44 720 732 15358046 10.1016/j.jacc.2004.07.001
Farmer JA Pleiotropic effects of statins Curr Atheroscler Rep 2000 2 208 217 11122746
Yang Z Kozai T van der LB Viswambharan H Lachat M Turina MI Malinski T Luscher TF HMG-CoA reductase inhibition improves endothelial cell function and inhibits smooth muscle cell proliferation in human saphenous veins J Am Coll Cardiol 2000 36 1691 1697 11079678 10.1016/S0735-1097(00)00924-4
Ming XF Burgering BM Wennstrom S Claesson-Welsh L Heldin CH Bos JL Kozma SC Thomas G Activation of p70/p85 S6 kinase by a pathway independent of p21ras Nature 1994 371 426 429 8090223 10.1038/371426a0
Kerner A Gruberg L Kapeliovich M Grenadier E Late stent thrombosis after implantation of a sirolimus-eluting stent Catheter Cardiovasc Interv 2003 60 505 508 14624431 10.1002/ccd.10712
Virmani R Guagliumi G Farb A Musumeci G Grieco N Motta T Mihalcsik L Tespili M Valsecchi O Kolodgie FD Localized hypersensitivity and late coronary thrombosis secondary to a sirolimus-eluting stent: should we be cautious? Circulation 2004 109 701 705 14744976 10.1161/01.CIR.0000116202.41966.D4
Eisenberg MJ Drug-eluting stents: some bare facts Lancet 2004 364 1466 1467 15500872 10.1016/S0140-6736(04)17287-5
Lemos PA Lee CH Degertekin M Saia F Tanabe K Arampatzis CA Hoye A van DM Sianos G Smits PC de FP van der Giessen WJ van Domburg RT Serruys PW Early outcome after sirolimus-eluting stent implantation in patients with acute coronary syndromes: insights from the Rapamycin-Eluting Stent Evaluated At Rotterdam Cardiology Hospital (RESEARCH) registry J Am Coll Cardiol 2003 41 2093 2099 12798587 10.1016/S0735-1097(03)00429-7
Colombo A Moses JW Morice MC Ludwig J Holmes DR JrSpanos V Louvard Y Desmedt B Di MC Leon MB Randomized study to evaluate sirolimus-eluting stents implanted at coronary bifurcation lesions Circulation 2004 109 1244 1249 14981005 10.1161/01.CIR.0000118474.71662.E3
Murphy BE Drug-eluting stents: some first-generation problems Am Heart Hosp J 2004 2 80 84 15604848
Jeanmart H Malo O Carrier M Nickner C Desjardins N Perrault LP Comparative study of cyclosporine and tacrolimus vs newer immunosuppressants mycophenolate mofetil and rapamycin on coronary endothelial function J Heart Lung Transplant 2002 21 990 998 12231370 10.1016/S1053-2498(02)00429-1
Babinska A Markell MS Salifu MO Akoad M Ehrlich YH Kornecki E Enhancement of human platelet aggregation and secretion induced by rapamycin Nephrol Dial Transplant 1998 13 3153 3159 9870481 10.1093/ndt/13.12.3153
Walpoth BH Pavlicek M Celik B Nicolaus B Schaffner T Althaus U Hess OM Carrel T Morris RE Prevention of neointimal proliferation by immunosuppression in synthetic vascular grafts Eur J Cardiothorac Surg 2001 19 487 492 11306318 10.1016/S1010-7940(01)00582-6
Guagliumi G Farb A Musumeci G Valsecchi O Tespili M Motta T Virmani R Images in cardiovascular medicine. Sirolimus-eluting stent implanted in human coronary artery for 16 months: pathological findings Circulation 2003 107 1340 1341 12628958 10.1161/01.CIR.0000062700.42060.6F
Fukuda D Sata M Tanaka K Nagai R Potent inhibitory effect of sirolimus on circulating vascular progenitor cells Circulation 2005 111 926 931 15710768 10.1161/01.CIR.0000155612.47040.17
Iakovou I Sangiorgi GM Stankovic G Corvaja N Vitrella G Ferraro M Colombo A Results and follow-up after implantation of four or more sirolimus-eluting stents in the same patient Catheter Cardiovasc Interv 2005 64 436 439 15789401 10.1002/ccd.20305
Guba M Yezhelyev M Eichhorn ME Schmid G Ischenko I Papyan A Graeb C Seeliger H Geissler EK Jauch KW Bruns CJ Rapamycin induces tumor-specific thrombosis via tissue factor in the presence of VEGF Blood 2005 105 4463 4469 15671443 10.1182/blood-2004-09-3540
Wolfrum S Jensen KS Liao JK Endothelium-dependent effects of statins Arterioscler Thromb Vasc Biol 2003 23 729 736 12615672 10.1161/01.ATV.0000063385.12476.A7
Hay N Sonenberg N Upstream and downstream of mTOR Genes Dev 2004 18 1926 1945 15314020 10.1101/gad.1212704
|
16018822
|
PMC1190166
|
CC BY
|
2021-01-04 16:30:06
|
no
|
BMC Cardiovasc Disord. 2005 Jul 15; 5:22
|
utf-8
|
BMC Cardiovasc Disord
| 2,005 |
10.1186/1471-2261-5-22
|
oa_comm
|
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-341610722210.1186/1471-2296-6-34Study ProtocolProcalcitonin-guided antibiotic use versus a standard approach for acute respiratory tract infections in primary care: study protocol for a randomised controlled trial and baseline characteristics of participating general practitioners [ISRCTN73182671] Briel Matthias [email protected] Mirjam [email protected] Jim [email protected] Philipp [email protected] Peter [email protected]ériat Pierre [email protected] Heiner C [email protected]üller Beat [email protected] Basel Institute for Clinical Epidemiology, University Hospital Basel, CH-4031 Basel, Switzerland2 Clinic of Endocrinology, Diabetes & Clinical Nutrition, Department of Internal Medicine, University Hospital Basel, CH-4031 Basel, Switzerland3 Department of Chemical Pathology, University Hospital Basel, CH-4031 Basel, Switzerland4 General practice, In den Neumatten 63, CH-4125 Riehen, Switzerland2005 18 8 2005 6 34 34 11 5 2005 18 8 2005 Copyright © 2005 Briel et al; licensee BioMed Central Ltd.2005Briel et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Acute respiratory tract infections (ARTI) are among the most frequent reasons for consultations in primary care. Although predominantly viral in origin, ARTI often lead to the prescription of antibiotics for ambulatory patients, mainly because it is difficult to distinguish between viral and bacterial infections. Unnecessary antibiotic use, however, is associated with increased drug expenditure, side effects and antibiotic resistance. A novel approach is to guide antibiotic therapy by procalcitonin (ProCT), since serum levels of ProCT are elevated in bacterial infections but remain lower in viral infections and inflammatory diseases.
The aim of this trial is to compare a ProCT-guided antibiotic therapy with a standard approach based on evidence-based guidelines for patients with ARTI in primary care.
Methods/Design
This is a randomised controlled trial in primary care with an open intervention. Adult patients judged by their general practitioner (GP) to need antibiotics for ARTI are randomised in equal numbers either to standard antibiotic therapy or to ProCT-guided antibiotic therapy. Patients are followed-up after 1 week by their GP and after 2 and 4 weeks by phone interviews carried out by medical students blinded to the goal of the trial.
Exclusion criteria for patients are antibiotic use in the previous 28 days, psychiatric disorders or inability to give written informed consent, not being fluent in German, severe immunosuppression, intravenous drug use, cystic fibrosis, active tuberculosis, or need for immediate hospitalisation.
The primary endpoint is days with restrictions from ARTI within 14 days after randomisation. Secondary outcomes are antibiotic use in terms of antibiotic prescription rate and duration of antibiotic treatment in days, days off work and days with side-effects from medication within 14 days, and relapse rate from the infection within 28 days after randomisation.
Discussion
We aim to include 600 patients from 50 general practices in the Northwest of Switzerland. Data from the registry of the Swiss Medical Association suggests that our recruited GPs are representative of all eligible GPs with respect to age, proportion of female physicians, specialisation, years of postgraduate training and years in private practice.
==== Body
Background
Acute respiratory tract infections (ARTI) are among the most frequent reasons for seeking ambulatory care [1]. ARTI in the context of this study include common cold, pharyngitis, tonsillitis, rhinosinusitis, tracheo-bronchitis, otitis media, acute exacerbations of asthma and of chronic obstructive pulmonary disease (COPD), and community acquired pneumonia. As much as 75% of antibiotics are prescribed for ARTI, despite the mainly viral origin [2-8].
Criteria often used in clinical practice to distinguish bacterial from viral infections of the respiratory tract include fever, dyspnea, purulent sputum, chest X-ray infiltrates, C-reactive protein, leucocyte count, and recovery of a pathogen from the respiratory tract or from blood cultures [9]. However, these are all non-specific symptoms and hence differentiation between viral and bacterial ARTI remains a diagnostic challenge [10]. Moreover, when antibiotic treatment is initiated, the optimal duration of antibiotic treatment for ARTI has not been determined [11,12]. In community acquired pneumonia an antibiotic treatment duration of 10 to 14 days is generally recommended, although data from intervention trials are lacking [13].
Unnecessary antibiotic use (i.e. number of prescriptions and duration of treatment) for ARTI not only increases drug expenditure [14] and the risk of adverse events [15], but also results in selection of resistant microorganisms [16]. Thereby, it constitutes an important public health problem [17]. For combating the increase in resistant infections a decrease of the excess antibiotic use is paramount [18]. There are only few intervention studies that have reported a successful reduction of antibiotic use in ambulatory care [19-25]. Most of these studies were not conducted in ARTI or have methodological limitations.
A novel approach to guide antimicrobial therapy is to prescribe antibiotics based on the level of biomarkers, specifically, calcitonin precursors, including procalcitonin (ProCT). Circulating levels of ProCT are elevated in systemic bacterial infections but remain relatively low in viral infections and inflammatory diseases [26,27]. In severe bacterial infections the use of ProCT significantly improves the sensitivity and specificity of the clinical diagnosis of infection [28]. A recent systematic review and meta-analysis found that ProCT is superior compared to C-reactive protein for the diagnosis of bacterial infections [29]. Most recently, we gathered evidence that both antibiotic prescription and treatment duration could be safely and markedly reduced in hospitalised patients with lower respiratory tract infections using ProCT-stewardship [30,31]. In successfully treated infections, circulating ProCT levels decrease in a log-linear pattern and have a plasma half life of 24 hours. In contrast, prolonged elevated plasma ProCT levels indicate adverse outcome [26,27].
Several studies indicate that the main reasons for antibiotic prescription in ambulatory patients with ARTI are non-medical and related to the physician-patient relationship, patients' expectations and beliefs about the benefit of antibiotics [32,33]. Thus, in theory a reduction of antibiotic prescriptions and duration can also be achieved by the implementation of guidelines [34]. However, in practice physician education and guidelines dissemination for ARTI management usually show no clinically relevant effect [25,35,36].
Methods/Design
Study aims
The objective of this trial is to evaluate, if a ProCT-guided diagnostic and therapeutic strategy leads to a similar outcome and reduced total antibiotic use for patients with ARTI in primary care compared to a standard approach recommended by current guidelines.
Study design and setting
This is a prospective, randomised, controlled, open intervention trial in primary care with appropriate power calculation. Adult patients suffering from ARTI, for whom the treating general practitioner (GP) decides to give antibiotic treatment on the basis of evidence-based guidelines, are randomised to routine antibiotic therapy or ProCT-guided antibiotic treatment. The pathway by which patients are recruited and followed-up is given in Figure 1.
Figure 1 Summary of the trial design.
Ethical considerations
The Ethics Committee of the University Hospital Basel, Switzerland, approved the study protocol which is in compliance with the Helsinki Declaration. Written informed consent was obtained from all participating GPs. All recruited patients have to give written informed consent.
The trial is supervised by an independent monitoring board that is not involved in the design and conduct of the trial, or in the recruitment of patients. The board consists of a general internist in primary care, an infectious disease specialist and a pneumologist.
Participants
We invited all GPs of two cantons (Basel-Stadt and Basel-Land) in the Northwest of Switzerland to participate in the trial. Of 345 GPs contacted, 53 working at 50 practices gave written informed consent and were included (Figure 2).
Figure 2 Flow diagram of recruited general practitioners.
From December 2004 GPs included in this study consecutively screen all eligible adults (aged 18 years or older) with symptoms (first experienced within the previous 28 days) of acute infection of the respiratory system. Inclusion criteria for patients are a consultation for common cold, pharyngitis, tonsillitis, rhinosinusitis, tracheo-bronchitis, otitis media, influenza, acute exacerbations of asthma or COPD, or community acquired pneumonia, the GPs intention to prescribe antibiotics on the basis of evidence-based guidelines, and written informed consent. Exclusion criteria for patients are antibiotic use in the previous 28 days, intravenous drug use, psychiatric disorders or inability to give written informed consent, not being fluent in German, severe immunosuppression (e.g. in HIV-infection, after solid organ transplantation or under chemotherapy), cystic fibrosis, active tuberculosis, and need for immediate hospitalisation. Study practices complete the trial after including 20 patients or at the anticipated end of the trial in December 2005.
Randomisation
Allocation of patients to either treatment group is concealed by using a centralised randomisation procedure with a computer generated list produced by an independent statistician otherwise not involved in the trial. Randomisation is stratified by GP practice.
Interventions
All participating GPs received instructions about the protocol and the details of the clinical trial in a 1-hour seminar. They were asked to consecutively enrol all patients with ARTI that they judge to be in need of antibiotic treatment according to guidelines. They were told how to call the central randomisation unit, and how to fill in the necessary study forms.
Evidence-based guidelines
HCB and MB developed guidelines for the management of ARTI based on evidence-based US-position papers which were endorsed by the Centers for Disease Control and Prevention, the American Academy of Family Medicine, the American College of Physicians, and the Infectious Disease Society of America [37-42]. We systematically searched MEDLINE and the Cochrane Library to update this evidence with recent controlled clinical trials. A panel of local primary care providers, infectious disease experts, and clinical epidemiologists reviewed the guidelines and made suggestions for adaptation to local conditions. We distributed the guidelines as a booklet (see ) and presented them in a 2-hour seminar to all participating GPs.
Procalcitonin test
We measure ProCT by using a newly developed time-resolved amplified cryptate emission (TRACE) technology assay (Kryptor PCT, Brahms, Henningsdorf, Germany). This assay is based on a sheep polyclonal antibody against calcitonin and a monoclonal antibody against katacalcin, which bind to the calcitonin and katacalcin sequence of calcitonin precursor molecules. The assay has an improved functional assay sensitivity of 0.06 μg/l – i.e., three to ten fold above normal mean values. Assay time is 19 min with 20–50 μl of plasma or serum. The test is performed at the central laboratory of the University Hospital Basel, and results can be communicated to participating GPs within 2–4 h depending on the location of the practice.
Data collection and management
We obtained baseline data on all eligible GPs from the registry of the Swiss Medical Association. These data suggested that included GPs are representative of all eligible GPs with respect to age, years in private practice, years of postgraduate training, years since diploma, specialisation, and percentage of female physicians (Table 1).
Table 1 Baseline characteristics of general practitioners
Study GPs All eligible GPs
n = 53 n = 345
Age – median [IQR] 51 [42 – 55] 53 [47 – 58]
Female physicians – n (%) 9 (17) 64 (19)
Specialisation
General medicine – n (%) 25 (47) 188 (54)
Internal medicine – n (%) 26 (49) 148 (43)
Other – n (%) 2 (3.8) 9 (2.6)
Years in private practice – median [IQR] 15 [6.2 – 20] 16 [9.0 – 23]
Years of postgraduate training – median [IQR] 8.8 [7.7 – 9.7] 8.9 [7.6 – 11]
Years since diploma – median [IQR] 25 [15 – 29] 26 [14 – 31]
When a participating GP intends to give antibiotic treatment to an eligible patient based on clinical criteria and the patient gives written informed consent, the GP calls the study centre and the patient is randomly allocated to one treatment group or the other. The GP then takes a blood sample from the patient and sends it by courier service to the laboratory of clinical chemistry at the University Hospital Basel. This laboratory measures ProCT in all patients. Additionally, the GP documents patient baseline data on signs and symptoms, diagnostic procedures, diagnosis, co-morbidity and prescribed medication.
Where patients are randomised to the ProCT-arm, GPs will be informed about ProCT results and given recommendations about appropriate antibiotic therapy within 2–4 h after the blood is taken depending on the location of the practice. A cut-off ProCT level of 0.1 μg/l is used to rule out a bacterial respiratory tract infection. This value is identical to the cut-off used for the evaluation of patients in the emergency department of the University Hospital Basel [30]. In patients with a ProCT level below 0.1 μg/l, the diagnosis of a bacterial respiratory tract infection is considered highly unlikely, and the GP is encouraged to look for viral or alternative causes. Accordingly, the use of antibiotics is discouraged. In patients with a ProCT level above 0.25 μg/l, a bacterial respiratory tract infection is considered the most likely diagnosis and the use of antibiotics is recommended. For ProCT levels from 0.1 to 0.25 μg/l, a bacterial infection is unlikely and antibiotic treatment is not advocated.
The GP then informs the patient about antibiotic treatment by phone. Patients in whom antibiotics are given will be asked to use a delayed prescription or to come back to the practice to pick up the antibiotic there. For patients in whom antibiotics are withheld based on ProCT levels of 0.25 μg/l or below, a follow-up measurement of ProCT within 24 hours is mandatory. If the ProCT level on this initial follow-up is >0.25 mg/l or if it has increased by more than 50% from its initial value without clinical improvement of the patient, the use of antibiotics is recommended.
In the ProCT group, all patients treated with antibiotics will be reassessed at Day 3. Discontinuation of antibiotic treatment is recommended if the ProCT level has decreased at least to 0.25 μg/l or below.
GPs draw blood samples and document therapy at each follow-up visit. They also collect information on days with restrictions and days off work at 1 week (6–8 days) after randomisation for all patients. Medical students, blinded to treatment allocation of patients and to the goal of the trial, will conduct standardised follow-up interviews at 14 and 28 days by phone. The patient flow will be monitored according to current guidelines and in agreement with the CONSORT statement [43]. We will use Teleform® (Cardwell, Cardiff, GB) for data entry.
Adverse events
Any serious adverse event is reported by fax to the principle investigator within 24 hours. We define a severe event independent of group allocation as hospitalisation for any reason, any complication related to infection such as sepsis, abscess etc., or allergic reaction due to the received therapy, or death that occurs within 28 days following the inclusion of the patient into the trial.
Outcomes and hypotheses
The primary outcome is days with restrictions from ARTI within 14 days after randomisation. Secondary outcomes are antibiotic use in terms of antibiotic prescription rate and duration of antibiotic treatment in days, days off work and days with side-effects from medication within 14 days, and relapse rate from ARTI within 28 days after randomisation.
Our hypothesis is that the clinical outcome for patients with ARTI will be no worse under ProCT-guided treatment, but patients with ProCT-guided treatment will have lower total antibiotic use; specifically a 20% lower antibiotic prescription rate and a 20% shorter antibiotic duration compared to patients treated under the standard approach.
Sample size considerations
This is a non-inferiority trial. We aim to show that on average ProCT-guided antibiotic management leads to at most one day more with restrictions than a standard approach. We consider a type I error rate of 5% and a type II error rate of 10% (i.e. 90% power) appropriate in this situation. In a previous trial in patients with acute respiratory tract infections (ISRCTN57824788), the standard deviation in the number of days with restrictions from ARTI was 4 days for those patients prescribed antibiotics. Given this previous estimate of the variability in the primary outcome, 275 patients are needed per treatment group [44]. Allowing for a loss to follow-up of 10% gives 306 patients per treatment group, or a total of 612 patients. This sample size will allow us to estimate the reduction in antibiotic use between the two arms to within ± 6%.
With a maximum of 20 patients recruited per practice we will probably need at least 35 participating general practices. To assess between-GP-variability in a sensitivity analysis, we will need a minimum of 10 patients (preferably 15) recruited per practice.
Statistical analysis
With a non-inferiority trial, an intent-to-treat analysis is not necessarily conservative. For the primary outcome, we will need to provide several intent-to-treat analyses under different assumptions about patients who do not complete the trial, and a per-protocol analysis. Per-protocol analyses are planned for all secondary outcomes [44].
All outcomes will be analysed by a generalised linear model, assuming an appropriate distribution for each outcome and using the same set of covariates. These covariates will be: age, sex, education and a baseline score of the degree of restricted activity reported by the patient. 95% confidence intervals will be reported for the difference between treatment groups.
For the primary outcome, the assumed distribution will be normal; that is, analysis will be by multivariate linear regression. As a sensitivity analysis, the GP will be added to the model as a random effect and the model re-fit to the data from all GPs with only one GP per practice and with at least 10 patients per GP. Our experience from a previous trial (ISRCTN57824788) is that the difference between GPs has little influence on patient reported outcomes.
For secondary outcomes, assumed distributions will be binomial (i.e. for prescription of antibiotics) or normal (i.e. duration of antibiotic treatment). Duration of antibiotic treatment will be analysed only for those patients who receive antibiotics. We will also report a confidence interval for the difference in the antibiotic prescription rate between treatment groups. As a sensitivity analysis, we will repeat this calculation using a method appropriate for a cluster sample, where each GP forms a cluster and using only data from GPs with one GP per practice and with at least 10 patients per GP.
Time plan for the study
Patient recruitment began in December 2004 and is planned to continue until December 2005. By April 2005, 213 patients (35% of target) have been recruited into the trial.
Discussion
The present trial is the first randomised controlled trial to evaluate whether ProCT testing in a primary care setting reduces antibiotic use for ARTI without compromising patient relevant outcomes. In case of success, implementing this new approach into daily practice could largely improve the management of ARTI in primary care by avoiding unnecessary antibiotic use and preventing antibiotic resistance.
Our trial may have some limitations. First, this is an open intervention trial, and GPs may learn from their experience with ProCT testing and improve their clinical judgement. We cannot control for this bias, but at least this bias will be conservative for outcomes such as the antibiotic prescription rate and the duration of antibiotic therapy. Second, we expect to have recruited highly motivated primary care physicians, interested in the research question and able to provide high quality data. Motivated, interested GPs might be more reluctant to prescribe antibiotics for ARTI; thus they might consider patients for antibiotic treatment which are on average sicker than patients considered by disinterested GPs. However, we believe that ProCT-guided ARTI management will lead to a reduced antibiotic use even in such a setting of motivated GPs and the potential bias will be conservative. Third, while measurement of ProCT at a central laboratory is not ideal for routine primary care, there are still a considerable number of general practices that send blood samples daily to a laboratory for analysis of C-reactive protein or leucocytes. Therefore this should also be feasible for ProCT until a near-patient test, which is currently being developed, becomes widely available.
Abbreviations
ARTI, acute respiratory tract infections
GP, general practitioner
ProCT, procalcitonin
Competing interests
BM has served as consultant and received payments from BRAHMS (the manufacturer of procalcitonin assays) for speaking engagements, travel expenses, or research. The other authors declare that they have no competing interests.
Authors' contributions
BM and HCB conceived of the study. All authors participated in the development of research protocols and in the design of the study. JY resolved statistical issues. MB drafted the manuscript. All authors read and corrected draft versions of the manuscript and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors would like to thank Ursula Schild and Fausta Chiaverio for their contribution to the research team. We are grateful to the members of the Ethics Committee of Basel, Professor Kummer (Chairman), for their helpful comments on our study-protocol and to the members of the Data Monitoring Board, Professor Werner Zimmerli, Dr. Michael Gonon and Dr. Ferdinand Martius, for their engagement. We also would like to thank Christian Schindler for providing the randomisation scheme. Finally we would like to thank all of the participating general practitioners and practice nurses for their interest in our project.
The study is funded by the University Hospital Basel and the Swiss National Science Foundation (project number 3300C0-107772/1). BRAHMS (Henningsdorf, Germany) provided assay material.
We also acknowledge the support provided by the Swiss Medical Association (FMH).
==== Refs
Schappert SM Ambulatory care visits to physician offices, hospital outpatient departments, and emergency departments: United States, 1997 Vital Health Stat 13 1999 i 39
Arroll B Kenealy T Antibiotics for the common cold and acute purulent rhinitis. The Cochrane Database of Systematic Reviews 2002, Issue 3 Art No :CD000247 2002
Del Mar CB Glasziou PP Spinks AB Antibiotics for sore throat. The Cochrane Database of Sytematic Reviews 2004, Issue 2 Art No :CD000023 2004
Williams JW Aguilar C Cornell J Chiquette E Dolor RJ Makela M Holleman DR Simel DL Antibiotics for acute maxillary sinusitis. The Cochrane Database of Systematic Reviews 2002, Issue 3 Art No :CD000243 2002
Smucny J Fahey T Becker L Glazier R Antibiotics for acute bronchitis. The Cochrane Database of Systematic Reviews 2004, Issue 4 Art No :CD000245 2004
Glasziou PP Del Mar CB Sanders SL Hayem M Antibiotics for acute otitis media in children. The Cochrane Database of Systematic Reviews 2003, Issue 4 Art No :CD000219 2003
Graham V Lasserson TJ Rowe BH Antibiotics for acute asthma. The Cochrane Database of Systematic Reviews 2001, Issue 2 Art No :CD002741 2001
Bjerre LM Verheij TJM Kochen MM Antibiotics for community acquired pneumonia in adult outpatients. The Cochrane Database of Systematic Reviews 2004, Issue 2 Art No :CD002109 2004
Griffin JJ Meduri GU New approaches in the diagnosis of nosocomial pneumonia Med Clin North Am 1994 78 1091 1122 8078371
Gonzales R Corbett K The culture of antibiotics Am J Med 1999 107 525 526 10569312 10.1016/S0002-9343(99)00302-2
Cunha BA The antibiotic treatment of community-acquired, atypical, and nosocomial pneumonias Med Clin North Am 1995 79 581 597 7752730
Niederman MS Mandell LA Anzueto A Bass JB Broughton WA Campbell GD Dean N File T Fine MJ Gross PA Martinez F Marrie TJ Plouffe JF Ramirez J Sarosi GA Torres A Wilson R Yu VL Guidelines for the management of adults with community-acquired pneumonia. Diagnosis, assessment of severity, antimicrobial therapy, and prevention Am J Respir Crit Care Med 2001 163 1730 1754 11401897
Halm EA Teirstein AS Clinical practice. Management of community-acquired pneumonia N Engl J Med 2002 347 2039 2045 12490686 10.1056/NEJMcp020499
Mainous AGIII Hueston WJ The cost of antibiotics in treating upper respiratory tract infections in a medicaid population Arch Fam Med 1998 7 45 49 9443698 10.1001/archfami.7.1.45
Freeman J Rosner BA McGowan JEJ Adverse effects of nosocomial infection J Infect Dis 1979 140 732 740 528790
Schito GC Debbia EA Marchese A The evolving threat of antibiotic resistance in Europe: new data from the Alexander Project J Antimicrob Chemother 2000 46 Suppl T1 3 9 10997593
Wenzel RP The antibiotic pipeline--challenges, costs, and values N Engl J Med 2004 351 523 526 15295041 10.1056/NEJMp048093
Guillemot D Courvalin P Better control of antibiotic resistance Clin Infect Dis 2001 33 542 547 11462192 10.1086/322583
Raz R Porat V Ephros M Can an educational program improve the diagnosis and treatment of pharyngotonsillitis in the ambulatory care setting? Isr J Med Sci 1995 31 432 435 7607871
Santoso B Suryawati S Prawaitasari JE Small group intervention vs formal seminar for improving appropriate drug use Soc Sci Med 1996 42 1163 1168 8737434 10.1016/0277-9536(95)00390-8
Bexell A Lwando E von Hofsten B Tembo S Eriksson B Diwan VK Improving drug use through continuing education: a randomized controlled trial in Zambia J Clin Epidemiol 1996 49 355 357 8676185 10.1016/0895-4356(95)00059-3
De Santis G Harvey KJ Howard D Mashford ML Moulds RF Improving the quality of antibiotic prescription patterns in general practice. The role of educational intervention Med J Aust 1994 160 502 505 8170427
Welschen I Kuyvenhoven MM Hoes AW Verheij TJ Effectiveness of a multiple intervention to reduce antibiotic prescribing for respiratory tract symptoms in primary care: randomised controlled trial BMJ 2004 329 431 15297305 10.1136/bmj.38182.591238.EB
Zwar N Wolk J Gordon J Sanson-Fisher R Kehoe L Influencing antibiotic prescribing in general practice: a trial of prescriber feedback and management guidelines Fam Pract 1999 16 495 500 10533946 10.1093/fampra/16.5.495
Gonzales R Steiner JF Lum A Barrett PHJ Decreasing antibiotic use in ambulatory practice: impact of a multidimensional intervention on the treatment of uncomplicated acute bronchitis in adults JAMA 1999 281 1512 1519 10227321 10.1001/jama.281.16.1512
Assicot M Gendrel D Carsin H Raymond J Guilbaud J Bohuon C High serum procalcitonin concentrations in patients with sepsis and infection Lancet 1993 341 515 518 8094770 10.1016/0140-6736(93)90277-N
Muller B Becker KL Schachinger H Rickenbacher PR Huber PR Zimmerli W Ritz R Calcitonin precursors are reliable markers of sepsis in a medical intensive care unit Crit Care Med 2000 28 977 983 10809269 10.1097/00003246-200004000-00011
Harbarth S Holeckova K Froidevaux C Pittet D Ricou B Grau GE Vadas L Pugin J Diagnostic value of procalcitonin, interleukin-6, and interleukin-8 in critically ill patients admitted with suspected sepsis Am J Respir Crit Care Med 2001 164 396 402 11500339
Simon L Gauvin F Amre DK Saint-Louis P Lacroix J Serum procalcitonin and C-reactive protein levels as markers of bacterial infection: a systematic review and meta-analysis Clin Infect Dis 2004 39 206 217 15307030 10.1086/421997
Christ-Crain M Jaccard-Stolz D Bingisser R Gencay MM Huber PR Tamm M Muller B Effect of procalcitonin-guided treatment on antibiotic use and outcome in lower respiratory tract infections: cluster-randomised, single-blinded intervention trial Lancet 2004 363 600 607 14987884 10.1016/S0140-6736(04)15591-8
Christ-Crain M Jaccard-Stolz D Bingisser R Huber PR Leuppi J Mueller C Tamm M Muller B Procalcitonin guidance significantly reduces antibiotic duration in community acquired pneumonia. Proc 25th Int Congress of Intensive and Emergency Medicine (ISICEM) 2005 P166
Scott JG Cohen D DiCicco-Bloom B Orzano AJ Jaen CR Crabtree BF Antibiotic use in acute respiratory infections and the ways patients pressure physicians for a prescription J Fam Pract 2001 50 853 858 11674887
Kumar S Little P Britten N Why do general practitioners prescribe antibiotics for sore throat? Grounded theory interview study BMJ 2003 326 138 12531847 10.1136/bmj.326.7381.138
Ruttimann S Keck B Hartmeier C Maetzel A Bucher HC Long-term antibiotic cost savings from a comprehensive intervention program in a medical department of a university-affiliated teaching hospital Clin Infect Dis 2004 38 348 356 14727204 10.1086/380964
Poses RM Cebul RD Wigton RS You can lead a horse to water--improving physicians' knowledge of probabilities may not affect their decisions Med Decis Making 1995 15 65 75 7898300
Wensing M Grol R Single and combined strategies for implementing changes in primary care: a literature review Int J Qual Health Care 1994 6 115 132 7953212
Gonzales R Bartlett JG Besser RE Cooper RJ Hickner JM Hoffman JR Sande MA Principles of appropriate antibiotic use for treatment of acute respiratory tract infections in adults: background, specific aims, and methods Ann Intern Med 2001 134 479 486 11255524
Gonzales R Bartlett JG Besser RE Hickner JM Hoffman JR Sande MA Principles of appropriate antibiotic use for treatment of nonspecific upper respiratory tract infections in adults: background Ann Intern Med 2001 134 490 494 11255526
Hickner JM Bartlett JG Besser RE Gonzales R Hoffman JR Sande MA Principles of appropriate antibiotic use for acute rhinosinusitis in adults: background Ann Intern Med 2001 134 498 505 11255528
Cooper RJ Hoffman JR Bartlett JG Besser RE Gonzales R Hickner JM Sande MA Principles of appropriate antibiotic use for acute pharyngitis in adults: background Ann Intern Med 2001 134 509 517 11255530
Gonzales R Bartlett JG Besser RE Cooper RJ Hickner JM Hoffman JR Sande MA Principles of appropriate antibiotic use for treatment of uncomplicated acute bronchitis: background Ann Intern Med 2001 134 521 529 11255532
Snow V Lascher S Mottur-Pilson C Evidence base for management of acute exacerbations of chronic obstructive pulmonary disease Ann Intern Med 2001 134 595 599 11281744
Moher D Schulz KF Altman DG The CONSORT statement: revised recommendations for improving the quality of reports of parallel-group randomized trials Ann Intern Med 2001 134 657 662 11304106
Jones B Jarvis P Lewis JA Ebbutt AF Trials to assess equivalence: the importance of rigorous methods BMJ 1996 313 36 39 8664772
|
16107222
|
PMC1190167
|
CC BY
|
2021-01-04 16:29:12
|
no
|
BMC Fam Pract. 2005 Aug 18; 6:34
|
utf-8
|
BMC Fam Pract
| 2,005 |
10.1186/1471-2296-6-34
|
oa_comm
|
==== Front
BMC GenetBMC Genetics1471-2156BioMed Central London 1471-2156-6-401602662010.1186/1471-2156-6-40Research ArticleThe missense mutation in Abcg5 gene in spontaneously hypertensive rats (SHR) segregates with phytosterolemia but not hypertension Chen Jianliang [email protected] Ashok [email protected] Shuqin [email protected] Wayne R [email protected] Michael E [email protected] Hongwei [email protected] Patrick [email protected] Gerald [email protected] Shailendra B [email protected] Division of Endocrinology, Diabetes and Medical Genetics, Medical University of South Carolina, STR 541, 114 Doughty Street, Charleston, SC 29403, USA2 Research Service and Medical Service, Department of Veterans Affairs Medical Center, East Orange, NJ 07019, USA3 Department of Pathology and laboratory Medicine, University of Cincinnati College of Medicine, Cincinnati, Ohio 45267. USA4 Division of Nephrology, Medical University of South Carolina, Charleston, SC 29403, USA2005 18 7 2005 6 40 40 10 2 2005 18 7 2005 Copyright © 2005 Chen et al; licensee BioMed Central Ltd.2005Chen et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Sitosterolemia is a recessively inherited disorder in humans that is associated with premature atherosclerotic disease. Mutations in ABCG5 or ABCG8, comprising the sitosterolemia locus, STSL, are now known to cause this disease. Three in-bred strains of rats, WKY, SHR and SHRSP, are known to be sitosterolemic, hypertensive and they carry a missense 'mutation' in a conserved residue of Abcg5, Gly583Cys. Since these rat strains are also know to carry mutations at other genetic loci and the extent of phytosterolemia is only moderate, it is important to verify that the mutations in Abcg5 are causative for phytosterolemia and whether they contribute to hypertension.
Methods
To investigate whether the missense change in Abcg5 is responsible for the sitosterolemia we performed a segregation analysis in 103 F2 rats from a SHR × SD cross. Additionally, we measured tail-cuff blood pressure and measured intestinal lipid transport to identify possible mechanisms whereby this mutation causes sitosterolemia.
Results
Segregation analysis showed that the inheritance of the Gly583Cys mutation Abcg5 segregated with elevated plant sterols and this pattern was recessive, proving that this genetic change is responsible for the sitosterolemia in these rat strains. Tail-cuff monitoring of blood pressure in conscious animals showed no significant differences between wild-type, heterozygous and homozygous mutant F2 rats, suggesting that this alteration may not be a significant determinant of hypertension in these rats on a chow diet.
Conclusion
This study shows that the previously identified Gly583Cys change in Abcg5 in three hypertension-susceptible rats is responsible for the sitosterolemia, but may not be a major determinant of blood pressure in these rats.
==== Body
Background
Sitosterolemia is an autosomal recessive disease, characterized by significantly increased plasma levels of plant sterols (such as sitosterol, campesterol), and is associated with premature atherosclerotic disease [1]. This disease has been mapped to a single locus, STSL, on human chromosome 2p21 [2,3]. Mutations in both alleles of one of two genes, ABCG5 or ABCG8, that comprise this locus, are now known to cause this disease [4-6]. No phytosterolemic patient with a single mutant ABCG5 allele and a mutant ABCG8 allele has been reported, suggesting these genes are not only linked physically, but their protein products may act as obligate heterodimers. ABCG5 and ABCG8 encode for sterolin-1 and sterolin-2 respectively. These genes are expressed in the liver, gall bladder and intestine and are implicated in determining biliary sterol excretion and selectivity of sterol absorption at the apical surfaces of the enterocytes [7-12]. Studies in mice deficient for Abcg5 or Abcg8, as well as one that over-expresses these genes, have confirmed that sterolins may be the major determinants of biliary sterol secretion [13-18]. Their role in preventing non-cholesterol sterol absorption in the intestine remains to be clarified.
Prior to the identification of the gene defects in sitosterolemia, the presence of sitosterolemia had been reported in the spontaneously hypertensive rat (SHR), although the molecular basis of this phenotype had not been investigated [19-25]. Sitosterolemia, hemolysis (a feature also present in human sitosterolemia) and an increased mortality when fed certain plant oils (enriched in plant sterols) have also been reported in SHR and SHRSP rats [20]. The SHR, SHRSP and their 'parental' strain, WKY rats have now been shown to carry a homozygous G1757T alteration in Abcg5 that results in a Gly583Cys change [26,27]. All of these rat strains are also sitosterolemic. However, all of these rats are highly in-bred and have been maintained by brother-sister matings [28]. These rats are likely to carry a number of mutated genes as part of their genetic burden. At least 4 'mutated' loci are known to be part of this burden; Cd36 [29], Srebp-1 [30], Kat-2 [31] and Abcg5 [26,27]. Although the Gly583Cys affects a highly conserved residue (conserved from Fugu to Man) [27], given the known genetic burden in the SHR line, it remains possible that genetic alterations at other loci may be responsible for the sitosterolemia. Support for this possibility has recently come from a quantitative trait genetic mapping study for sitosterolemia involving two mouse strains, C57Bl/6J and CASA/Rk [32,33]. Although the plasma sitosterol level was genetically determined in these mouse strains, loci that accounted for this did not map to the murine STSL and thus other unidentified genes were hypothesized to be involved [32,33].
We report here genetic and biochemical analyses of F2 rats from a SHR × SD cross to test the hypothesis that the Gly583Cys is responsible for the sitosterolemia. In addition, since hypertension is a known phenotype of the SHR and since feeding diets rich in plant sterols seem to decrease their life-span, we also investigated the possibility that this genetic change may also be important in playing a role in hypertension.
Results
Segregation analyses of plasma cholesterol and plant sterols
To ensure there were no significant differences in growth curves of the F2 animals by Abcg5 genotype, body weights were measured serially and analyzed only after the genotypes had been determined at the end of the 14 weeks (Fig. 1). The genetic status at the STSL locus did not affect growth parameters. Blood from 14-week old F2 animals fed a chow diet was obtained after a 4 h fast and analyzed for cholesterol, as well as plant sterols, determined by GC analyses (see Methods). All sterol analyses were performed independent of the genotype analyses. Fig. 2 shows the distribution of cholesterol (Fig 2A) and sitosterol (Fig 2B) in these animals. The distribution of cholesterol was slightly skewed to the right, with a mean cholesterol value of 59.5 mg/dl. The distribution of sitosterol was significantly right-skewed and suggested a bi-modal pattern.
Figure 1 Growth Curves of F2 mice. The body weights of male (top set of lines) and female (bottom set of lines) F2 rats were monitored with time. At 17 weeks, most of the male F2 animals were sacrificed, or were used to determine intestinal absorption (see Fig. 5). At 17 weeks, female F2 rats were placed on a more defined plant sterol diet and their drinking water substituted with 1% NaCl. No differences in body weight gains between any of the Abcg5 genotypes (wild-type, WT; heterozygous, HET and homozygous mutant, HMZ) were observed.
Figure 2 Sterol distribution in F2 rats at 14 weeks age. Blood for sterol analyses by GC was drawn at 14 weeks age (see Methods). The frequency distribution of cholesterol, in deciles, was almost Gaussian, with a slight right skew (panel A). However, plasma sitosterol, expressed in 0.5 mg/dl increments, was significantly right skewed (panel B).
All animals were genotyped for the Abcg5 G1757T mutation and grouped according to wild-type (WT), heterozygous (HET) or homozygous mutant (HMZ) and their plasma cholesterol (Fig. 3A) and sitosterol (Fig. 3B) values plotted as individual values. Two investigators, each blinded to each other's results performed the genotype and sterol analyses separately. There were no significant differences in plasma cholesterol values between any of the genotypes (Fig. 3A). All but two data points (highlighted in Fig. 3B) showed a robust segregation of homozygous inheritance of mutant alleles with elevated plant sterol levels. The two exceptions were a WT rat with an elevated plant sterol and a homozygous mutant rat that exhibited low levels of sitosterol. Unfortunately, the plant sterol levels were made available only after all of these animals had been sacrificed and it was not possible to re-sample the blood to re-confirm the plasma values, although archived DNA allowed us to confirm the genotypes. Thus an error in sample mis-identification could not be confirmed.
Figure 3 Plasma sterols as a function of Abcg5 genotypes. Cholesterol (panel A) and sitosterol (panel B) were segregated in all F2 animals by Abcg5 Gly583Cys. While there were no differences between cholesterol values between any of the genotypes, F2 rats who were homozygous for Cys583 (HMZ) showed significantly higher plasma sitosterol values (*, P < 0.05, mean 5.12 ± 0.25 mg/dl) compared to either heterozygous animals (HET, mean 2.68 ± 0.16 mg/dl) or wild-type (WT, mean 2.25 ± 0.20 mg/dl). These data were analyzed with the inclusion of two 'outliers' (arrows), one in the WT group and one in the HMZ group, where re-analyses to confirm these values were not possible (see Text).
Plasma levels of campesterol, campestanol and total plant sterols were also significantly elevated in the HMZ animals compared to WT and HET animals (Fig. 4A, B and D), with non-significant increases in sitostanol (Fig. 4C). It is also interesting to note that a small number of the heterozygous animals had plasma sitosterol, campesterol and total plant sterol values as high as those from rats homozygous for the Abcg5 mutation. All of these measurements were obtained on animals fed a chow diet and a diet challenge to distinguish between these various genotypes was not performed.
Figure 4 Segregation analyses of other plant sterols by genotype. Segregation of campesterol (panel A), its metabolite, campestanol (panel B), the metabolite of sitosterol, sitostanol (panel C), or total plant sterols (panel D) are as shown. In all cases, except for sitostanol, the homozygous F2 rats had significantly elevated levels (* P < 0.05).
Intestinal absorption of lipid
To identify possible mechanisms for the elevated plant sterols, we performed lymph duct cannulation in WT, HET and HMZ male rats and monitored absorption of cholesterol, and fat. Although we attempted lymph duct cannulation in 30 animals, we obtained data in only 3 WT, 4 HET and 4 HMZ animals due to either a high mortality or unacceptable lymph flow. Lymphatic flow rates in successfully cannulated rats are shown in Fig. 5A. There were no differences by genotype. Following a bolus delivery of radioactive cholesterol and triolein (see Methods), no differences were observed between WT, HET or HMZ rats for recovery of radioactivity in the lymph for either triglyceride or cholesterol (Fig. 5A and B), suggesting that these pathways were relatively normal. Lymphatic recovery for absorption of sitosterol was not performed.
Figure 5 Intestinal lipid absorption in F2 rats. Male F2 rats had their lymph ducts cannulated as described (see Methods) and had lymph flow quantitiated to ensure adequate and stable flow (panel A). Labeled triolein (panel B) or cholesterol (panel C was administered via the duodenal tube, and the recovery of radio-isotope in the lymph quantitated with time. No differences in the appearance of label for either triolein (triglyceride absorption) or cholesterol was observed between WT, HET or HMZ F2 rats.
Biliary secretion of plasma cholesterol and plant sterols
To further elucidate the mechanism by which plant sterols were accumulating in the Abcg5 mutant animals, we collected and analyzed bile for sterols in 24-week old after they had been fed a defined diet containing 300 mg/kg of sitosterol (Fig. 6). There were no differences in the amounts of cholesterol present in the bile from any of the three genotypes (Fig 6A). However, significantly more sitosterol was present in the bile of mutant rats compared to either HET or WT rats (Fig 6B), suggesting that the excretory pathways are functional and not affected by this mutation. Thus the increased plasma plant sterol levels were reflected by the increased biliary secretion.
Figure 6 Biliary cholesterol and sitosterol concentrations. Female F2 rats, on a defined sitosterol diet (see text) had their bile ducts cannulated and bile collected under gravity and analyzed for cholesterol (panel A) or sitosterol (panel B). No differences in the cholesterol content of bile were noted between any of the genotypes. Note that the concentration of cholesterol is approximately 50-fold more than sitosterol. Interestingly, the sitosterol concentration in the F2 HMZ rats was increased significantly by ~ 50% relative to WT and HET rats (panel B).
Effect of Abcg5 mutation on blood pressure
We used the tail-cuff measurement technique to monitor systolic blood pressure in these rats on a chow diet, with time. All blood pressure measurements were accumulated independent of the genotype results and only after all blood pressure measurements had been accumulated were the data analyzed by genotype (Fig. 7). In male (Fig 7A) or female (Fig 7B) mice on a chow diet, no differences up to 16 weeks of age were noted. For comparison, male SD rats had a mean systolic blood pressure of 128 mm Hg and the value for SHR was 197 mm Hg.
Figure 7 Tail cuff blood pressure in male and female F2 rats with time. Tail cuff measured blood pressures for male F2 rats (panel A) and female F2 rats (panel B) were monitored with time and analyzed by genotype (solid squares, WT; solid diamonds, HET and solid triangles, HMZ). No significant differences were noted between any of these genotypes. For comparison, the average blood pressures of rats at age 16 weeks for SD and SHR were 127 mm Hg and 197 mm Hg respectively. Additionally, female rats at 17 weeks were placed on a defined sitosterol diet and fed 1% NaCl drinking water (see bar, panel B). Neither significant increases in blood pressures, nor any differences between the genotypes were observed.
In a separate study, female mice were switched to a defined diet (containing 300 mg/kg plant) and had 1% NaCl added to the drinking water and blood pressures recorded for a further 5 weeks (Fig 7B, see bar). No differences in blood pressure were seen in any of the genotypes, nor was there a progressive increase in overall blood pressures recorded.
Discussion
Sitosterolemia is a rare autosomal recessive inherited metabolic disorder of man that results in the accumulation on non-cholesterol sterols from the diet and is associated with premature coronary heart disease, hemolytic anemia and the formation of tendon and tuberous xanthomas [1]. The genetic defect(s) has now been mapped and identified. Mutations in one of two genes organized in tandem on human chromosome 2p21, ABCG5 or ABCG8, cause this defect [34]. As part of the literature survey, we observed that elevations of plant sterols had been reported in rats commonly used for hypertension and stroke research, WKY, SHR and SHRSP [19-25]. Interestingly, a number of studies had reported previously that alteration in diets not only increased the hemolysis associated with the presence of sitosterol in these rats, but had a dramatic effect on life-expectancy as well as led to an increase in hypertension [20,24]. Thus SHRSP rats fed canola or soy-bean oils supplemented with phytosterols led to a significantly shortened life-span relative to oils low in plant sterols [19]. Other investigators have reported the association of hypertension in these rat strains when fed diets supplemented with sitosterol, suggesting that hypertension may be modified by plant sterol accumulation [22].
Examination of SHR, SHRSP and WKY rats showed that these rats carried a homozygous missense in Abcg5, Gly583Cys, affecting a residue that shows conservation from Fugu to Man, despite the relative conservative change with respect to the amino acids involved [26,27]. At present, there are no control strains for WKY, SHR or SHRSP, since these strains have been maintained by brother-sister matings for >80 generations. Additionally, some of these strains have been maintained by different investigators for several generations raising the possibility that there may also be sub-strain variations between different colonies. Although these strains have been used extensively in genetic analyses of hypertension and a number of loci mapped for this trait, none of these loci map to the rat STSL locus on chromosome 6 [35]. In a literature search of genetic changes reported for these strains, we were able to compile at least 3 other loci that are mutated in the SHR strain (Cd36, Srebp-1 and Kat-2), suggesting the possibility that the missense 'mutation' identified in Abcg5 is not causative for sitosterolemia (which is also moderate in comparison to levels seen in patients with sitosterolemia who have levels frequently >15 mg/dl), but associated as a result of the in-breeding [29-31]. Additionally, a murine QTL study of plasma sitosterol levels showed that plant sterol levels were strongly influenced genetically, but mapping studies showed that this trait did not map to the murine STSL on chromosome 17 [33], raising the possibility that other loci/factors may be involved. Finally, in a study involving the GH and Norwegian Brown rat strains (neither of which are sitosterolemic), a QTL for hypertension was mapped to the region encompassing the rat STSL. However, we sequenced the rat STSL in both of these strains and did not identify any differences, suggesting that variations at the STSL in the rat may not be contributing to the hypertension [27]. We therefore undertook this study to test the hypothesis that the Gly583Cys was causative for sitosterolemia. We also measured blood pressure to test the hypothesis that this mutation was responsible for contributing to an elevation in blood pressure in these affected strains.
Segregation analyses in a large F2 population (SHR × SD) showed that homozygosity for Gly583Cys segregated with elevated plasma plant sterol levels in all but two animals. Of these latter animals, we could not verify that the samples had been inadvertently mis-labeled, as we could not repeat plasma analysis as the animals had been sacrificed before the plant sterol analyses were available. If we exclude these animals, all animals that are homozygous for the Gly583Cys are also sitosterolemic on a rodent chow diet, proving that this genetic change was responsible for the elevation in dietary non-cholesterol sterols.
We also tracked blood pressures, using tail cuff measurements, in all weaned animals for up to 4 months. No differences in blood pressures by genotype were observed in these F2 animals. In a sub-study, at ~ 4 months, female rats were switched to a more defined diet that had a higher level of plant sterols (300 mg/kg compared to ~ 130 mg/kg in most rodent chow preparations). This was done as normal rodent chow is known to lead to variable sterol absorption. Additionally, the rats were given drinking water supplemented with 1% NaCl to increase any manifestation of hypertension. Despite these interventions, rats homozygous for the Gly583Cys change did not show any significant increases in hypertension, relative to wild-type or heterozygous animals. Thus, in our study, we could not demonstrate an increase in blood pressures, contrary to other published data. There are several explanations for this discrepant finding. Firstly, our animals are F2 animals and since a number of loci are implicated in the hypertension phenotype, these should segregate independently of the STSL locus (except for one QTL for hypertension identified in the GH × BN experiment, but not reported for crosses involving SHR). Secondly, the levels of plant sterols we employed in our dietary studies are considerably less than used by others. In almost all of the studies reported where the exact amount of plant sterols were reported, these have been almost 10-fold or more higher than those used in our current study. Our justification for not using such large quantities is that it is unlikely such intakes reflect amounts that would be normally consumed and thus may not be 'physiological'. Thirdly, the blood pressure measurements were undertaken in animals fed plain drinking water and in all of the reported studies where hypertension was monitored, supplementation with a high salt intake (in water, diet or both) has been used. Finally, the majority of studies using higher plant sterols also used increase plant oils as well and the effect of these on altering the phenotype is not easy to discern.
We are aware of the limitations of the tail-cuff technique for blood pressure determinations and subtle, but significant differences, in blood pressure may have been masked by the use of this technique. Our data indicate that variations at the STSL locus may not be responsible for major effects in contributing to blood pressure. Any subtle effects will need to be explored in rat strains that are congenic for Gly583Cys on either the SHR or SD backgrounds.
Although the Gly583Cys change affects an amino acid residue that is conserved from Fugu to Man in Abcg5, this amino acid substitution is a relatively conservative change, with no both amino acids being aliphatic and small. To delineate possible mechanisms by which this genetic change altered dietary sterol trafficking, we compared intestinal absorption for cholesterol and triglyceride, using lymphatic duct cannulation studies. These showed that neither cholesterol absorption, nor triglyceride absorption (digestion, uptake, re-synthesis and secretion) were affected by this change. Note that patients with sitosterolemia, mutated in Abcg8, have been reported to manifest hyper-absorption of cholesterol, as well as plant sterols [34]. Unfortunately, sitosterol absorption was not determined at the same time, as these studies were performed by a core (PT, University of Cincinnati) not routinely using sitosterol. Biliary excretion of cholesterol and sitosterol was unaffected by the Gly583Cys alterations, although the homozygous mutant F2 rats showed increased amounts of plant sterols in their bile. Thus, although not definitive, this mutation may cause an increase in the intestinal absorption of non-cholesterol sterols, leading to increased plasma concentrations. Despite an increase in biliary secretion, a small net retention may result, accounting for the sitosterolemia. In this context, it should be pointed out that the levels of sitosterolemia observed in these rats are considerably lower than those seen in patients with sitosterolemia. Future experiments, using congenic mice and lymph duct cannulation will allow us to examine this hypothesis and delineate pathophysiology of this missense mutation.
Conclusion
We report here segregation analyses that prove that the Gly583Cys alteration identified in WKY, SHR and SHRSP strains is causative for sitosterolemia, that this alteration may not be a significant contributor to the hypertension phenotype, and that it may be causing sitosterolemia by primarily increasing intestinal absorption on non-cholesterol sterols. Our study also highlights the need to develop lines congenic for the missense mutation on an SD background and the wild-type residue on the SHR background to allow for the study of Abcg5 function on its role in the diet-induced decrease in longevity in these hypertensive strains.
Methods
Animal breeding and husbandry
The Animal Care and Use Committee of Medical University of South Carolina approved all animal protocols. Inbred SHR (SHR/NCrlBR) rats and outbred CD® (SD) IGS BR rats were purchased from Charles River Laboratories (CRL, Wilmington, MA). Animals were housed 2–3 rats/cage in a temperature-controlled room (22 ± 2°C) with a 12-hour light/dark cycle and fed Teklad Sterilized Rodent Diet (W) 8656 (Harlan, Madison, WI) with free access to water. To produce F1 progeny, (1 male and 1 female) SHR and (1 male and 1 female) SD were cross-mated and the F1 resultant progeny randomly mated (4 breeding pairs) to produce 103 F2 progeny. All F2 animals were fed standard rodent chow and had free access to water. At 17 weeks of age, female rats were fed a high sitosterol diet (sitosterol, 300 mg/kg and cholesterol, 150 mg/kg, Harlan Teklad diet) and their drinking water supplemented with 1% Sodium chloride for 6 weeks. Males were shipped to University of Cincinnati (PT lab) for lymph duct cannulation.
Blood pressure and body weight measurements
Blood pressure (BP) was measured in the warmed, conscious, restrained state by using the photoelectric oscillometric tail-cuff method using a Natsume KN-210 machine (Natsume Seisakusho Co. Ltd., Japan). Animals were accustomized to the apparatus for 1 week, before recordings were collected. At each time point, at least 10 readings were obtained for each rat and the average of the readings taken. Body weights were recorded after each BP measurement. All data were accumulated independent of the genotype data.
Genotyping
Blood samples were drawn at 14 weeks of age from all rats. In addition, blood, was obtained at 20 and 24 weeks from female rats; the plasma was saved at -70°C for sterol analysis and genomic DNA, isolated from WBC, used for genotyping by PCR and Hae III digestion as described previously [27].
Bile collection
Under Isoflurane anesthesia, the bile duct was cannulated and bile collected for 15 minutes under gravity as described previously [16].
Sterol analyses
Plasma and bile samples were processed according to the methods described by Heinemann [36]. Plasma total cholesterol levels were measured by using gas-liquid chromatography (GC), and plant sterol levels determined by GC-mass spectrometry with 5-cholestane and epicoprostanol as internal standards, respectively, as described previously. Total plant sterols included campestanol, campesterol, sitostanol and sitosterol. Campestanol and sitostanol are 5-dihydro derivatives of campesterol and sitosterol. Sterol analyses and genotype analyses were performed independently and only after completion of both were results linked.
Intestinal absorption
The intestinal lymph duct was cannulated using a clear vinyl tubing (o.d., 0.8 mm) according to the method of Bollman et al. [37]. The second soft silicone tubing (o.d., 1.6 mm) was installed in the stomach through the fundus into the duodenum and secured with a purse-string suture. A purse-string suture (6-0 silk) and cyanoacrylate glue were used to secure the duodenal infusion tube. The gastroduodenal and lymph duct tubes were externalized through the right flank of the rat to allow easy access for infusion and collection. The skin, abdominal musculature and peritoneal layers were closed in a single layer (4-0 silk). The rats were kept in restraining cages at an ambient temperature of 30°C and allowed to recover from the anesthesia. The next morning, a lipid emulsion containing 6 μmol of 3H-triolein, 0.78 μmol of 14C-cholesterol, 0.87 μmol of egg phosphatidylcholine, and 5.7 μmol of sodium taurocholate (19 mM) in 1.5 ml of phosphate-buffered saline (pH 6.4) was infused at 1.5 ml/h for 6 h. Lymph was collected 1 h before the lipid infusion (fasting) and then hourly for 6 h. The volume of collected lymph was measured by weighting. Samples were counted for 10 min by a liquid scintillation spectrometer (Model TR 1900 tri-card; Packard).
Statistical analysis
Statistical analysis was performed by program Prism 2.0. Comparisons of parameters among 3 groups were analyzed by one-way ANOVA with Newman-Keuls's post-test and t-test. All data are expressed as mean ± SD.
Abbreviations
ABCG5, ATP binding cassette G5; SD. Sprague Dawley STSL, sitosterolemia locus; SHR, spontaneously hypertensive rat, SHRSP; SHR stroke-prone;
Authors' contributions
The studies were designed by JC, HY and SBP and all experimental data obtained by JC, SZ, AB, WF and PT. Data analyses were performed by JC, SBP, AB, PT and GS. The paper was written by JC and SBP and edited by all the authors.
Acknowledgements
We wish to thank Jana Fine for her expertise with tail-cuff blood pressure measurements and to Bibi for her expertise with sterol analyses. This work was supported by the National Institutes of Health, NHLBI HL060613 (SBP).
==== Refs
Bjorkhem I Boberg KM Scriver CR, Beaudet AL, Sly WS and Valle D Inborn errors in bile acid biosynthesis and storage of sterols other than cholesterol The Metabolic Basis of Inherited Disease 1995 2 7th New York, McGraw-Hill Inc. 2073 2102
Patel SB Salen G Hidaka H Kwiterovich PO Stalenhoef AF Miettinen TA Grundy SM Lee MH Rubenstein JS Polymeropoulos MH Brownstein MJ Mapping a gene involved in regulating dietary cholesterol absorption. The sitosterolemia locus is found at chromosome 2p21 J Clin Invest 1998 102 1041 1044 9727073
Lu K Lee MH Carpten JD Sekhon M Patel SB High-Resolution Physical and Transcript Map of Human Chromosome 2p21 Containing the Sitosterolemia Locus European Journal of Human Genetics 2001 9 364 374 11378825 10.1038/sj.ejhg.5200627
Berge KE Tian H Graf GA Yu L Grishin NV Schultz J Kwiterovich P Shan B Barnes R Hobbs HH Accumulation of dietary cholesterol in sitosterolemia caused by mutations in adjacent ABC transporters Science 2000 290 1771 1775 11099417 10.1126/science.290.5497.1771
Lee MH Lu K Hazard S Yu H Shulenin S Hidaka H Kojima H Allikmets R Sakuma N Pegoraro R Srivastava AK Salen G Dean M Patel SB Identification of a gene, ABCG5, important in the regulation of dietary cholesterol absorption Nature Genetics 2001 27 79 83 11138003 10.1038/87170
Lu K Lee MH Hazard S Brooks-Wilson A Hidaka H Kojima H Ose L Stanlenhoef AFH Mietinnen T Bjorkhem I Brukert E A. P Brewer HB Salen G Dean M Srivastava A Patel SB Two genes that map to the STSL locus cause sitosterolemia: Genomic structure and spectrum of mutations involving sterolin-1 and sterolin-2, encoded by ABCG5 and ABCG8 respectively American Journal of Human Genetics 2001 69 278 290 11452359 10.1086/321294
Graf GA Li WP Gerard RD Gelissen I White A Cohen JC Hobbs HH Coexpression of ATP-binding cassette proteins ABCG5 and ABCG8 permits their transport to the apical surface J Clin Invest 2002 110 659 669 12208867 10.1172/JCI200216000
Kipp H Arias IM Trafficking of canalicular ABC transporters in hepatocytes Annu Rev Physiol 2002 64 595 608 11826281 10.1146/annurev.physiol.64.081501.155793
Lu K Lee MH Yu H Zhou Y Sandell SA Salen G Patel SB Molecular cloning, genomic organization, genetic variations, and characterization of murine sterolin genes Abcg5 and Abcg8 J Lipid Res 2002 43 565 578 11907139
Repa JJ Berge KE Pomajzl C Richardson JA Hobbs H Mangelsdorf DJ Regulation of ATP-binding cassette sterol transporters ABCG5 and ABCG8 by the liver X receptors alpha and beta J Biol Chem 2002 277 18793 18800 11901146 10.1074/jbc.M109927200
Klett EL Lee MH Adams DB Chavin KD Patel SB Localization of ABCG5 and ABCG8 proteins in human liver, gall bladder and intestine BMC Gastroenterol 2004 4 21 15383151 10.1186/1471-230X-4-21
Tauscher A Kuver R ABCG5 and ABCG8 are expressed in gallbladder epithelial cells Biochem Biophys Res Commun 2003 307 1021 1028 12878214 10.1016/S0006-291X(03)01296-8
Yu L Li-Hawkins J Hammer RE Berge KE Horton JD Cohen JC Hobbs HH Overexpression of ABCG5 and ABCG8 promotes biliary cholesterol secretion and reduces fractional absorption of dietary cholesterol J Clin Invest 2002 110 671 680 12208868 10.1172/JCI200216001
Yu L Hammer RE Li-Hawkins J Von Bergmann K Lutjohann D Cohen JC Hobbs HH Disruption of Abcg5 and Abcg8 in mice reveals their crucial role in biliary cholesterol secretion Proc Natl Acad Sci U S A 2002 99 16237 16242 12444248 10.1073/pnas.252582399
Graf GA Yu L Li WP Gerard R Tuma PL Cohen JC Hobbs HH ABCG5 and ABCG8 are obligate heterodimers for protein trafficking and biliary cholesterol excretion J Biol Chem 2003 278 48275 48282 14504269 10.1074/jbc.M310223200
Klett EL Lu K Kosters A Vink E Lee MH Altenburg M Shefer S Batta AK Yu H Chen J Klein R Looije N Oude-Elferink R Groen AK Maeda N Salen G Patel SB A mouse model of sitosterolemia: absence of Abcg8/sterolin-2 results in failure to secrete biliary cholesterol BMC Med 2004 2 5 15040800 10.1186/1741-7015-2-5
Kosters A Frijters RJ Kunne C Vink E Schneiders MS Schaap FG Nibbering CP Patel SB Groen AK Diosgenin-induced biliary cholesterol secretion in mice requires Abcg8 Hepatology 2005 41 141 150 15619238 10.1002/hep.20540
Yu L Gupta S Xu F Liverman AD Moschetta A Mangelsdorf DJ Repa JJ Hobbs HH Cohen JC Expression of ABCG5 and ABCG8 is required for regulation of biliary cholesterol secretion J Biol Chem 2005 280 8742 8747 15611112 10.1074/jbc.M411080200
Ratnayake WM Plouffe L Hollywood R L'Abbe MR Hidiroglou N Sarwar G Mueller R Influence of sources of dietary oils on the life span of stroke-prone spontaneously hypertensive rats Lipids 2000 35 409 420 10858026
Ratnayake WM L'Abbe MR Mueller R Hayward S Plouffe L Hollywood R Trick K Vegetable oils high in phytosterols make erythrocytes less deformable and shorten the life span of stroke-prone spontaneously hypertensive rats J Nutr 2000 130 1166 1178 10801914
Ikeda I Nakagiri H Sugano M Ohara S Hamada T Nonaka M Imaizumi K Mechanisms of phytosterolemia in stroke-prone spontaneously hypertensive and WKY rats Metabolism 2001 50 1361 1368 11699058 10.1053/meta.2001.26739
Ogawa H Yamamoto K Kamisako T Meguro T Phytosterol additives increase blood pressure and promote stroke onset in salt-loaded stroke-prone spontaneously hypertensive rats Clin Exp Pharmacol Physiol 2003 30 919 924 14678230 10.1111/j.1440-1681.2003.03939.x
Naito Y Kasama K Yoshida H Ohara N Thirteen-week dietary intake of rapeseed oil or soybean oil as the only dietary fat in Wistar Kyoto rats-change in blood pressure Food Chem Toxicol 2000 38 811 816 10930702 10.1016/S0278-6915(00)00076-4
Naito Y Konishi C Ohara N Blood coagulation and osmolar tolerance of erythrocytes in stroke-prone spontaneously hypertensive rats given rapeseed oil or soybean oil as the only dietary fat Toxicol Lett 2000 116 209 215 10996482 10.1016/S0378-4274(00)00224-1
Naito Y Yoshida H Nagata T Tanaka A Ono H Ohara N Dietary intake of rapeseed oil or soybean oil as the only fat nutrient in spontaneously hypertensive rats and Wistar Kyoto rats - blood pressure and pathophysiology Toxicology 2000 146 197 208 10814852 10.1016/S0300-483X(00)00175-X
Scoggan KA Gruber H Lariviere K A missense mutation in the Abcg5 gene causes phytosterolemia in SHR, stroke-prone SHR, and WKY rats Journal of Lipid Research 2003 44 911 916 12611906 10.1194/jlr.M200438-JLR200
Yu H Pandit B Klett E Lee MH Lu K Helou K Ikeda I Egashira N Sato M Klein R Batta A Salen G Patel SB The rat STSL locus: characterization, chromosomal assignment, and genetic variations in sitosterolemic hypertensive rats BMC Cardiovasc Disord 2003 3 4 12783625 10.1186/1471-2261-3-4
Kurtz TW Montano M Chan L Kabra P Molecular evidence of genetic heterogeneity in Wistar-Kyoto rats: implications for research with the spontaneously hypertensive rat Hypertension 1989 13 188 192 2914738
Aitman TJ Glazier AM Wallace CA Cooper LD Norsworthy PJ Wahid FN Al-Majali KM Trembling PM Mann CJ Shoulders CC Graf D St Lezin E Kurtz TW Kren V Pravenec M Ibrahimi A Abumrad NA Stanton LW Scott J Identification of Cd36 (Fat) as an insulin-resistance gene causing defective fatty acid and glucose metabolism in hypertensive rats Nat Genet 1999 21 76 83 9916795 10.1038/5013
Pravenec M Jansa P Kostka V Zidek V Kren V Forejt J Kurtz TW Identification of a mutation in ADD1/SREBP-1 in the spontaneously hypertensive rat Mamm Genome 2001 12 295 298 11309661 10.1007/s003350010273
Kwok JB Kapoor R Gotoda T Iwamoto Y Iizuka Y Yamada N Isaacs KE Kushwaha VV Church WB Schofield PR Kapoor V A missense mutation in kynurenine aminotransferase-1 in spontaneously hypertensive rats J Biol Chem 2002 277 35779 35782 12145272 10.1074/jbc.C200303200
Sehayek E Genetic regulation of cholesterol absorption and plasma plant sterol levels: commonalities and differences J Lipid Res 2003 44 2030 2038 12897193 10.1194/jlr.R300008-JLR200
Sehayek E Duncan EM Lutjohann D Von Bergmann K Ono JG Batta AK Salen G Breslow JL Loci on chromosomes 14 and 2, distinct from ABCG5/ABCG8, regulate plasma plant sterol levels in a C57BL/6J x CASA/Rk intercross Proc Natl Acad Sci U S A 2002 99 16215 16219 12446833 10.1073/pnas.212640599
Lee MH Lu K Patel SB Genetic basis of sitosterolemia Current Opinion in Lipidology 2001 12 141 149 11264985 10.1097/00041433-200104000-00007
Stoll M Kwitek-Black AE Cowley AWJ Harris EL Harrap SB Krieger JE Printz MP Provoost AP Sassard J Jacob HJ New target regions for human hypertension via comparative genomics Genome Res 2000 10 473 482 10779487 10.1101/gr.10.4.473
Heinemann T Axtmann G von Bergmann K Comparison of intestinal absorption of cholesterol with different plant sterols in man Eur J Clin Invest 1993 23 827 831 8143759
Bollman J Cain JC Grindlay JH Techniques for the collection of lymph from the liver, small intestine or thoracic duct of the rat Journal of Laboratory and Clinical Medicine 1949 33 1349 1352
|
16026620
|
PMC1190168
|
CC BY
|
2021-01-04 16:30:38
|
no
|
BMC Genet. 2005 Jul 18; 6:40
|
utf-8
|
BMC Genet
| 2,005 |
10.1186/1471-2156-6-40
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-1021604280210.1186/1471-2164-6-102Research ArticleComputational tradeoffs in multiplex PCR assay design for SNP genotyping Rachlin John [email protected] Chunming [email protected] Charles [email protected] Simon [email protected] Bioinformatics program, Boston University, Boston MA 02215, USA2 Centre for Emerging Infectious Diseases, The Chinese University of Hong Kong, Prince of Wales Hospital, Shatin, New Territories, Hong Kong Special Administrative Region, Hong Kong3 Department of Biomedical Engineering, Boston University, MA 02215, USA4 Center for Advanced Biotechnology, Boston University, MA 02215, USA5 Center for Advanced Genomic Technologies, Boston University, MA 02215, USA6 SEQUENOM, Inc., San Diego, CA 92121-1331, USA2005 25 7 2005 6 102 102 28 3 2005 25 7 2005 Copyright © 2005 Rachlin et al; licensee BioMed Central Ltd.2005Rachlin et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Multiplex PCR is a key technology for detecting infectious microorganisms, whole-genome sequencing, forensic analysis, and for enabling flexible yet low-cost genotyping. However, the design of a multiplex PCR assays requires the consideration of multiple competing objectives and physical constraints, and extensive computational analysis must be performed in order to identify the possible formation of primer-dimers that can negatively impact product yield.
Results
This paper examines the computational design limits of multiplex PCR in the context of SNP genotyping and examines tradeoffs associated with several key design factors including multiplexing level (the number of primer pairs per tube), coverage (the % of SNP whose associated primers are actually assigned to one of several available tube), and tube-size uniformity. We also examine how design performance depends on the total number of available SNPs from which to choose, and primer stringency criterial. We show that finding high-multiplexing/high-coverage designs is subject to a computational phase transition, becoming dramatically more difficult when the probability of primer pair interaction exceeds a critical threshold. The precise location of this critical transition point depends on the number of available SNPs and the level of multiplexing required. We also demonstrate how coverage performance is impacted by the number of available snps, primer selection criteria, and target multiplexing levels.
Conclusion
The presence of a phase transition suggests limits to scaling Multiplex PCR performance for high-throughput genomics applications. Achieving broad SNP coverage rapidly transitions from being very easy to very hard as the target multiplexing level (# of primer pairs per tube) increases. The onset of a phase transition can be "delayed" by having a larger pool of SNPs, or loosening primer selection constraints so as to increase the number of candidate primer pairs per SNP, though the latter may produce other adverse effects. The resulting design performance tradeoffs define a benchmark that can serve as the basis for comparing competing multiplex PCR design optimization algorithms and can also provide general rules-of-thumb to experimentalists seeking to understand the performance limits of standard multiplex PCR.
==== Body
Background
The PCR (Polymerase Chain Reaction) method of DNA amplification has had a profound impact on biotechnology and biological research. Multiplex PCR is an extension of the standard PCR protocol in which multiple loci are amplified simultaneously in order to save time, improve throughput, and reduce the total cost of reagents. Applications for PCR and Multiplex PCR abound including quantitative gene expression [1-4], haplotyping [5], whole-genome closure [6,7], detection of genetically modified organisms [8], forensic analysis, including human identification and paternity testing [9,10] diagnosis of infectious diseases [11,12], and for anti-bioterror applications aimed at detecting biological agents such as Anthrax [13]
Multiplex PCR has recently emerged as a core enabling technology for high-throughput SNP genotyping [14-16], and variations on the standard protocol are being actively explored and in some cases more widely commercialized. It is in this context of genotyping that we focus our discussion of multiplex PCR assay design. Thus we will typically refer to multiplexing SNPs (rather than primers) but our treatment is readily applicable to most other PCR applications. Genomic variations in the form of Single Nucleotide Polymorphisms (SNPs) and associated haplotypes continue to garner tremendous interest particularly in the context of pharmacogenomic initiatives aimed at understanding the connection between individual genetic traits, drug response, and disease susceptibility [17-21]. Broad adaptation of genotyping technologies in clinical settings will depend on their cost and inherent clinical value and may be significantly impacted by ethical and legal considerations. Recent technological developments in PCR-based genotyping based on primer extension with universal PCR primers [22] have demonstrated very high (~100-plex) multiplexing levels, although the use of common primers does introduce other issues including the greater potential for cross-contamination.
Multiplex PCR assay design is a multi-objective optimization problem involving intrinsic performance tradeoffs. The key objectives we consider in this paper include the number of SNPs per tube (multiplex level) and the percentage of SNPs assigned to full tubes (coverage). We further require that all resulting tubes achieve uniform levels of multiplexing with the idea that doing so facilitates automation in a high-throughput environment. While lower coverage may be acceptable in initial survey studies involving many (104-106) SNPs, achieving high (>95%) coverage becomes obviously more important when the focus of investigation has been narrowed to a relatively small (102-103) set of SNPs each of which is suspected of having some biological or pharmacological impact.
The question we address in this paper is whether there are fundamental limitations to our ability to design assays that achieve multiplexing levels of arbitrary size using standard multiplex PCR protocols. While multiplex PCR is an established technique, its usefulness as the basis for future high-throughput platforms depends critically on scalability. We introduce a new framework of "multi-node graphs" to model the multiplex PCR problem. We show that the problem of finding high-multiplexing/high-coverage designs is subject to a computational phase transition, becoming dramatically more difficult when the probability that two primers are mutually compatible drops below a critical threshold. This probability depends on fundamental primer selection criteria typically selected to avoid the formation of primer dimers. For standard criteria, we can identify where such a transition occurs, and show that it is consistent with typical multiplexing levels. The precise location of this critical transition point will also depend on N, the number of available SNPs. For a given level of coverage, the level of achievable multiplex is proportional to log(N). We further quantify design performance tradeoffs using two best-fit tube assignment algorithms on human SNP data.
Results
Phase transitions in multiplex PCR complexity
Our first main result reported in this section can be succinctly stated as follows: for an assay with N SNPs and approximately S candidate primers for each SNP we devise a relatively efficient algorithm that can achieve almost perfect coverage with tubes of size approximately O(log NS). Unfortunately the coverage drops dramatically if the multiplex level is increased. (See [23] for a formal analysis.) This result is similar in spirit (but not in detail) to related observations made about other graph problems.
In recent years, it has been shown that for broad classes of computationally intractable problems, there exist certain phase-transition boundaries across which the nature of the solutions and the computational effort needed to identify a solution changes dramatically [24]. When attempting to design multiplex PCR assays with high coverage, we observe a similar computational behavior on simulations using a novel graph formulation we call a multi-node graph (see Methods section). This graph representation consists of nodes representing SNPs and edges connecting two multiplex-compatible SNPs. Two SNPs are multiplex compatible if none of their associated primers interact. To model the fact that SNP compatibility depends on the assigned primers, we allow for the presence of an edge matrix Euv. In a multi-node graph, Euv[i][j] = 1 when node u with primer set i is compatible to node v with primer set j. We induced a random multi-node graph by setting Euv[i][j] = 1 with probability P for all node pairs u and v, in states i and j respectively. Using a simple greedy algorithm (see Methods section) we find that our ability to achieve high (>95%) coverage for randomly generated multi-node graphs critically depends on the compatibility probability, P, (or conversely the interaction probability (1-P)) as well as the target level of multiplexing. These results are presented in Figure 1. By sampling from chromosome 21 of the human genome, the actual probability that two SNPs are compatible is approximately 0.299. Figure 1 would suggest, therefore, that designing 10-plex assays from N = 1,200 SNPs is generally straightforward, but that increasing multiplex performance to 15- to 20-plex or beyond becomes extremely problematic. This appears to be consistent with current design practice though we emphasize that the location of the phase transition depends on both the total number of SNPs and the number of candidate primer pairs per SNP. We recognize, furthermore, that random multi-node graphs only approximately model the multiplex assay design problem because primer pair candidates derived from real sequence data are not truly independent. For example, primer pairs may share a forward or reverse primer, or they may significantly overlap. In addition, in the process of assay design optimization, primers within a single tube may take on certain sequence characteristics (e.g., high GT / GA / CT / CA content) that are intrinsically less likely to interact, and thus make higher-than-expected coverage possible for a given multiplexing target.
Figure 1 Phase transition in full-tube coverage as a function of SNP-SNP compatibility probability. These results are based on a simulations where the controlling parameter P denotes the probability that two SNPs are compatible. Two SNPs are compatible if their associated primers are all pair-wise compatible. This simulation is based on N = 1,200 SNPS and S = 500 primer pairs per SNP. In reality, this compatibility probability, P, depends on the stringency by which primer pairs are tested for cross-interactions. As we increase the target multiplexing level, higher compatibility, beyond what are normally obtained using standard primer selection criteria is required, suggesting fundamental barriers to increasing target multiplexing levels.
Multiplex PCR performance on human SNP data
Next we implemented two multiplex PCR assay design algorithms and applied them to real SNP data obtained from the dbSNP database. We prescreened the 84,393 chromosome 21 SNPs contained in build 116 of dbSNP [25] for class 1 SNPs (strict single nucleotide polymorphisms) containing at least 200 bases of non-low-complexity sequence both upstream and downstream from the target SNP. This reduced our working set to 18,498 SNPs, 21.9% of the original total, from which 1,200 SNPs were randomly selected for experimental purposes. The GC content of the 401-base flanking sequence surrounding (including the SNPs themselves) was 41.9% +/- 11.0% in line with a 41% GC content for chromosome 21 and for the human genome as a whole [26].
Our two best-fit greedy algorithms are designed to simultaneously assign primer candidates to SNPs and to partition SNPs into individual tubes in an effort to maximize both multiplexing level and coverage. See Methods for complete details. While best fit algorithms are relatively simple, one can actually show theoretically that these results appear to be as good as expected (on average) in graphs with this level of density. One version which we call "Fixed-Assignment Best Fit" assigns SNPs in random order to the fullest compatible tube, and as the name suggests, once a SNP is assigned to a tube it is fixed. Neither its assigned tube nor its associated primers are ever modified. If no compatible tube can be found, the SNP is left unassigned, reducing total coverage. We considered a second variation on the best-fit approach called "Flexible Assignment Best Fit" in which SNPs already assigned to a tube can be removed under special conditions in order to accommodate a new SNP assignment. Special rules of the algorithm guarantee that the algorithm will eventually terminate with increasingly high probability. Figure 2 demonstrates the precise nature of the tradeoff between multiplexing and SNP coverage for a fixed number of SNPs (N = 1,200) for both best fit methods.
Figure 2 Coverage vs. target multiplex level using two different best-fit tube assignment strategies. These results were all based on N = 1,200 for varying target multiplexing level M. In each trial, the number of allowed tubes is limited to . Full-tube coverage, the percentage of SNPs assigned to full tubes, of close to 80% is achieved at a multiplexing level of 20, though it drops rapidly for higher multiplexing levels. The graph shows a significant improvement in one algorithm over the other, demonstrating that such tradeoffs can be used to effectively compare and contrast competing optimization strategies.
Multiplex PCR coverage performance tradeoffs
Next we employed the fixed-assignment best-fit algorithm to generate coverage curves for target multiplexing levels M = 10, 20, 30 while varying numbers of SNPs. We considered SNP sets containing between 100 and 1200 SNPs. Figure 3 presents our results. With 200 SNPs, 80% coverage could be achieved with 10-plex assays, but this drops to 40% coverage using 20-plex assays. However, if we increase the number of SNPs to 1200, then for 20-plex assays, coverage increases from approximately 40% to 80%. This graph shows that regardless of the multiplexing level desired, coverage increases with the number of SNPs but with diminishing returns. More precisely, for fixed multiplexing level M, coverage is roughly proportional to log(N).
Figure 3 Multiplex PCR performance tradeoffs. A closer examination of the Fixed Assignment Best-Fit algorithm reveals tradeoffs between the available number of SNPs, N, the target multiplexing level, M, and full-tube coverage. The dip at N = 1000, M = 30 is an artifact of the algorithm which strictly limits the number of tubes to = 34 tubes. Since M does not divide N evenly, the algorithm ends up partially filling the excess tube rather than working harder to fill the remaining 33 tubes to full 30-plex capacity.
Discussion
There is extensive literature on the general principles of PCR primer design [27-31]. This work has led to a number of software applications, most notably Primer3 and various extensions [32-34]. A fast dynamic programming formulation for testing primers for pair-wise compatibility has also been developed [35].
The application of Multiplex PCR has increased steadily over the past decade, requiring more sophisticated primer selection protocols. Different algorithms may favor particular objectives, or may be designed for particular technology platforms. In general, the problem of identifying primer pairs to maximize the multiplexing level of a single assay has been shown to be NP-complete by Nicodeme and Steyaert [36] who also present an approximation algorithm that eliminates 3' base complementarity while addressing product size constraints. They also consider electrophoresis distance constraints that require two amplicons to have some minimum length difference so that they can be distinguished after being processed through an electrophoresis gel. Additionally, SNP detection methods based on primer extension protocols in conjunction with mass spectrometry must take into account the resolution of the mass spectrometer as for example with the matching algorithms presented by Aumann, Manisterski, and Yakhini [37].
Whereas this paper focuses on the dual problem of assigning primers and partitioning SNPs into multiplex-compatible tubes, an entirely different multiplex PCR problem is concerned with finding a minimal number of primers necessary to amplify a maximum number of targets over a single experiment [38] or over multiple experiments [39].
Our best fit approach is motivated by the theoretical analysis provided by Alon and Furedi [40] who show that a greedy algorithm in standard graphs produces an independent set of size log N, and moreover this approach can be extended to produce a full cover of the graph. The multi-node graph is, in practice, substantially more complex to cover, however theoretical analysis suggests that the behavior is similar to standard graphs. The sketch of the proof is as follows. Formal details are provided elsewhere [23].
1. For a multi-node graph with N nodes and S states per node, we create a corresponding standard graph with NS nodes. (Each state in the multi-node graph is a unique node.)
2. We add random edges with probability P getting O(N2 S2 P) edges. Then we remove all the edges between nodes that are connecting representatives of the same node. The total number of edges removed is N S2 P. This means independently of the number S of representatives per node we remove roughly 1/N of the total number of edges.
3. If 1/N << P then this removal does not greatly effect the resulting graph and the probability that their exists a clique of size K on a graph of size NS applies to a multi-node graph of size N with S representatives per node.
Conclusion
In this paper, we quantified some of the critical tradeoffs involved in the multi-objective design of multiplex PCR assays and demonstrated a phase transition suggesting that achieving high-coverage designs becomes dramatically more difficult when SNP compatibility probabilities fall below a certain critical threshold. Explicit consideration of tradeoffs in multiplex PCR design is useful in helping researchers to design effective and reliable assays within the computational limits of the problem. Furthermore, such tradeoffs provide a natural basis for comparing and contrasting novel computational techniques aimed at improving one or more objectives. Although we have attempted to rely on standard design criteria, further laboratory testing is required to validate the design criteria used as the basis of this analysis. In the future we will seek to further improve our current tradeoff benchmarks with the development of novel algorithms and to apply our techniques to the design of high-multiplexing assays that achieve broad coverage of the complete human genome. We have also developed a web-enabled Multiplex PCR assay design system known as MuPlex [41] that also serves as a testing ground for on-going algorithmic development.
Methods
Multi-node graphs: a novel formulation for the multiplex PCR problem
Designing one or more multiplex assays for SNPs with preselected primers is equivalent to finding a clique in a graph G where nodes are SNPS and edges connect pairwise multiplex compatible SNPs, i.e., two SNPs whose primers can be pooled with a single tube without forming primer dimers. Equivalently, we can model the problem using the inverse graph whose edges denote non-compatible or interacting SNPs (i.e., SNPs whose associated primers incur at least one interaction) in which case the objective is to identify an independent set rather than a clique. Theoretical bounds for covering a graph with disjoint cliques can be found in [42].
The multiplex PCR problem is more general in that we impact the graph topology by choice of primers. We use the term "multi-node graph" to denote a graph whose nodes have multiple states. In a multi-node graph, an edge matrix Euv is attached to each pair of nodes, (u, v). If node u (in state i) is multiplex compatible with node v (in state j) then Euv[i][j]= 1. Otherwise, Euv[i][j] = 0. As illustrated in Figure 4, nodes W, X, and Y are pair-wise compatible when in certain states and incompatible in other states. The nodes W, X, Y form a 3-clique (3-plex) when in states 7, 2, and 4 respectively. The multiplex PCR design problem is equivalent to choosing a state assignment to each node in a graph to achieve maximal covering (including as many nodes as possible) with disjoint cliques of size M in a multi-node graph.
Figure 4 Multi-node graphs. A multi-node graph is a convenient way of formalizing the multiplex PCR problem. In multi-node graphs, individual nodes can take on one or more states. In this figure, an edge between two nodes, X and Y, is determined by the state of the two nodes, or more specifically, an edge matrix EXY connecting nodes X and Y. There is no restriction on the number of states per node, and each node may contain a different number of states.
In the context of multiplex PCR, our goal is to identify a set of uniformly sized disjoint cliques in a multi-node graph. This involves the dual problem of selecting node states (primer pair selections for each SNP) and identification of the cliques themselves corresponding to multiplex-compatible SNP sets.
Selecting candidate primers
We generated candidate forward and reverse primer pair candidates for each of the 1,200 SNPs according to the selection criteria listed in Table 1. In addition, if more than one valid primer shared a given 3' position, all but the shortest was automatically discarded as redundant. We employed two separate tests for primer-primer interaction, one based on a standard local alignment to detect stretches of complementary sequence, the other based on the worst-case ΔG of the 3'-tail of one primer interacting somewhere along the strand of another primer. These interaction criteria were used for screening individual primers as well as forward-reverse primer pairs, and for determining compatibility of two SNPs within a single multiplex PCR assay. Primer selection stringency is obviously a critical factor impacting the performance of any multiplex PCR design process. Fewer primers more carefully chosen may be more likely to produce a working assay but could undermine one's ability to identify high-multiplex designs. We have attempted to select primers based on a number of commonly employed selection criteria, recognizing that our overall performance would be directly impacted by any particular selection criteria.
Table 1 Primer design selection criteria. These criteria are used, where applicable, for determining the compatibility of forward and reverse primers within a given locus and for pair-wise compatibility between primers for different loci.
Parameter Allowed Range
Length 17 – 24 bases
% GC 35 – 65%
Tm (nearest neighbor) 57.0 – 70.0 C
Tm Difference 3.0 C (for both candidate primer pairs and for all primers within a particular multiplex tube)
Base repeats ≤ 3 bases maximum
Product Size 60 – 200 bases
Distance to SNP 177 bases (5' end) 5 bases (3' end)
Self complementarity local alignment score ≤ 8.0 (match = 1.0, mismatch = 1.0, gap = -2.0)
3'-Tail alignment ΔG ≥ -4.5 kCal/mol
Using the above primer selection criteria, we generated an average of 1555.8 +/- 1249.4 primer pair candidates per SNP. Twenty-two of our 1,200 SNPs (1.8%) produced no valid primer candidates. For each SNP we randomly selected a primer pair and constructed the corresponding compatibility graph, where edges connect two compatible SNPs. The graph density was 0.299 +/- 0.005 over 10 random trials. As noted earlier, a random graph model is an imprecise representation for the multiplex PCR primer design problem because primer pair candidates are not independent, and because the primers for SNPs within a tube tend to become dominated by one of four non-interacting base pair combinations (G-T, G-A, C-T, or C-A.)
To further understand how SNP-pair compatibility probabilities depend on primer selection criteria, we randomized these criteria and evaluated the resulting graph density for 50 SNPs randomly chosen from human chromosome 21. While certain oligo-specific parameters were kept fixed, pair-wise compatibility criteria were chosen at random within certain specified ranges. The specific compatibility criteria we considered were:
• Complementary Sequence Local Alignment Score: Allowed to vary between +4 and +10 (assuming base match = +1, mismatch = -1, gap = -2). Typically, a threshold of +8.0 is used. A smaller scoring threshold is more stringent as we are disallowing primer pairs with less complementary sequence.
• 3' Tail ΔG Alignment Score: This score measures the worst-case alignment of the 3' tail of one primer along any other part of another sequence. Current practice suggests that it is the 3' tail that is most critical to ensuring proper primer ligation. We allowed this threshold to range from 0.0 to -9.0 kCal/mol. (-4.0 to -6.0 kCal / mol is probably reasonable, although this scoring method is not widely used.) A higher (less negative) cutoff is more stringent as we are disallowing less-energetically favorable interactions.
• Melting Temperature (Tm) Range: The allowed melting temperature between any two primer pairs. A Tm difference of 3.0 degrees K is fairly standard. Here we allow the Tm difference to vary between 0.5 and 7.5 degrees K. (A smaller Tm difference is more stringent.)
The random selection of thresholds and the resulting generation of valid primers constituted a single trial. Within each trial, we randomly selected one primer pair for each locus and computed the resulting density of a graph where nodes represent particular SNPs having an assigned primer pair, and edges connect two multiplex-compatible SNPs. For two SNPs to be mutually compatible, all four primers must be pair-wise compatible using the selected thresholds. These thresholds are thus used to screen both intra- and inter-SNP primer pairs. The graph density was computed as the average of 10 densities each resulting from 10 random primer selection rounds. Figure 5 shows the relationship between graph density and the 3' tail interaction threshold.
Figure 5 Primer-primer compatibility probability and primer selection stringency. This figure shows how a number of primer selection criteria impact the overall probability that two primers will be mutually compatible. If compatible primers are connected by edges in a graph, the resulting probability is equivalent to the graph density. This figure plots graph density as a function of 3' ΔG interaction. Each point represents a single trial where additional primer compatibility thresholds were randomly chosen within specified ranges. Considered were the 3' tail ΔG interaction, complementary sequence local alignment score, and melting temperature (Tm) difference. Points that are more red allow for high Tm differences while points that are more blue require smaller Tm differences. The impact of local alignment score thresholds, while not shown explicitly, is indirectly revealed by the multiple tiers (bands) across the graph, the lowest corresponding to score = +4 and the highest to score = +6 to +10.
Each point in the graph represents a particular trial. Red points have a high Tm difference threshold while blue points have a low Tm difference threshold. As expected, the resulting compatibility graph density increases as we loosen the constraint on the interaction ΔG. For a given ΔG cutoff, tighter constraints on the Tm difference will naturally tend to reduced SNP pair compatibility. The multiple performance tiers (clearly seen as multiple red bands across the chart) reflect different cutoffs for the standard complementary sequence local alignment score, the lowest being for Score = 4, while at the top, scores of 6–10 blend together in this figure. At ΔG = -4.5 kCal/mol, we expect graph densities no greater than about 30%. In other words, the probability that two SNPs, each assigned random primer pairs, are multiplex compatible is only about 30%.
Algorithms
The fixed-assignment best-fit algorithm
We define a benchmark best-fit strategy for designing uniform M-plexes as follows:
Let GN be a multi-node graph with N nodes (SNPs).
Let M be the desired clique size (multiplexing level).
Let T= [T1, T2, ..., TN/M] cliques (tubes), initially empty.
Let CANTASSIGN be a set of unassignable nodes, initially empty.
1. Choose a node, u∊N and u∉CANTASSIGN at random.
2. Assign a random state (primer pair) to the node
3. Find all cliques in T that are compatible with u. (A clique is compatible if the size of the clique is less than M, and every node in the clique has an edge connected to u according to the appropriate edge matrices. Thus, if u is added to the clique, then it is still a clique. If there are one or more compatible cliques, we assign u to the largest clique, otherwise, we leave u unassigned.
4. For all unassigned nodes, choose a different state, ensuring that no state is chosen more than once. If a particular unassigned node u has no such state, then we add u to CANTASSIGN.
5. Repeat steps 1–4 until there are no more nodes to choose from. The resulting set of cliques, T, define our final solution. We measure coverage as a percentage of the nodes assigned to full tubes.
For N nodes (SNPs) and S states (primer pair candidates) per node, pre-computing and storing all possible primer-pair interactions in a multi-node graph would require O(N2S2) time and space. Therefore, when applying the above algorithm to real-world SNPs with potentially hundreds of candidate primer pairs (states) per SNP (node), we determine primer interactions (edge interactions) as needed in order to test a particular SNP for tube compatibility.
Step 3 provides our definition of "best-fit." The best tube is defined as the largest compatible tube. The rational for this strategy is that while each tube assignment reduces our probability of finding another assignment to the same tube, the reduction in probability is minimized when we make an assignment to the tube that is currently the largest. It should be noted, however, that in practice very little change in the performance of the algorithm is observed if we assign nodes to a random compatible tube, to the smallest compatible tube, or even to the first compatible tube. This is due to the fact that the probability of assigning a particular SNP to a particular tube decreases exponentially as the size of the tube (clique) grows larger. For sufficiently large cliques, it is unlikely that a SNP with a particular assigned primer set will be compatible with any tube. Most often, when a compatible tube is found, it is the only compatible tube among the available choices, and thus there is little or no resulting difference in how we actually choose our tube.
Flexible-assignment best-fit
As noted above, we expect the probability of finding a compatible tube for a particular SNP in a particular state to decrease exponentially as the tube size increases. Suppose in attempting to assign a SNP to a particular tube, we find that it is incompatible with one other SNP which was originally assigned when the tube was small. It is relatively easy to find a compatible tube when they contain few SNPs because there are fewer interactions to consider. By contrast, if we find a nearly-compatible SNP when the tube is large, we should attempt to accommodate the SNP with the idea that low-probability SNP/Tube assignments are relatively rare and should thus be maintained whenever possible.
In general, suppose we have a tube containing k SNPs (a clique with k nodes). We test a SNP for compatibility with the tube and find that d ≤ k SNPs are incompatible with the SNP. Suppose furthermore that these d SNPs were assigned to the tube when the tube was of size k1, k2, ... kd. We claim that it is valuable to substitute our test SNP for the d incompatible SNPs whenever the total probability of assigning the d incompatible SNPs is greater than the probability of assigning the test SNP to a tube that excludes these d SNPs. That is:
If there are multiple tubes satisfying the above condition, we assign the SNP to the tube where is maximized. In assessing these assignment probabilities, additional consideration could be given to the number of candidate primer pairs maintained by each SNP, as it is less likely that we will identify a compatible tube for SNPs with relatively few candidates with the idea that we should be more reluctant to remove such SNPs once a compatible assignment is determined. In this version, however, we considered only the tube sizes at the time each SNP is assigned. In the Fixed-Assignment Best-Fit algorithm, the size of each tube is monotonically increasing. Thus, if a SNP with a given primer candidate is not compatible with any tube, the primer pair candidate can be removed from further consideration and furthermore, if no primer candidate is compatible for any tube, then the SNP can be discarded from further consider. By contrast, the Flexible-Assignment Best-Fit algorithm requires that we reconsider the compatibility of SNPs and their associated primers within any modified tube. We thus specifically track which candidates have been tested on which tubes, and we update this status each time a particular tube is modified.
It is clear that the Flexible Assignment Best-Fit algorithm will eventually terminate with increasingly high probability because every SNP substitution we perform within a particular tube is assessed as having strictly lower probability than previous assignments.
Authors' contributions
John Rachlin and Simon Kasif are responsible for the overall algorithmic design and theoretical analysis presented in this paper. John Rachlin and Simon Kasif introduced the framework of multi-node graphs in consultation with Noga Alon. John Rachlin implemented the algorithms and produced the initial results for both simulated and real human SNP data. John Rachlin discovered the phase transitions during the analysis of the simulated data which were explained by a theoretical analysis developed initially by Noga Alon and Vera Asodi and expanded and reported in [23]. Charles Cantor proposed the problem and provided the overall direction and oversight of the project. Chunming Ding was involved in the overall biotechnology design and also helped identify the representative primer selection criteria used as the basis for the optimization criteria used by the system. All authors contributed to outlining, writing and editing the manuscript with John Rachlin carrying the lion share of the write-up.
Acknowledgements
This work is supported in part by NSF grants DBI-0239435 and ITR-048715 and NHGRI grant #1R33HG002850-01A1. The authors thank Noga Alon and Richard Beigel for many profound insights and suggestions.
==== Refs
Ding C Cantor CR A high-throughput gene expression analysis technique using competitive PCR and matrix-assisted laser desorption ionization time-of-flight MS Proc Natl Acad Sci U S A 2003 100 3059 3064 12624187 10.1073/pnas.0630494100
Ding C Maier E Roscher AA Braun A Cantor CR Simultaneous quantitative and allele-specific expression analysis with real competitive PCR BMC Genet 2004 5 8 15128429 10.1186/1471-2156-5-8
Becker-Andre M Hahlbrock K Absolute mRNA quantification using the polymerase chain reaction (PCR). A novel approach by a PCR aided transcript titration assay (PATTY) Nucleic Acids Res 1989 17 9437 9446 2479917
Gibson UE Heid CA Williams PM A novel method for real time quantitative RT-PCR Genome Res 1996 6 995 1001 8908519
Ding C Cantor CR Direct molecular haplotyping of long-range genomic DNA with M1-PCR Proc Natl Acad Sci U S A 2003 100 7449 7453 12802015 10.1073/pnas.1232475100
Beigel R Alon N Apaydin SM Fortnow L Kasif S An Optimal Multiplex PCR Protocol for Closing Gaps in Whole Genomes RECOMB 2001
Tettelin H Radune D Kasif S Khouri H Salzberg SL Optimized multiplex PCR: efficiently closing a whole-genome shotgun sequencing project Genomics 1999 62 500 507 10644449 10.1006/geno.1999.6048
Jensen Nested Multiplex Polymerase Chain Reaction for the Determination of DNA from Genetically Modified Corn and Soybeans American Biotechnology Laboratory 2004
Inagaki S Yamamoto Y Doi Y Takata T Ishikawa T Imabayashi K Yoshitome K Miyaishi S Ishizu H A new 39-plex analysis method for SNPs including 15 blood group loci Forensic Sci Int 2004 144 45 57 15240020 10.1016/j.forsciint.2004.03.005
Hanson EK Ballantyne J A highly discriminating 21 locus Y-STR "megaplex" system designed to augment the minimal haplotype loci for forensic casework J Forensic Sci 2004 49 40 51 14979342 10.1520/JFS2003209
Elnifro EM Ashshi AM Cooper RJ Klapper PE Multiplex PCR: optimization and application in diagnostic virology Clin Microbiol Rev 2000 13 559 570 11023957 10.1128/CMR.13.4.559-570.2000
Read SJ Kurtz JB Laboratory diagnosis of common viral infections of the central nervous system by using a single multiplex PCR screening assay J Clin Microbiol 1999 37 1352 1355 10203485
Oggioni MR Meacci F Carattoli A Ciervo A Orru G Cassone A Pozzi G Protocol for real-time PCR identification of anthrax spores from nasal swabs after broth enrichment J Clin Microbiol 2002 40 3956 3963 12409358 10.1128/JCM.40.11.3956-3963.2002
Kwok PY Methods for genotyping single nucleotide polymorphisms Annu Rev Genomics Hum Genet 2001 2 235 258 11701650 10.1146/annurev.genom.2.1.235
Shi MM Enabling large-scale pharmacogenetic studies by high-throughput mutation detection and genotyping technologies Clin Chem 2001 47 164 172 11159763
Shi MM Bleavins MR de la iglesia FA Technologies for detecting genetic polymorphisms in pharmacogenomics Mol Diagn 1999 4 343 351 10671645
Sachidanandam R Weissman D Schmidt SC Kakol JM Stein LD Marth G Sherry S Mullikin JC Mortimore BJ Willey DL Hunt SE Cole CG Coggill PC Rice CM Ning Z Rogers J Bentley DR Kwok PY Mardis ER Yeh RT Schultz B Cook L Davenport R Dante M Fulton L Hillier L Waterston RH McPherson JD Gilman B Schaffner S Van Etten WJ Reich D Higgins J Daly MJ Blumenstiel B Baldwin J Stange-Thomann N Zody MC Linton L Lander ES Altshuler D A map of human genome sequence variation containing 1.42 million single nucleotide polymorphisms Nature 2001 409 928 933 11237013 10.1038/35057149
Kammerer S Roth RB Reneland R Marnellos G Hoyal CR Markward NJ Ebner F Kiechle M Schwarz-Boeger U Griffiths LR Ulbrich C Chrobok K Forster G Praetorius GM Meyer P Rehbock J Cantor CR Nelson MR Braun A Large-scale association study identifies ICAM gene region as breast and prostate cancer susceptibility locus Cancer Res 2004 64 8906 8910 15604251
Mooser V Waterworth DM Isenhour T Middleton L Cardiovascular pharmacogenetics in the SNP era J Thromb Haemost 2003 1 1398 1402 12871273 10.1046/j.1538-7836.2003.00272.x
Altshuler D Pollara VJ Cowles CR Van Etten WJ Baldwin J Linton L Lander ES An SNP map of the human genome generated by reduced representation shotgun sequencing Nature 2000 407 513 516 11029002 10.1038/35035083
Gabriel SB Schaffner SF Nguyen H Moore JM Roy J Blumenstiel B Higgins J DeFelice M Lochner A Faggart M Liu-Cordero SN Rotimi C Adeyemo A Cooper R Ward R Lander ES Daly MJ Altshuler D The structure of haplotype blocks in the human genome Science 2002 296 2225 2229 12029063 10.1126/science.1069424
Oliphant A Barker DL Stuelpnagel JR Chee MS BeadArray technology: enabling an accurate, cost-effective approach to high-throughput genotyping Biotechniques 2002 Suppl 56 8, 60-1 12083399
Alon N Asodi V Beigel. R Cantor C Kasif S Rachlin J Multi-node Graphs and their Applications In preparation
Hogg T. HBWC Phase Transitions and the Search Problem Artificial Intelligence 1996 81 1 15 10.1016/0004-3702(95)00044-5
Sherry ST Ward MH Kholodov M Baker J Phan L Smigielski EM Sirotkin K dbSNP: the NCBI database of genetic variation Nucleic Acids Res 2001 29 308 311 11125122 10.1093/nar/29.1.308
Venter JC Adams MD Myers EW Li PW Mural RJ Sutton GG Smith HO Yandell M Evans CA Holt RA Gocayne JD Amanatides P Ballew RM Huson DH Wortman JR Zhang Q Kodira CD Zheng XH Chen L Skupski M Subramanian G Thomas PD Zhang J Gabor Miklos GL Nelson C Broder S Clark AG Nadeau J McKusick VA Zinder N Levine AJ Roberts RJ Simon M Slayman C Hunkapiller M Bolanos R Delcher A Dew I Fasulo D Flanigan M Florea L Halpern A Hannenhalli S Kravitz S Levy S Mobarry C Reinert K Remington K Abu-Threideh J Beasley E Biddick K Bonazzi V Brandon R Cargill M Chandramouliswaran I Charlab R Chaturvedi K Deng Z Di Francesco V Dunn P Eilbeck K Evangelista C Gabrielian AE Gan W Ge W Gong F Gu Z Guan P Heiman TJ Higgins ME Ji RR Ke Z Ketchum KA Lai Z Lei Y Li Z Li J Liang Y Lin X Lu F Merkulov GV Milshina N Moore HM Naik AK Narayan VA Neelam B Nusskern D Rusch DB Salzberg S Shao W Shue B Sun J Wang Z Wang A Wang X Wang J Wei M Wides R Xiao C Yan C Yao A Ye J Zhan M Zhang W Zhang H Zhao Q Zheng L Zhong F Zhong W Zhu S Zhao S Gilbert D Baumhueter S Spier G Carter C Cravchik A Woodage T Ali F An H Awe A Baldwin D Baden H Barnstead M Barrow I Beeson K Busam D Carver A Center A Cheng ML Curry L Danaher S Davenport L Desilets R Dietz S Dodson K Doup L Ferriera S Garg N Gluecksmann A Hart B Haynes J Haynes C Heiner C Hladun S Hostin D Houck J Howland T Ibegwam C Johnson J Kalush F Kline L Koduru S Love A Mann F May D McCawley S McIntosh T McMullen I Moy M Moy L Murphy B Nelson K Pfannkoch C Pratts E Puri V Qureshi H Reardon M Rodriguez R Rogers YH Romblad D Ruhfel B Scott R Sitter C Smallwood M Stewart E Strong R Suh E Thomas R Tint NN Tse S Vech C Wang G Wetter J Williams S Williams M Windsor S Winn-Deen E Wolfe K Zaveri J Zaveri K Abril JF Guigo R Campbell MJ Sjolander KV Karlak B Kejariwal A Mi H Lazareva B Hatton T Narechania A Diemer K Muruganujan A Guo N Sato S Bafna V Istrail S Lippert R Schwartz R Walenz B Yooseph S Allen D Basu A Baxendale J Blick L Caminha M Carnes-Stine J Caulk P Chiang YH Coyne M Dahlke C Mays A Dombroski M Donnelly M Ely D Esparham S Fosler C Gire H Glanowski S Glasser K Glodek A Gorokhov M Graham K Gropman B Harris M Heil J Henderson S Hoover J Jennings D Jordan C Jordan J Kasha J Kagan L Kraft C Levitsky A Lewis M Liu X Lopez J Ma D Majoros W McDaniel J Murphy S Newman M Nguyen T Nguyen N Nodell M Pan S Peck J Peterson M Rowe W Sanders R Scott J Simpson M Smith T Sprague A Stockwell T Turner R Venter E Wang M Wen M Wu D Wu M Xia A Zandieh A Zhu X The sequence of the human genome Science 2001 291 1304 1351 11181995 10.1126/science.1058040
Chen BY Janes HW Chen S Computer programs for PCR primer design and analysis Methods Mol Biol 2002 192 19 29 12494633
Dieffenbach CW Lowe TM Dveksler GS General concepts for PCR primer design PCR Methods Appl 1993 3 S30 7 8118394
Gorelenkov V Antipov A Lejnine S Daraselia N Yuryev A Set of novel tools for PCR primer design Biotechniques 2001 31 1326 1330 11768662
Hyndman DL Mitsuhashi M PCR primer design Methods Mol Biol 2003 226 81 88 12958488
Vieux EF Kwok PY Miller RD Primer design for PCR and sequencing in high-throughput analysis of SNPs Biotechniques 2002 Suppl 28 30, 32 12083394
Rozen S Skaletsky H Primer3 on the WWW for general users and for biologist programmers Methods Mol Biol 2000 132 365 386 10547847
van Baren MJ Heutink P The PCR suite Bioinformatics 2004 20 591 593 14751986 10.1093/bioinformatics/btg473
Weckx S De Rijk P Van Broeckhoven C Del-Favero J SNPbox: a modular software package for large-scale primer design Bioinformatics 2005 21 385 387 15347573 10.1093/bioinformatics/bti006
Kampke T Kieninger M Mecklenburg M Efficient primer design algorithms Bioinformatics 2001 17 214 225 11294787 10.1093/bioinformatics/17.3.214
Nicodeme P Steyaert JM Selecting optimal oligonucleotide primers for multiplex PCR Proc Int Conf Intell Syst Mol Biol 1997 5 210 213 9322038
Aumann Y Manisterski E Yakhini Z Designing Optimally Multiplexed SNP Genotyping Assays Proc Algorithms in Bioinformatics : Third International Workshop (WABI) 2003 Springer-Verlag 320 338
Doi K Imai H Greedy Algorithms for Finding a Small Set of Primers Satisfying Cover and Length Resolution Conditions in PCR Experiments Genome Inform Ser Workshop Genome Inform 1997 8 43 52 11072304
Doi K Imai H A Greedy Algorithm for Minimizing the Number of Primers in Multiple PCR Experiments Genome Inform Ser Workshop Genome Inform 1999 10 73 82 11072344
Alon N Furedi Z Spanning subgraphs of random graphs Graphs and Combinatorics 1992 8 91 94 10.1007/BF01271712
Rachlin J Ding C Cantor C Kasif S MuPlex: multi-objective multiplex PCR assay design Nucleic Acids Res 2005 33 W544 W547 15980531 10.1093/nar/gki377
Alon N Spencer JH The probabilistic method Wiley-Interscience series in discrete mathematics and optimization 2000 2nd New York , Wiley xvi, 301 p.
|
16042802
|
PMC1190169
|
CC BY
|
2021-01-04 16:32:49
|
no
|
BMC Genomics. 2005 Jul 25; 6:102
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-102
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-981601181010.1186/1471-2164-6-98Research ArticleMuscle regeneration in dystrophin-deficient mdx mice studied by gene expression profiling Turk R [email protected] E [email protected] Meijer EJ [email protected] Ommen G-JB [email protected] Dunnen JT [email protected]'t Hoen PAC [email protected] Center for Human and Clinical Genetics, Leiden University Medical Center, Wassenaarseweg 72, 2333 AL Leiden, Nederland2 Leiden Genome Technology Center, Leiden University Medical Center, Wassenaarseweg 72, 2333 AL Leiden, Nederland3 Department of Physiology and Biophysics, Howard Hughes Medical Institute, University of Iowa, 400 Eckstein Medical Research Building, Iowa City, IA52240-1101, U.S.A2005 13 7 2005 6 98 98 8 2 2005 13 7 2005 Copyright © 2005 Turk et al; licensee BioMed Central Ltd.2005Turk et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Duchenne muscular dystrophy (DMD), caused by mutations in the dystrophin gene, is lethal. In contrast, dystrophin-deficient mdx mice recover due to effective regeneration of affected muscle tissue. To characterize the molecular processes associated with regeneration, we compared gene expression levels in hindlimb muscle tissue of mdx and control mice at 9 timepoints, ranging from 1–20 weeks of age.
Results
Out of 7776 genes, 1735 were differentially expressed between mdx and control muscle at at least one timepoint (p < 0.05 after Bonferroni correction). We found that genes coding for components of the dystrophin-associated glycoprotein complex are generally downregulated in the mdx mouse. Based on functional characteristics such as membrane localization, signal transduction, and transcriptional activation, 166 differentially expressed genes with possible functions in regeneration were analyzed in more detail. The majority of these genes peak at the age of 8 weeks, where the regeneration activity is maximal. The following pathways are activated, as shown by upregulation of multiple members per signalling pathway: the Notch-Delta pathway that plays a role in the activation of satellite cells, and the Bmp15 and Neuregulin 3 signalling pathways that may regulate proliferation and differentiation of satellite cells. In DMD patients, only few of the identified regeneration-associated genes were found activated, indicating less efficient regeneration processes in humans.
Conclusion
Based on the observed expression profiles, we describe a model for muscle regeneration in mdx mice, which may provide new leads for development of DMD therapies based on the improvement of muscle regeneration efficacy.
==== Body
Background
Duchenne muscular dystrophy (DMD) is caused by mutations in the gene encoding dystrophin, a subsarcolemmal protein functioning within the dystrophin-associated glycoprotein complex (DGC)[1,2]. This complex connects the intracellular cytoskeleton to the extracellular matrix. The DGC is concentrated at the Z-lines of the sarcomere and confers the transmission of force across the muscle fibre[3]. Disruption of this link results in membrane instability, which eventually leads to sarcolemmal ruptures[4,5]. Influx of extracellular calcium alters molecular processes like muscle contraction and activates proteolytic activity. Affected muscle fibres become necrotic or apoptotic, and release mitogenic chemoattractants, which initiate inflammatory processes [6-8]. Cycles of degeneration and regeneration eventually lead to irreversible muscle wasting and replacement by fibrotic and adipose tissue.
Muscle has the potential to regenerate by activation of undifferentiated myogenic precursor cells (satellite cells), which are normally quiescent and situated between the basal membrane and the myofibers[9,10]. Upon activation, satellite cells proliferate and divide asymmetrically, with the daughter cells having divergent cell fates[11]. Only one of the daughter cells differentiates, progresses towards the myoblast-stadium, and subsequently fuses with other myoblasts or with damaged muscle fibres to induce muscle fibre repair. The other daughter cell remains in a proliferating state or returns to quiescence[12]. Genetic mutations responsible for DMD are also present in satellite cells. Hence, the ability to restore normal muscle function remains obstructed. A small number of muscle fibres are able to produce functional dystrophin, mostly due to secondary mutations in myogenic precursor cells which restore the reading frame[13]. However, these so-called revertant fibres are in a too small minority to alleviate the pathology of the dystrophin-deficiency. Exhaustion of the satellite cell pool due to degeneration and regeneration cycles is thought to critically contribute to the disease[14].
The mdx mouse model for DMD has a spontaneous mutation in exon 23 of the Dmd gene, introducing a premature stopcodon[15,16]. The pathology of the mdx mouse is characterized by histologically well-defined stages with similarity to the human pathology. Neonatal muscle tissue appears to be unaffected. Necrotic or apoptotic processes in combination with inflammation emerge at approximately 3 weeks of age[15]. Regeneration processes are initiated around the age of 6 weeks and continue while alternating with ongoing degeneration until 12 weeks of age [17-19]. Contrary to the lethal human pathology, the mdx mouse somehow recovers from the progressive muscle wasting, and does not show the accumulation of connective and adipose tissue[17,20]. However, mdx mice do show a decline in their regeneration capacity at advanced age (>65 weeks), while necrotic processes persist[21]. Since the degeneration processes are similar to those seen in human pathology, the regenerational differences may hold one of the clues of restoration of proper muscle function.
Although previous studies have studied gene expression levels in the mdx mouse [22-27], regeneration processes were not studied in full detail. We studied the regeneration process through genome-wide monitoring of gene expression levels[28] in healthy control and mdx mice at 9 time points from 1 to 20 weeks of age, while putting emphasis on time points where regenerative activity is maximal (6–12 weeks), a period which was not analysed in detail in a previous time course study[23]. According to the temporal gene expression profiles, we determined which pathways are active during regeneration with respect to normal muscle aging. The majority of identified genes presented in this study have not been described before and provide a substantial addition to the elucidation of the temporal phasing of degeneration and regeneration. By careful annotation based on existing literature, new light is shed on the pathology and subsequent recovery in the mdx mouse. Furthermore, we compared gene expression profiles to those of human DMD patients and found only modest overlap in regeneration-associated genes. This confirms that regeneration is no longer an active process at the age at which the patients were profiled (5–12 years old).
Results and discussion
Global comparison of mdx and control mice
Gene expression levels were determined in hindlimb muscle tissue from mdx and control mice at 9 time points, ranging from 1 to 20 weeks. Differential gene expression levels were calculated per time point by subtraction of the average normalized intensities of control samples from those of the mdx samples to correct for normal aging processes. The effects of the normal aging processes on gene expression are relatively minor, and are discussed below. Statistical significance was calculated per time point by performing a Student's t-test. Differential gene expression was considered significant when p-values were lower than 0.05 after applying a Bonferroni correction for multiple testing (p ≤ 6.43 × 10-6). Out of 7,776 temporal gene expression profiles 1,735 were selected, which satisfied the significance criterion at one or more time points [see Additional file 1].
The number of differentially expressed genes per time point changes considerably during the time course, an effect also shown in a previous study by Porter et al.[27] (Figure 1). The number of differentially expressed genes peaks at the age of 8–12 weeks, coinciding with the period of maximal muscle regeneration. Interestingly, also at the first two time points (1 and 2.5 weeks of age), where the histology of the mdx muscle is not different from that of control mice, a large number of genes was differentially expressed, indicating differences in muscle development in dystrophin-deficient animals. The majority of these genes (553/677 and 355/407, respectively) also show statistically significant differences in expression at later time points. This overlap can be explained by the assumption that the repertoire of gene products used for muscle growth and development also functions in muscle regeneration.
Figure 1 Amount of differentially expressed genes. The number of statistically significantly differentially expressed genes between mdx and control mice measured across 9 consecutive timepoints from 1 to 20 weeks in this study (continuous line) are compared to the number of differentially expressed genes found in the study of Porter et al.[27] (dashed line).
In this report we will describe the expression changes of two main categories in more detail: genes coding for proteins within the costamer and the dystrophin-associated glycoprotein complex (DGC), and genes involved in regeneration.
The Dystrophin-Glycoprotein Complex
The effect of dystrophin-deficiency on expression levels of dystrophin-glycoprotein complex (DGC)-related genes, or genes with associated functional relevance within the costamer has not been reported in previous gene expression profiling studies [22-26], with the exception of the study of Porter and co-workers[27]. In the Porter study, a downregulation of dystrophin was reported, but no changes in gene expression of other components of the DGC. A selection of 52 genes was made based on an overview by Ervasti et al. of members of the costameric protein network[29] (Additional file 4). According to the statistical selection criteria, 4 genes were upregulated (Figure 2A) and 12 were downregulated (Figure 2B) in the mdx mouse. Although a decrease in dystrophin expression was found in our study, the stringent statistical criteria were not met.
Figure 2 Costameric and DGC related gene expression. Fold change in gene expression levels between mdx and control muscle tissue measured across 9 consecutive timepoints from 1 to 20 weeks of costameric and DGC related genes. Figure 2A shows the fold changes of the statistically significantly upregulated genes over time; Figure 2B shows the downregulated genes.
Upregulated DGC related genes
We find that the retina-specific isoform of dystrophin (Dp260) is expressed in skeletal muscle of mdx mice, whereas Dp260 cannot be detected in hindlimb muscle of control mice. Expression of Dp260 was detected by an oligonucleotide probe within the unique first exon of this transcript. The promoter of the Dp260 isoform resides in intron 29, downstream of the mdx mutation (exon23). Transgenic mdx mice, which overexpress Dp260 via an alpha-actin promoter, show a restoration of a stable association between costameric actin and the sarcolemma, a re-assembly of the DGC, and an overall alleviation of the pathology[30]. Increased transcription initiation of Dp260 might therefore be a natural adaptation for the lack of the muscle specific isoform of dystrophin. However, in contrast to the artificially raised expression by the alpha-actin promoter, the expression of Dp260 in the mdx mouse through the original promoter does not seem to be strong enough to compensate for loss of the full-length muscle specific isoform.
We found a continuous upregulation of alpha-dystrobrevin (Dtna) gene expression with maximum differential expression at 4 weeks in mdx mice. Dystrobrevin is a phosphotyrosine-containing protein localized at both the sarcolemma and the postsynaptic side of the neuromuscular junction (NMJ), where it binds to either dystrophin or utrophin [31-34]. Dtna has been described to function as a signalling mediator within the DGC[34]. Transcription of Dtna is activated when myoblasts differentiate into multinucleated myotubes[35]. Newey et al. reported that Dtna-protein levels are significantly reduced in the mdx mouse at the sarcolemma, whereas the protein level was unchanged at the NMJ. This would be consistent with a stabilizing action of Dtna upon binding to dystrophin or utrophin, since dystrophin is not present at the sarcolemma, whereas utrophin is expressed at the NMJ. They proposed a model, where localized translation of Dtna transcripts contributes to synapse formation[36]. Upregulation of Dtna in the mdx mouse might indicate an attempt to compensate for the increased turnover of the protein, in order to stabilize the post-synaptic side of neuromuscular junctions of affected muscle fibres, and retain neuronal connection.
Our results show a continuous upregulation of LIM domain protein 3 (Ldb3, Cypher/ZASP). Studies in Ldb3 knock-out mice demonstrated that ablation of Ldb3 eradicates the structural integrity of the Z-line in contracting striated muscle and causes a severe form of congenital myopathy[37]. Upregulation of Ldb3 indicates the necessity for stabilization of the Z-line in mdx mice, compensating the undermining effect of dystrophin-deficiency.
Downregulated DGC related genes
It can be seen that gene expression levels of several core-proteins of the DGC, e.g. the transmembrane proteins dystroglycan (Dag1), sarcospan (Sspn), and two members of the sarcoglycan-complex (Sgcd, Sgcg), are lower in mdx mice, over the whole time course. Lower expression levels were also detected in other members of the sarcoglycan complex (Sgca, Sgcb, Sgce) and in dystrophin (Dmd, oligonucleotide at the 3' end), but these were not statistically significant. The decrease in expression of DGC-related genes was most prominent during regeneration (8–12 weeks).
Interestingly, DGC related gene expression levels restore to pre-regeneration levels subsequent to the regeneration period, but remain lower than normal (control) level. Similarly, protein levels of core-proteins of the DGC have been shown to be severely reduced in dystrophin-deficient mdx mice[38]. It is suggested that the secondary displacement of DGC core-proteins is due to a decrease in protein synthesis and/or assembly, or due to an increase in protein degradation. Similarly, in sarcoglycan-deficiencies the absence of a single subunit causes the loss or strong reduction of the entire sarcoglycan protein complex [39-43]. Since our study reveals a downregulation of mRNA levels of the DGC core-proteins, we conclude that alterations in transcriptional activity also contribute to the decrease in protein levels. As transcription of members of the DGC is likely to be co-ordinately regulated[44], downregulation of these members as seen in mdx mice can occur via inhibition or downregulation of shared transcriptional activators.
Regeneration
In the mdx mouse, regeneration of affected muscle tissue is most prominent at the age of 6–12 weeks, after which a stabilized condition is reached. To identify pathways active in regeneration, we studied five categories of differentially expressed genes covering major functional characteristics of regenerative tissues (trophic factors, proteases, membrane associated proteins, signal transduction, and transcription, Additional file 5). This selection of 166 genes was typed for temporal effects during regeneration, and the pathways to which they belong. Since signal transduction pathways are still poorly annotated in current genomic databases, these pathways were constructed from the literature. In our study, a pathway is only considered activated or repressed, when multiple members show differential gene expression.
Temporal effects during regeneration
Differential gene expression profiles, based on the ratio between mdx and control mice, were scaled to the first time point. Differential expression profiles can therefore be compared independent of the ratio level, which enables the detection of temporal effects. Using k-means clustering (k = 6), differential gene expression profiles were classified according to their temporal similarity. The unscaled temporal effects of mdx and control gene expression profiles are shown per cluster for the up- and downregulated genes separately (Figure 3). Genes, which show an upregulation in gene expression during the regenerative phase, are present in clusters 1 (n = 27), cluster 2 (n = 36), and cluster 4 (n = 21). The temporal effect is determined by the gene expression profile of the mdx mouse, since gene expression is continuously low without temporal changes in the control mouse. Downregulation of gene expression in the mdx mouse during regeneration is primarily seen in cluster 3 (n = 23). During normal aging, which can be seen in the control mouse, gene expression increases until the age of 10 weeks, followed by a slow decrease. During the regenerative phase in the mdx mouse, however, the expression of these genes is downregulated markedly.
Figure 3 Temporal effects during regeneration. K-means clustering (k = 6) classifies gene expression profiles according to similarity in temporal patterns based on the scaled differential gene expression levels (grey line). For each cluster the up- and downregulated genes are shown separately. Unscaled differential gene expression levels are shown (black line), which are representative for the ratio between the mdx gene expression levels (dashed line) and control gene expression levels (small dashed line). Relative gene expression levels are obtained after normalization and coincide with the natural logarithm.
Notch-Delta pathway
Gene expression levels of a number of genes functioning in the Notch-Delta pathway are upregulated (Notch1, Notch2, Hr), whereas others (Dxd26, Dvl, Dvl2) are downregulated in the mdx mouse at 8 weeks of age (Figure 4A). The gene expression of Dll3 and Numb are switched on in the mdx mouse, while no gene expression can be detected in the control mouse (Supplemental Table 1 [see Additional file 4]). The differential expression of the upregulated genes is mostly increased during the regeneration period (6 to 12 weeks) (Figure 4A). For several genes in the Notch-Delta pathway, quantitative RT-PCR experiments were performed to confirm the temporal expression profiles found on the microarray. In accordance with the microarray results, quantitative RT-PCR experiments demonstrated higher expression of Notch2, Numb and myogenin in mdx than in control mice at all ages (Figure 5).
Figure 4 Reconstruction of active regeneration pathways. Regeneration-associated pathways were constructed based on differentially expressed genes, literature study and gene ontology. Expression levels of genes in the Notch-Delta pathway (Panel A), the Bmp15 pathway (Panel B), and the Neuregulin3 pathway (Panel C) are plotted as fold-changes between mdx and control mice, as a function of age. In the pathway diagrams, filled boxes refer to upregulation (red) and downregulation (green) at 8 weeks. Outlined boxes refer to upregulation (red) and downregulation (green) at other timepoints than 8 weeks. Shaded grey boxes represent genes which are not detected, or are not represented on the microarray. White boxes represent genes that show no differential expression between mdx and control mice.
Figure 5 Confirmation of microarray data for genes in the Notch-Delta pathway by quantitative RT-PCR. Expression levels of Notch2 (Panel A and B), Numb (Panel C and D), and Myogenin (Panel E and F) in mdx (grey circles) and wild-type (black squares) mice at 1 to 20 weeks of age were measured by quantitative RT-PCR (Panel A, C, E) and expression microarrays (Panel B, D, F). Expression levels relative to those in 1 week-old wild-type mice are plotted on a logarithmic scale (natural logarithm).
Previous work by Conboy et al.[11] indicates the role of the Notch-Delta signalling pathway in the regulation of proliferation versus differentiation of asymmetrically dividing satellite cells by Notch or Numb, respectively. According to Delfini et al.[45], Notch is expressed in immature myoblasts, while Delta (Dll) expressing cells are more advanced in myogenesis (post-mitotic myoblasts and muscle fibres). Notch activation is thought to inhibit transcription factors containing a basic helix loop helix domain (bHLH) [46-50], via the induction of Hairy and Enhancer of Split 1 (Hes1)[51], thereby inhibiting myogenic differentiation. Numb-expressing cells are able to undergo myogenic differentiation, because the Notch-Delta pathway is inhibited (Figure 4A). Based on our results, the Notch-Delta signalling pathway, notably the expression of Notch or Numb, is responsible for the determination towards proliferation or differentiation of activated satellite cells in the mdx mouse. Since gene expression profiling detects proliferation and differentiation processes simultaneously, satellite cell activation and commitment are ongoing, parallel processes.
Bmp pathway
Various members of a Bmp-associated pathway (Bmp15, Gdf9, Bmpr1a, Madh4, Inhbc, Inhbe, Inhba, and Idb2) are differentially expressed in the mdx mouse (Figure 4B). The gene expression of Bmp15, Bmpr1a, Inhbc, Inhbe, and Idb2 is switched on in the mdx mouse, while expression cannot be detected in the control mouse. Bone Morphogenetic Protein 15 (Bmp15) is a member of the transforming growth factor-β (TGF-β) family. Bmp15 induces transcription of Inhibitors of DNA Binding proteins (Idb1-3) via binding to Bone Morphogenetic Protein Receptor type I (Bmpr1a), and the downstream transportation of Smad-complex (Madh8-Madh4) to the nucleus[52,53]. Idb-proteins function as positive regulators of cell growth by binding to Retinoblastoma 1 (Rb1). This leads to the activation of the E2f transcription factor, which plays a role in cell-cycle regulation. Furthermore, Idb-proteins inhibit myogenic differentiation through binding to MRFs[54]. The activation of Idb2 in the pre-regeneration period is indicative of an inhibition of the myogenic differentiation. This inhibition seems to be alleviated during the regeneration period by a decrease in differential expression of Idb2. Although most of the differentially expressed genes in the Bmp pathway are continuously upregulated, the expression of a number of genes peaks during regeneration (Inhbc, Inhbe, Bmp15, and Gdf9)(Figure 4B). The Inhibin proteins (Inhbc/e), likely to be antagonists of Bone Morphogenetic Proteins[55], are also upregulated. Altogether, this points to a positively and negatively controlled regulation of the Bmp15 pathway. Our data suggests that the Bmp15 pathway has an important function in the balancing of proliferation and differentiation of myoblasts, necessary for effective upscaling of muscle-mass.
Neuregulin pathway
In our study we found that several members of the Epidermal Growth Factor-like (EGF-like) Neuregulin pathway are differentially expressed (Figure 4C). The signalling cascade is activated by the binding of Neuregulin3 (Nrg3) to the extracellular domain of the upregulated protein tyrosine kinase v-erb-a erythroblastic leukemia viral oncogene homolog 4 (ErbB4)[56]. Both Nrg3 and ErbB4 are expressed in the mdx mouse and cannot be detected in the control mouse. The interaction between Nrg3 and ErbB4 activates epidermal growth factor-like signal transduction via binding of the adaptor protein Growth factor receptor bound protein 2 (Grb2), which peaks at the initiation of regeneration (Figure 4C). Grb2 can activate Mitogen activated kinase kinase 1 (Map3k1)[57], whose differential gene expression is increased at 2.5 weeks as well as during regeneration. The activation of the MAP kinase pathways eventually leads to transcriptional induction (reviewed in[58]) through members of Activating protein complex 1 (Ap1), like Jund1 and Jun.
Depending on the protein complexes formed, specific transcription activation will lead to different biological processes ranging from proliferation to differentiation. Furthermore, we find that other Grb-interacting proteins like Vav 1 oncogene (Vav1)[59], and p21-activated kinase 1 (Pak1)[60] are switched on and upregulated, respectively. The differential gene expression of Vav1 increases during the initiation of regeneration, and might play a role in the clustering of integrins for cell adhesion[61]. Pak1 differential gene expression is increased during regeneration. Downstream genes activated by Pak1 regulate cytoskeletal dynamics, proliferation and cell survival signalling[60]. Furthermore, our results show that the Erk pathway is downregulated in the mdx mouse as well as the p38 pathway.
Comparisons with other studies
In contrast to previously published studies of temporal gene expression profiling in the mdx mouse [25-27], we primarily focus on regeneration. The majority of differentially expressed genes in regeneration (148 out of 166) in our study have not been reported as differentially expressed in the mdx mouse in other studies. Apart from important differences in gene coverage (the Affymetrix U74v2 GeneChips used in the other studies lack probe sets for 22/166 genes), we explain the limited overlap by differences in cut-off levels: as we applied very stringent statistical tests, we could avoid setting a cut-off level for the fold change, thereby picking up genes with small but consistent fold changes, which can be biologically very relevant, especially in the case of transcription factors. This may also explain the large difference between the study of Porter et al. and our study in the number of genes found differentially expressed at the early timepoints (Figure 1), where mainly subtle expression changes are expected.
Goetsch et al. reported results from a gene expression profiling studies during muscle regeneration induced by cardiotoxin injection in wildtype mice[62]. The authors concluded that muscle regeneration is a complex process that requires the coordinated modulation of the inflammatory response, myogenic precursor cells, growth factors, and the extracellular matrix for complete regeneration of muscle architecture. A similar study of cardiotoxin-induced muscle regeneration, recently published by Zhao and Hoffman, reported that embryonic positional cues (Wnt, Shh, and Bmp) were not induced, whereas expression of factors involved in satellite cell proliferation and differentiation (MRFs, Pax, Notch1, and FGFR4) was recapitulated[63]. Our study, which also asserts satellite cell activation, proliferation, and differentiation, shows differences in muscle regeneration between mdx and wildtype mice. Bmp and EGF-like signalling pathways are activated during regeneration in the mdx mouse, as well as upregulation of members of the Notch-Delta pathway. In contrast to the upregulation of Pax7 in wildtype mice, we have found upregulation of Pax3.
Altogether, these findings suggest that dystrophin-deficiency might lead to enhanced regeneration processes in hindlimb muscles over and above those found in wildtype mice. It is likely that the regeneration pathways identified in our study are also active in the mdx diaphragm, given that the expression of their downstream targets, the muscle regulatory factors myf-5, myoD, and myogenin, are even more elevated in mdx diaphragm than in hindlimb[64]. As demonstrated in another recent study[65], the regeneration capacity per se is high and does not explain why the muscle wasting in the mdx diaphragm is more severe than in the hindlimb. Other factors such as higher workload and different involvement of the immune system are likely to contribute.
To discern active processes between the lethal human and regenerative murine dystrophin-deficiency, the selected murine gene expression profiles at 8 weeks of age were compared to those of DMD patients[66]. Out of 166 regeneration-associated transcripts, 19 genes could be detected that are differentially expressed in both the human and murine muscular dystrophy (Additional file 5). Seven of these overlapping genes showed an opposite differential expression between human DMD and mdx, of which Platelet derived growth factor beta (Pdgfb) and Paired box 3 (Pax3) are discussed below.
Gene expression of Pdgfb is upregulated in the mdx mouse (18.6 fold), where it is downregulated (-1.7-fold) in DMD patients. In the mdx mouse, Pdgfb shows an increase in gene expression during regeneration (present in cluster 4, Figure 3). Pdgfb was immunolocalized in infiltrating macrophages, regenerating muscle fibres, and myofibre nuclei of affected dystrophic muscle tissue[67]. The mitogen Pdgfb stimulates myoblast proliferation, while inhibiting myoblast differentiation[68]. It seems to have a similar role during regeneration. Paired box 3 (Pax3) gene expression is activated in the mdx mouse relative to the control. Its gene expression increases during regeneration, peaking at 12 weeks of age, while hPax3 is downregulated in DMD patients (-1.6-fold). Pax3 is capable of activating the expression of the muscle regulatory factors Myod1, Myf5 or Myogenin, and thereby activating the myogenic program[69].
The limited amount of overlapping genes between mdx mice and human DMD patients, as well as a number of genes showing opposite expression (i.e. Pdgfb and Pax3), suggests that processes active in regenerating mouse muscle are not active in human patients at the time gene expression was profiled (Age: 5–12 years old). This corresponds with clinical findings that patients older than 5 years have surpassed active regeneration processes[70]. The discovery of genes showing opposite regulation may partly explain the differences in regeneration efficiency and lethal manifestation of the pathology between dystrophin-deficient human and murine muscles.
Conclusion
Mdx mice lack a functional DGC at the sarcolemma. As a consequence, gene expression of most DGC members is downregulated. Mdx mice suffer from massive muscle fibre necrosis starting at the age of 3 weeks. Regenerative processes, starting approximately at the age of 6 weeks, largely restore muscle tissue architecture, although muscle fibres remain centrally nucleated. Recovered muscles of mdx mice have slightly diminished strength and higher fatigability. By analysing temporal expression profiles in mdx and control mice, we have identified genes and pathways involved in regeneration. The expression of these genes peaks between the ages of 6–12 weeks. Based on the observation that several of these genes are not expressed in control muscle and based on gene ontology classification and further literature annotation, we suggest a role for these genes in activation, proliferation, or differentiation of satellite cells and myoblasts. We propose the following model (Figure 6). Muscular dystrophy leads to muscle fibre necrosis, which attracts inflammatory cells, and release of trophic factors. These factors activate quiescent satellite cells, which as a consequence start to proliferate and differentiate. This divergent cell-fate is controlled by the Notch-Delta pathway. Activated satellite cells differentiate to myoblasts, which proliferate and differentiate as well. The balance between these cell-fates may be regulated by the level of Numb and the activation of Bmp15 and Nrg3 signalling pathways. Differentiation of myoblasts eventually leads to fusion with affected muscle fibres or to the formation of new muscle fibres. The genes and pathways active in regeneration are reminiscent of embryonic myogenesis. We hypothesize that the newly formed or repaired muscle fibres are similar to those in the pre-necrotic phase, but are more able to adapt to dystrophin-deficiency through remodelling of muscle structure and fibre composition. Since many regeneration-related genes remain higher expressed in mdx than in control muscle, it seems that regeneration processes are active throughout the life span of the animal. Regenerative processes appear to be most effective when mice reach adulthood, and normal growth processes cease. Regeneration and muscle development are both dependent on the availability of satellite cells, and these processes will therefore compete for the satellite cell availability when activated simultaneously. In the human situation, regeneration processes seem to be exhausted before growth is finished. Together with the accumulation of fibrotic and adipose tissue, exhaustion is thought to be the reason of the lethal manifestation of the disease in human patients. Prolongation of the regenerative capacity by activating the described pathways and/or replenishment with a pool of 'regeneration-primed' cells may therefore provide an attractive strategy in the treatment of muscular dystrophy.
Figure 6 Schematic model of processes in regeneration.
Methods
Target preparation and hybridisation
Hindlimb muscle tissue was isolated from mdx (C57Bl/10ScSnmdx/J (Jackson) × C57Bl/6NCrl (Charles River) × CBA/JCrl (Charles River)) and control (C57Bl/10ScSnOlaHsd, Harland) at the ages of 1, 2 1/2, 4, 6, 8, 10, 12, 14, and 20 weeks (2 individuals per time point per strain). Total RNA was isolated as described previously (Turk et al., submitted). cRNA was prepared by linear amplification and concurrent incorporation of amino-allyl UTP, followed by chemical coupling to monoreactive Cy3 or Cy5 dyes[71]. Labelled targets (1.5 μg cRNA per target) were hybridised overnight on prehybridised murine oligonucleotide microarrays (65-mer with 5'-hexylaminolinker, Sigma-Genosys mouse 7.5K oligonucleotide library, printed in duplicate) using an automatic hybridisation station (GeneTac, Genomic Solutions). Posthybridisation washes were performed as described previously[71].
Data analysis
Hybridisations were performed in a dye-swap fashion using temporal loop-designs [72-74], enabling optimal detection of gene expression differences between adjacent time points for both mdx and control targets [see Additional file 2] Microarrays were analysed by GenePix Pro 3 feature extraction software (Axon). Local background-corrected, median spot intensities were normalized simultaneously for all microarray experiments using Variance Stabilization and Normalization (VSN) in R[75]. This transformation coincides with the natural logarithm for the high intensities. Array data has been made available through the GEO data repository of the National Centre for Biotechnology Information under series GSE1574. Averaged (arithmetic mean) normalized intensities were calculated per gene per time point for mdx and control samples based upon 8 data points (2 biological replicates with 4 technical replicates each). This method is more efficient than ratio-based calculations for each hybridisation[76]. Genes were considered expressed when the average normalized intensity was higher than the background level. The background level was determined by calculating the averaged normalized intensity of 157 empty spots in all experiments plus 3 standard deviations.
Fold-changes in gene expression were calculated on a linear scale by subtraction of averaged normalized intensities of control samples from mdx samples at each time point, followed by returning e raised to the power of the difference. Maximum fold-changes per gene were determined according to the highest fold-change within the time course. Statistically significant differential gene expression per gene per time point was calculated between mdx and control samples by performing a two-tailed Student's t-test assuming equal distributions. Significance levels were set at 0.05 after applying a Bonferroni correction for multiple testing. Gene expression profiles were taken in consideration when at least one time point showed statistically significant differential gene expression between mdx and control samples.
Comparisons with other gene expression studies were facilitated by the program GeneHopper[77] that links annotations for different platforms.
Clustering
Temporal differential gene expression profiles were scaled to time point t = 1 week. Selected profiles were grouped using k-means clustering into a predetermined number of clusters according to the correlation similarity measure (Spotfire DecisionSite 7.1.1, Functional Genomics package). Clustering was initiated using evenly spaced profiles as algorithm. This method generates profiles to be used as centroids that are evenly distributed between the minimum and maximum value for each variable in the selected profiles (from Spotfire DecisionSite User's Guide and Reference Manual).
Comparison with data from Duchenne patients
In a previous study, gene expression levels in 2 pools of muscle RNA from 5 Duchenne patients (aged 5–6 years and aged 10–12 years, respectively) and 2 pools of muscle RNA from non-dystrophic controls (aged 5–12 years and aged 4–13 years, respectively), were evaluated on Affymetrix U95A and U95Av2 GeneChips® [78]. We re-analysed the gene expression data to obtain expression levels for all genes, as well as the most recent annotation. To this end, publicly available cel-files were loaded in Rosetta Resolver® Gene Expression Analysis System v4.0 (Rosetta Biosoftware Inc., Seattle, WA). Data were processed and normalized with the Rosetta error model for Affymetrix U95A Genechips. Genes that showed differential expression (p < 0.001 and absolute fold-change >1.5) between the pools of dystrophic patients and non-dystrophic controls were exported and linked to mouse UniGene clusters with GeneHopper, based on HomoloGene annotation.
Functional annotation
Gene Ontology annotation was developed by Compugen, using nomenclature obtained from the Gene Ontology consortium . Additional information was retrieved via OMIM and LocusLink .
Quantitative RT-PCR
cDNA was synthesized by adding 40 ng of random hexamer primers to 1 μg of total RNA in a total volume of 11 μl. After denaturation for 10 min at 70°C and cooling for 10 min on ice, 4 μl of 5× first-strand buffer (MBI-Fermentas), 2 μl of 10 mM dNTPs and 2 μl (200 U/μl) RevertAid RNase H- (MBI-Fermentas) was added. The mixture was incubated for 10 min at room temperature 2 hours at 42°C. The cDNA synthesis was halted by heating at 70°>C for 10 min. Quantitative PCR assays, using 10 μl of 20× diluted cDNA, were run on a MyIQ real-time PCR detection system (BioRad), applying 36 cycles of 10 seconds denaturation (95°C), 20 seconds annealing (60°C), and 25 seconds extension (72°C). PCR mixtures contained 1× PCR Buffer (Roche), 3 mM MgCl2, 225 μM of each dNTP, 250 μg/ml BSA, 1× SYBR-Green (diluted from 10,000× stock, Molecular Probes), 10 nM fluorescein (BioRad), 2.5 U homemade Taq polymerase, 0.25 U AmpliTaq (Roche), and 10 pmol of forward and reverse primers (for sequences see Additional file 3). The PCR efficiencies, determined by analysis of a dilution series of a mixture of all cDNA samples over 5 orders of magnitude, ranged from 94.5% to 98% for the different primer pairs. Melting curves were analyzed to confirm single product formation. Gene expression levels were calculated using the gene expression macro provided by BioRad and normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH, stable expression in all samples) expression levels.
Abbreviations
DMD Duchenne muscular dystrophy
DGC Dystrophin-associated glycoprotein complex
NMJ Neuromuscular junction
VSN Variance stabilization and normalization
RT-PCR Reverse transcription followed by polymerase chain reaction
Authors' contributions
RT performed the microarray hybridisations, analysed the microarrays and wrote the draft of the paper. ES helped to set up the microarray technique and assisted with the analysis. EdM was involved in the breeding of the mice, the isolation of the tissues, and the quantitative PCR experiments. GJvO edited the manuscript. JdD conceived the study. P'tH was involved in the microarray analysis and participated in the writing of the paper.
Supplementary Material
Additional File 1
Averaged expression and significance levels for 1735 genes differentially expressed at at least one time point.
Click here for file
Additional File 4
Costameric and DGC related gene expression.
Click here for file
Additional File 5
Regeneration associated genes.
Click here for file
Additional File 2
Temporal loop design. Hybridisations were done using a temporal loop design for the mdx and the control samples separately. The temporal loop design balances dyes and samples, and provides low variance between adjacent timepoints[73]. The order of age is maintained in the hybridisation scheme; each target is hybridised with the target of the following timepoint. Hybridisations are indicated by Hyb ID. Green and red boxes indicate labelling of target with Cy3 and Cy5 respectively. The number in the boxes indicates the age of the mouse. The boxes linked by an arrow are identical samples.
Click here for file
Additional File 3
Primer sequences used for quantitative RT-PCR assays.
Click here for file
Acknowledgements
This work was supported by the Nederlandse Stichting voor Wetenschappelijk Onderzoek (NWO) and the Center for Medical Systems Biology (CMSB) established by the Netherlands Genomics Initiative / Netherlands Organisation for Scientific Research (NGI/NWO).
==== Refs
Hoffman EP Brown RHJ Kunkel LM Dystrophin: the protein product of the Duchenne muscular dystrophy locus Cell 1987 51 919 928 3319190 10.1016/0092-8674(87)90579-4
Ervasti JM Campbell KP Membrane organization of the dystrophin-glycoprotein complex Cell 1991 66 1121 1131 1913804 10.1016/0092-8674(91)90035-W
Straub V Bittner RE Leger JJ Voit T Direct visualization of the dystrophin network on skeletal muscle fiber membrane J Cell Biol 1992 119 1183 1191 1447296 10.1083/jcb.119.5.1183
Carpenter S Karpati G Duchenne muscular dystrophy: plasma membrane loss initiates muscle cell necrosis unless it is repaired Brain 1979 102 147 161 427527
Petrof BJ Shrager JB Stedman HH Kelly AM Sweeney HL Dystrophin protects the sarcolemma from stresses developed during muscle contraction Proc Natl Acad Sci U S A 1993 90 3710 3714 8475120
Bischoff R Control of satellite cell proliferation Adv Exp Med Biol 1990 280 147 157 2248134
Chen G Quinn LS Partial characterization of skeletal myoblast mitogens in mouse crushed muscle extract J Cell Physiol 1992 153 563 574 1447318 10.1002/jcp.1041530318
Jimena I Pena J Luque E Vaamonde R Muscle hypertrophy experimentally induced by administration of denervated muscle extract J Neuropathol Exp Neurol 1993 52 379 386 8355027
Bischoff R Engel AG and Franzini-Armstrong C The satellite cell and muscle regeneration Myology: basic and clinical 1994 3 2nd New York, McGraw-Hill 97 118
Charge SB Rudnicki MA Cellular and molecular regulation of muscle regeneration Physiol Rev 2004 84 209 238 14715915 10.1152/physrev.00019.2003
Conboy IM Rando TA The regulation of Notch signaling controls satellite cell activation and cell fate determination in postnatal myogenesis Dev Cell 2002 3 397 409 12361602 10.1016/S1534-5807(02)00254-X
Conboy IM Conboy MJ Smythe GM Rando TA Notch-mediated restoration of regenerative potential to aged muscle Science 2003 302 1575 1577 14645852 10.1126/science.1087573
Klein CJ Coovert DD Bulman DE Ray PN Mendell JR Burghes AH Somatic reversion/suppression in Duchenne muscular dystrophy (DMD): evidence supporting a frame-restoring mechanism in rare dystrophin-positive fibers Am J Hum Genet 1992 50 950 959 1570844
Jejurikar SS Kuzon WMJ Satellite cell depletion in degenerative skeletal muscle Apoptosis 2003 8 573 578 14574063 10.1023/A:1026127307457
Bulfield G Siller WG Wight PA Moore KJ X chromosome-linked muscular dystrophy (mdx) in the mouse Proc Natl Acad Sci U S A 1984 81 1189 1192 6583703
Sicinski P Geng Y Ryder-Cook AS Barnard EA Darlison MG Barnard PJ The molecular basis of muscular dystrophy in the mdx-mouse: a point mutation Science 1989 244 1578 1580 2662404
Dangain J Vrbova G Muscle development in mdx mutant mice Muscle Nerve 1984 7 700 704 6543918 10.1002/mus.880070903
Anderson JE Ovalle WK Bressler BH Electron microscopic and autoradiographic characterization of hindlimb muscle regeneration in the mdx mouse Anat Rec 1987 219 243 257 3425943 10.1002/ar.1092190305
DiMario JX Uzman A Strohman RC Fiber regeneration is not persistent in dystrophic (MDX) mouse skeletal muscle Dev Biol 1991 148 314 321 1936568 10.1016/0012-1606(91)90340-9
Torres LF Duchen LW The mutant mdx: inherited myopathy in the mouse. Morphological studies of nerves, muscles and end-plates Brain 1987 110 ( Pt 2) 269 299 3567525
Pastoret C Sebille A mdx mice show progressive weakness and muscle deterioration with age J Neurol Sci 1995 129 97 105 7608742 10.1016/0022-510X(94)00276-T
Tkatchenko AV Le Cam G Leger JJ Dechesne CA Large-scale analysis of differential gene expression in the hindlimb muscles and diaphragm of mdx mouse Biochim Biophys Acta 2000 1500 17 30 10564714
Porter JD Khanna S Kaminski HJ Rao JS Merriam AP Richmonds CR Leahy P Li J Guo W Andrade FH A chronic inflammatory response dominates the skeletal muscle molecular signature in dystrophin-deficient mdx mice Hum Mol Genet 2002 11 263 272 11823445 10.1093/hmg/11.3.263
Rouger K Le Cunff M Steenman M Potier MC Gibelin N Dechesne CA Leger JJ Global/temporal gene expression in diaphragm and hindlimb muscles of dystrophin-deficient (mdx) mice Am J Physiol Cell Physiol 2002 283 C773 C784 12176734
Boer JM de Meijer EJ Mank EM van Ommen GB den Dunnen JT Expression profiling in stably regenerating skeletal muscle of dystrophin-deficient mdx mice Neuromuscul Disord 2002 12 S118 S124 12206806 10.1016/S0960-8966(02)00092-5
Tseng BS Zhao P Pattison JS Gordon SE Granchelli JA Madsen RW Folk LC Hoffman EP Booth FW Regenerated mdx mouse skeletal muscle shows differential mRNA expression J Appl Physiol 2002 93 537 545 12133862
Porter JD Merriam AP Leahy P Gong B Khanna S Dissection of temporal gene expression signatures of affected and spared muscle groups in dystrophin-deficient (mdx) mice Hum Mol Genet 2003 12 1813 1821 12874102 10.1093/hmg/ddg197
Schena M Shalon D Davis RW Brown PO Quantitative monitoring of gene expression patterns with a complementary DNA microarray Science 1995 270 467 470 7569999
Ervasti JM Costameres: the Achilles' heel of Herculean muscle J Biol Chem 2003 278 13591 13594 12556452 10.1074/jbc.R200021200
Warner LE DelloRusso C Crawford RW Rybakova IN Patel JR Ervasti JM Chamberlain JS Expression of Dp260 in muscle tethers the actin cytoskeleton to the dystrophin-glycoprotein complex and partially prevents dystrophy Hum Mol Genet 2002 11 1095 1105 11978768 10.1093/hmg/11.9.1095
Wagner KR Cohen JB Huganir RL The 87K postsynaptic membrane protein from Torpedo is a protein-tyrosine kinase substrate homologous to dystrophin Neuron 1993 10 511 522 8461138 10.1016/0896-6273(93)90338-R
Blake DJ Nawrotzki R Peters MF Froehner SC Davies KE Isoform diversity of dystrobrevin, the murine 87-kDa postsynaptic protein J Biol Chem 1996 271 7802 7810 8631824 10.1074/jbc.271.13.7802
Sadoulet-Puccio HM Rajala M Kunkel LM Dystrobrevin and dystrophin: an interaction through coiled-coil motifs Proc Natl Acad Sci U S A 1997 94 12413 12418 9356463 10.1073/pnas.94.23.12413
Peters MF Sadoulet-Puccio HM Grady MR Kramarcy NR Kunkel LM Sanes JR Sealock R Froehner SC Differential membrane localization and intermolecular associations of alpha-dystrobrevin isoforms in skeletal muscle J Cell Biol 1998 142 1269 1278 9732287 10.1083/jcb.142.5.1269
Holzfeind PJ Ambrose HJ Newey SE Nawrotzki RA Blake DJ Davies KE Tissue-selective expression of alpha-dystrobrevin is determined by multiple promoters J Biol Chem 1999 274 6250 6258 10037712 10.1074/jbc.274.10.6250
Newey SE Gramolini AO Wu J Holzfeind P Jasmin BJ Davies KE Blake DJ A novel mechanism for modulating synaptic gene expression: differential localization of alpha-dystrobrevin transcripts in skeletal muscle Mol Cell Neurosci 2001 17 127 140 11161474 10.1006/mcne.2000.0918
Zhou Q Chu PH Huang C Cheng CF Martone ME Knoll G Shelton GD Evans S Chen J Ablation of Cypher, a PDZ-LIM domain Z-line protein, causes a severe form of congenital myopathy J Cell Biol 2001 155 605 612 11696561 10.1083/jcb.200107092
Ohlendieck K Campbell KP Dystrophin-associated proteins are greatly reduced in skeletal muscle from mdx mice J Cell Biol 1991 115 1685 1694 1757468 10.1083/jcb.115.6.1685
Roberds SL Leturcq F Allamand V Piccolo F Jeanpierre M Anderson RD Lim LE Lee JC Tome FMS Romero NB Fardeau M Beckmann JS Kaplan JC Campbell KP Missense mutations in the adhalin gene linked to autosomal recessive muscular dystrophy Cell 1994 78 625 633 8069911 10.1016/0092-8674(94)90527-4
Lim LE Duclos F Broux O Bourg N Sunada Y Allamand V Meyer J Richard I Moomaw C Slaughter C Tome C Fardeau F Jackson M Beckmann CE Campbell KP Beta-sarcoglycan: characterization and role in limb-girdle muscular dystrophy linked to 4q12 Nat Genet 1995 11 257 265 7581448 10.1038/ng1195-257
Noguchi S McNally EM Ben Othmane K Hagiwara Y Mizuno Y Yoshida M Yamamoto H Bonnemann CG Gussoni E Denton PH Kyriakides T Middleton L Hentati F Ben Hamida M Nonaka I Vance JM Kunkel LM Ozawa E Mutations in the dystrophin-associated protein gamma-sarcoglycan in chromosome 13 muscular dystrophy Science 1995 270 819 822 7481775
Bonnemann CG Modi R Noguchi S Mizuno Y Yoshida M Gussoni E McNally EM Duggan DJ Angelini C Hoffman EP Beta-sarcoglycan (A3b) mutations cause autosomal recessive muscular dystrophy with loss of the sarcoglycan complex Nat Genet 1995 11 266 273 7581449 10.1038/ng1195-266
Nigro V de Sa ME Piluso G Vainzof M Belsito A Politano L Puca AA Passos-Bueno MR Zatz M Autosomal recessive limb-girdle muscular dystrophy, LGMD2F, is caused by a mutation in the delta-sarcoglycan gene Nat Genet 1996 14 195 198 8841194 10.1038/ng1096-195
Wakabayashi-Takai E Noguchi S Ozawa E Identification of myogenesis-dependent transcriptional enhancers in promoter region of mouse gamma-sarcoglycan gene Eur J Biochem 2001 268 948 957 11179961 10.1046/j.1432-1327.2001.01954.x
Delfini M Hirsinger E Pourquie O Duprez D Delta 1-activated notch inhibits muscle differentiation without affecting Myf5 and Pax3 expression in chick limb myogenesis Development 2000 127 5213 5224 11060246
Kopan R Nye JS Weintraub H The intracellular domain of mouse Notch: a constitutively activated repressor of myogenesis directed at the basic helix-loop-helix region of MyoD Development 1994 120 2385 2396 7956819
Nye JS Kopan R Axel R An activated Notch suppresses neurogenesis and myogenesis but not gliogenesis in mammalian cells Development 1994 120 2421 2430 7956822
Shawber C Nofziger D Hsieh JJ Lindsell C Bogler O Hayward D Weinmaster G Notch signaling inhibits muscle cell differentiation through a CBF1-independent pathway Development 1996 122 3765 3773 9012498
Artavanis-Tsakonas S Rand MD Lake RJ Notch signaling: cell fate control and signal integration in development Science 1999 284 770 776 10221902 10.1126/science.284.5415.770
Delgado I Huang X Jones S Zhang L Hatcher R Gao B Zhang P Dynamic gene expression during the onset of myoblast differentiation in vitro Genomics 2003 82 109 121 12837262 10.1016/S0888-7543(03)00104-6
Kuroda K Tani S Tamura K Minoguchi S Kurooka H Honjo T Delta-induced Notch signaling mediated by RBP-J inhibits MyoD expression and myogenesis J Biol Chem 1999 274 7238 7244 10066785 10.1074/jbc.274.11.7238
Moore RK Otsuka F Shimasaki S Molecular basis of bone morphogenetic protein-15 signaling in granulosa cells J Biol Chem 2003 278 304 310 12419820 10.1074/jbc.M207362200
Miyazawa K Shinozaki M Hara T Furuya T Miyazono K Two major Smad pathways in TGF-beta superfamily signalling Genes Cells 2002 7 1191 1204 12485160 10.1046/j.1365-2443.2002.00599.x
Norton JD Deed RW Craggs G Sablitzky F Id helix-loop-helix proteins in cell growth and differentiation Trends Cell Biol 1998 8 58 65 9695810 10.1016/S0962-8924(97)01183-5
Wiater E Vale W Inhibin is an antagonist of bone morphogenetic protein signaling J Biol Chem 2003 278 7934 7941 12493742 10.1074/jbc.M209710200
Zhang D Sliwkowski MX Mark M Frantz G Akita R Sun Y Hillan K Crowley C Brush J Godowski PJ Neuregulin-3 (NRG3): a novel neural tissue-enriched protein that binds and activates ErbB4 Proc Natl Acad Sci U S A 1997 94 9562 9567 9275162 10.1073/pnas.94.18.9562
Pomerance M Multon MC Parker F Venot C Blondeau JP Tocque B Schweighoffer F Grb2 interaction with MEK-kinase 1 is involved in regulation of Jun-kinase activities in response to epidermal growth factor J Biol Chem 1998 273 24301 24304 9733714 10.1074/jbc.273.38.24301
Hazzalin CA Mahadevan LC MAPK-regulated transcription: a continuously variable gene switch? Nat Rev Mol Cell Biol 2002 3 30 40 11823796 10.1038/nrm715
Lee IS Liu Y Narazaki M Hibi M Kishimoto T Taga T Vav is associated with signal transducing molecules gp130, Grb2 and Erk2, and is tyrosine phosphorylated in response to interleukin-6 FEBS Lett 1997 401 133 137 9013873 10.1016/S0014-5793(96)01456-1
Puto LA Pestonjamasp K King CC Bokoch GM p21-activated kinase 1 (PAK1) interacts with the Grb2 adapter protein to couple to growth factor signaling J Biol Chem 2003 278 9388 9393 12522133 10.1074/jbc.M208414200
Krawczyk C Oliveira-dos-Santos A Sasaki T Griffiths E Ohashi PS Snapper S Alt F Penninger JM Vav1 controls integrin clustering and MHC/peptide-specific cell adhesion to antigen-presenting cells Immunity 2002 16 331 343 11911819 10.1016/S1074-7613(02)00291-1
Goetsch SC Hawke TJ Gallardo TD Richardson JA Garry DJ Transcriptional profiling and regulation of the extracellular matrix during muscle regeneration Physiol Genomics 2003 14 261 271 12799472
Zhao P Hoffman EP Embryonic myogenesis pathways in muscle regeneration Dev Dyn 2004 229 380 392 14745964 10.1002/dvdy.10457
Porter JD Merriam AP Leahy P Gong B Feuerman J Cheng G Khanna S Temporal gene expression profiling of dystrophin-deficient (mdx) mouse diaphragm identifies conserved and muscle group-specific mechanisms in the pathogenesis of muscular dystrophy Hum Mol Genet 2004 13 257 269 14681298 10.1093/hmg/ddh033
Matecki S Guibinga GH Petrof BJ Regenerative capacity of the dystrophic (mdx) diaphragm after induced injury Am J Physiol Regul Integr Comp Physiol 2004 287 R961 R968 15191902
Chen YW Zhao P Borup R Hoffman EP Expression profiling in the muscular dystrophies: identification of novel aspects of molecular pathophysiology J Cell Biol 2000 151 1321 1336 11121445 10.1083/jcb.151.6.1321
Zhao Y Haginoya K Sun G Dai H Onuma A Iinuma K Platelet-derived growth factor and its receptors are related to the progression of human muscular dystrophy: an immunohistochemical study J Pathol 2003 201 149 159 12950028 10.1002/path.1414
Jin P Sejersen T Ringertz NR Recombinant platelet-derived growth factor-BB stimulates growth and inhibits differentiation of rat L6 myoblasts J Biol Chem 1991 266 1245 1249 1702435
Maroto M Reshef R Munsterberg AE Koester S Goulding M Lassar AB Ectopic Pax-3 activates MyoD and Myf-5 expression in embryonic mesoderm and neural tissue Cell 1997 89 139 148 9094722 10.1016/S0092-8674(00)80190-7
Hoffman EP Partridge TA Genotype/phenotype correlations in Duchenne/becker dystrophy Molecular and Cell Biology of Muscular Dystrophy Molecular and Cell Biology of Human Diseases Wright DJM 1993 2 1st London, Chapman & Hall 12 36
't Hoen PA de Kort F van Ommen GJ den Dunnen JT Fluorescent labelling of cRNA for microarray applications Nucleic Acids Res 2003 31 e20 12595569 10.1093/nar/gng020
Churchill GA Fundamentals of experimental design for cDNA microarrays Nat Genet 2002 32 490 495 12454643 10.1038/ng1031
Yang YH Speed T Design issues for cDNA microarray experiments Nat Rev Genet 2002 3 579 588 12154381
Townsend JP Multifactorial experimental design and the transitivity of ratios with spotted DNA microarrays BMC Genomics 2003 4 41 14525623 10.1186/1471-2164-4-41
Huber W Von Heydebreck A Sultmann H Poustka A Vingron M Variance stabilization applied to microarray data calibration and to the quantification of differential expression Bioinformatics 2002 18 S96 S104 12169536
't Hoen PA Turk R Boer JM Sterrenburg E de Menezes RX van Ommen GJ den Dunnen JT Intensity-based analysis of two-colour microarrays enables efficient and flexible hybridization designs Nucleic Acids Res 2004 32 e41 14982960 10.1093/nar/gnh038
Svensson BA Kreeft AJ van Ommen GJ den Dunnen JT Boer JM GeneHopper: a web-based search engine to link gene-expression platforms through GenBank accession numbers Genome Biol 2003 4 R35 12734015 10.1186/gb-2003-4-5-r35
Bakay M Zhao P Chen J Hoffman EP A web-accessible complete transcriptome of normal human and DMD muscle Neuromuscul Disord 2002 12 S125 S141 12206807 10.1016/S0960-8966(02)00093-7
|
16011810
|
PMC1190170
|
CC BY
|
2021-01-04 16:32:48
|
no
|
BMC Genomics. 2005 Jul 13; 6:98
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-98
|
oa_comm
|
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-511608350410.1186/1472-6963-5-51Research ArticlePredictors of failed attendances in a multi-specialty outpatient centre using electronic databases Lee Vernon J [email protected] Arul [email protected] Mark I [email protected] Bala [email protected] Department of Clinical Epidemiology, Tan Tock Seng Hospital, Singapore2 Division of Operations, Tan Tock Seng Hospital, Singapore2005 6 8 2005 5 51 51 5 3 2005 6 8 2005 Copyright © 2005 Lee et al; licensee BioMed Central Ltd.2005Lee et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Failure to keep outpatient medical appointments results in inefficiencies and costs. The objective of this study is to show the factors in an existing electronic database that affect failed appointments and to develop a predictive probability model to increase the effectiveness of interventions.
Methods
A retrospective study was conducted on outpatient clinic attendances at Tan Tock Seng Hospital, Singapore from 2000 to 2004. 22864 patients were randomly sampled for analysis. The outcome measure was failed outpatient appointments according to each patient's latest appointment.
Results
Failures comprised of 21% of all appointments and 39% when using the patients' latest appointment. Using odds ratios from the mutliple logistic regression analysis, age group (0.75 to 0.84 for groups above 40 years compared to below 20 years), race (1.48 for Malays, 1.61 for Indians compared to Chinese), days from scheduling to appointment (2.38 for more than 21 days compared to less than 7 days), previous failed appointments (1.79 for more than 60% failures and 4.38 for no previous appointments, compared with less than 20% failures), provision of cell phone number (0.10 for providing numbers compared to otherwise) and distance from hospital (1.14 for more than 14 km compared to less than 6 km) were significantly associated with failed appointments. The predicted probability model's diagnostic accuracy to predict failures is more than 80%.
Conclusion
A few key variables have shown to adequately account for and predict failed appointments using existing electronic databases. These can be used to develop integrative technological solutions in the outpatient clinic.
==== Body
Background
Failure to comply with outpatient medical appointments is a perennial problem, affecting costs, causing scheduling conflicts, and interrupting continuity of care. Failed appointments in different outpatient settings have ranged from 12% to 42% [1-7]. The resulting economic costs range from £65 per failed appointment in the United Kingdom in 1997 [2] to 3–14% of total outpatient clinic income in the United States [8]. This problem may be compounded if non-compliance with appointments is an indication of poorer clinical outcomes [9]. Most studies on failed appointments focused on the socio-economic and demographic factors that affect failures [1,10-13]. Other factors studied include symptom duration or resolution, illness, long waiting periods, forgotten appointments, and other commitments [13-16]. Successful interventions have included reminders, giving the patient's choice of date, improved communication, and selective overbooking [2,10,17]. However, almost all studies were for specific specialties in small-scaled settings [2,5,8-13].
We wanted to determine the intrinsic and external factors affecting failed outpatient appointments using only routinely available data. Our objective was to examine the factors most associated with failed appointments in Singapore, and to devise a prognostic index that administrators may use to identify potential defaulters. The findings will allow administrators to account for these factors when scheduling attendances, and provide the platform for problem solving. Such a prognostic index will also allow targeting of patients at higher risk of defaulting hence reducing the costs of intervening in patients who do keep their appointment.
Methods
This was a retrospective cohort study on patients attending all outpatient clinics at Tan Tock Seng Hospital, a 1400 bed general hospital in Singapore. Data was obtained from the hospital's appointment systems database and included 3,212,789 outpatient appointments starting from the creation of the electronic database in August 2000, to July 2004. Cancelled or rescheduled appointments were excluded, and a computer generated random sample of 10% of patients was used.
Outcome measures and input factors
The outcome measure was failure of a patient to attend his most recent appointment, analysed for individual patients who had at least one visit from August 2001 to July 2004. This allowed us to have at least one year of appointment history (starting August 2000) for all patients.
A system-unique alphanumeric patient identifier was then used to sort all appointments by individual patients. The most recent appointment was then selected and coded as "actualised" if the patient registered during the scheduled clinic session, or "failure" if the patient did not attend the appointment. The same process was used to identify the appointment history for each patient. To account for the varying frequency and duration of follow-up between patients, we analysed past history of failed appointments as a proportion of all scheduled appointments, hence allowing us to use the entire database for the predicted probability model. Patients with no record of previous appointments within the entire database period starting August 2000 were classified separately. As the maximum inter-appointment duration is usually not longer than a year, we could assume that cases seen after August 2001 with no prior database records were correctly classified as having no prior appointments.
Other factors studied included the patient's gender, race, age-group, days from scheduling to appointment, percentage of previous appointment failures, provision of cell phone numbers, distance from place of residence, and hospital admission during the appointment or between scheduling and appointment. Reasons for failed appointments were not obtained as there was no routine provision for contacting patients who defaulted. Direct distance from the patient's residence to the hospital was computed from the address zip codes and categorised into 3 groups – less than 6 km (1–2 districts away), 6 to 14 km (3–4 districts away), and more than 14 km (outlying districts). The data was stratified by specialties by categorising all 47 sub-specialty departments into 6 functional groups – medical subspecialties, surgical departments, ear, nose, and throat (ENT), ophthalmology, therapy, and others.
Statistical methods
Data extraction and management was done in Microsoft Access and data analysis was performed using Stata [18]. All tests were conducted at the 5% level of significance and we reported the odds ratios and corresponding 95% confidence intervals.
We started with a univariate analysis on all variables by simple regression. As the effect of confounding has been previously shown to be important [19], multivariate analysis with a multiple logistic regression model was also performed starting from the most significant variable in the univariate analysis and adding the next most significant, using the likelihood ratio test to observe improvements in the model's fit. The coefficients from the logistic regression were used to formulate the predicted probability model. For the final model, we used a receiver-operating characteristic (ROC) curve to assess the model's discriminatory ability for appointment actualisation. The data was then stratified by the six specialty functional groups, and the final multiple logistic regression analysis repeated to observe for possible differences across specialty departments.
Results
Failed appointments accounted for 21% of all appointments in the database. From our sampling, a total of 22864 patients were included and of the most recent visit for individual patients, 39% of these appointments resulted in failures. Table 1 gives the characteristics of the study population. 26% had no previous appointment record and more than 40% of appointments were in excess of three weeks after scheduling. Only a small proportion were actually hospitalised prior to, or during the appointment date (2% and 1% respectively). The majority of patients (60%) provided a cell phone number.
Table 1 Demographic characteristics and univariate factors associated with failed appointments, with the corresponding number of subjects (n), odds ratios, confidence intervals, and p-values (overall n = 22864).
Variable n (%) OR 95% CI p-value
Gender
Male 12453 (54%) 1
Female 10411 (46%) 0.94 (0.90, 0.99) 0.035
Race
Chinese 16951 (74%) 1
Malay 2073 (9%) 1.51 (1.38, 1.66) <0.001
Indian 2120 (9%) 1.73 (1.58, 1.90) <0.001
Others 1715 (8%) 1.42 (1.29, 1.57) <0.001
Age group
Up to 20 years 2002 (9%) 1
21 to 30 years 4298 (19%) 0.99 (0.89, 1.10) 0.838
31 to 40 years 4190 (18%) 0.97 (0.87, 1.08) 0.621
41 to 50 years 3992 (17%) 0.76 (0.68, 0.84) <0.001
51 to 60 years 3265 (14%) 0.67 (0.60, 0.75) <0.001
More than 60 years 5137 (22%) 0.75 (0.68, 0.84) <0.001
Days from scheduling to actual appointment
Up to 7 days 5852 (26%) 1
7 to 21 days 7234 (32%) 1.05 (0.98, 1.13) 0.144
More than 21 days 9840 (43%) 1.24 (1.16, 1.33) <0.001
Percentage of previous failed appointments
Up to 20% 5288 (23%) 1
21% to 40% 3584 (16%) 1.14 (1.04, 1.25) 0.007
41% to 60% 3596 (16%) 1.41 (1.29, 1.55) <0.001
More than 60% 4414 (19%) 1.95 (1.79, 2.13) <0.001
No previous appointment 6044 (26%) 4.67 (4.31, 5.06) <0.001
Provided cell phone number 13813 (60%) 0.09 (0.09, 0.10) <0.001
Approximate distance from TTSH
<6 km 8114 (37%) 1
6 to 14 km 8427 (39%) 0.99 (0.93, 1.06) 0.844
>14 km 5248 (24%) 1.11 (1.04, 1.20) 0.003
Admitted
During appointment date 182 (1%) 0.86 (0.63, 1.16) 0.320
Between appointment scheduling date and actual appointment date 423 (2%) 0.87 (0.71, 1.07) 0.183
Department
Surgical 7961 (37%) 1
Medical 6848 (32%) 1.06 (0.99, 1.13) 0.106
ENT 1935 (9%) 1.14 (1.03, 1.27) 0.014
Ophthalmology 3721 (17%) 1.13 (1.04, 1.23) 0.004
Therapy 613 (3%) 2.64 (2.23, 3.11) <0.001
Others 574 (3%) 13.4 (10.49, 17.11) <0.001
Analysis
In the univariate analysis (Table 1), we found that gender, race, age group, days from scheduling to appointment, previous failed appointments, provision of cell phone number, distance from the hospital, and department were all significantly associated with failed appointments.
From the multiple logistic regression analysis (Table 2), age group, days from scheduling to appointment, previous failed appointments, provision of cell phone number, distance from hospital, and department were independently and significantly associated with failed appointments. Those older than 40 years had significantly lower odds of appointment failure than those below 20. Malays and Indians had significantly higher odds ratio (OR 1.48 and 1.61 respectively) compared to Chinese. Scheduling to appointment time was a good predictor, and longer times increased the likelihood of failure (OR 1.29 for 7 to 21 days, and 2.38 for more than 21 days). Prior appointment history was also strongly predictive of failure. Patients with more than 40% failed appointments had significantly higher odds compared to those with less than 20%. Patients without previous appointments had the highest odds ratio of 4.38. Those residing more than 14 km from the hospital had a significant odds of failure 1.14 times that of those residing less than 7 km away. Those providing cell phone numbers were least likely to have failed appointments, with an odds ratio of 0.10 (95% CI: 0.10–0.11). Compared to surgical appointments, ENT, ophthalmology, therapy, and others had significantly higher odds of failure. Variables which did not improve the model's fit were gender, and hospital admission during or prior to appointment.
Table 2 Multivariate factors associated with failed appointments with the corresponding odds ratios, confidence intervals, and p-values.
Variable* OR 95% CI p-value
Age group
Up to 20 years 1
21 to 30 years (x1) 0.96 (0.83, 1.11) 0.575
31 to 40 years (x2) 0.93 (0.81, 1.08) 0.335
41 to 50 years (x3) 0.75 (0.64, 0.86) <0.001
51 to 60 years (x4) 0.66 (0.57, 0.77) <0.001
More than 60 years (x5) 0.84 (0.73, 0.97) 0.019
Race
Chinese 1
Malay (x6) 1.48 (1.31, 1.68) <0.001
Indian (x7) 1.61 (1.42, 1.81) <0.001
Others (x8) 1.03 (0.89, 1.18) 0.716
Days from scheduling to actual appointment
Up to 7 days 1
7 to 21 days (x9) 1.29 (1.16, 1.42) <0.001
More than 21 days (x10) 2.38 (2.16, 2.62) <0.001
Percentage of previous failed appointments
Up to 20% 1
21% to 40% (x11) 0.96 (0.85, 1.09) 0.565
41% to 60% (x12) 1.21 (1.07, 1.36) <0.001
More than 60% (x13) 1.79 (1.60, 2.00) <0.001
No previous appointment (x14) 4.38 (3.95, 4.86) <0.001
Provided cell phone number (x15) 0.10 (0.09, 0.11) <0.001
Approximate distance from TTSH
<6 km 1
6 to 14 km (x16) 1.02 (0.94, 1.11) 0.596
>14 km (x17) 1.14 (1.04, 1.25) <0.001
Department
Surgical 1
Medical (x18) 0.92 (0.84, 0.99) 0.049
ENT (x19) 1.2 (1.05, 1.37) 0.008
Ophthalmology (x20) 1.13 (1.02, 1.26) 0.022
Therapy (x21) 4.73 (3.85, 5.82) <0.001
Others (x22) 20.22 (15.34, 26.65) <0.001
* The indicator variables as used in the predicted probability equation in Figure 1 are shown in brackets () next to the respective variables.
Predicted probability model
Based on the final model, we created a prognostic index to predict failed appointments. The predicted probability of failure (pi) was calculated using the equation shown in Figure 1.
Figure 1 Predicted probability equation for appointment failure derived from the multiple logistic regression model.
From the final model's receiver-operating characteristic curve (Figure 2), the area under the curve of 0.84 (95%CI: 0.83–0.85) indicates that the model's overall diagnostic accuracy in predicting failed appointments is good. Using a cut-off of p = 0.24, the model had a sensitivity of 80%, specificity of 70%, and an accuracy of 73%.
Figure 2 Receiver-operating characteristic curve of the final multiple logistic regression model for failed appointments.
Stratification by department
We also performed a stratified analysis of the final multivariate model for department groups (Table 3). Provision of cell phone numbers was the only factor negatively associated with failed appointments across all departments, while no previous appointments was positively associated throughout. More than 21 days from scheduling to appointment was positively associated for all departments except therapy, where there was an insignificant negative association. Patients older than 40 years were negatively correlated with failed appointments except for elderly ophthalmology patients.
Table 3 Stratified analysis of factors by key departments
Variable Surgical (n = 7961) Medical (n = 6848) ENT (n = 1935) Ophthalmology (n = 3721) Therapy (n = 613) Others (n = 574)
Age group
Up to 20 years 1 1 1 1 1 1
21 to 30 years 0.77* 0.99 0.88 1.36 1.12 0.87
31 to 40 years 0.85 0.81 0.76 1.25 1.12 1.23
41 to 50 years 0.71* 0.7* 0.52* 0.87 0.88 0.81
51 to 60 years 0.71* 0.66* 0.49* 0.62* 0.6 0.42
More than 60 years 0.89 0.69* 0.86 1.1 0.62 0.79
Race
Chinese 1 1 1 1 1 1
Malay 1.69* 1.53* 1.26 1.35 0.98 0.91
Indian 1.44* 1.67* 1.53 1.85* 1.11 5.28*
Others 1.01 0.95 1.19 1.17 0.85 0.7
Days from scheduling to actual appointment
Up to 7 days 1 1 1 1 1 1
7 to 21 days 1.46* 1.36* 1.1 0.95 2.07* 0.78
More than 21 days 2.77* 2.7* 2.28* 1.99* 0.64 2.8*
Percentage of previous failed appointments
Up to 20% 1 1 1 1 1 1
21% to 40% 0.7* 1.37* 0.49* 0.97 1.24 1.18
41% to 60% 0.86 1.7* 1.06 1.21 1.98* 1.06
More than 60% 1.26* 3.1* 1.39 1.41* 1.59 3.25*
No previous appointment 3.84* 5.96* 4.07* 3.24* 1.86 4.37*
Provided cell phone number 0.07* 0.1* 0.06* 0.17* 0.15* 0.15*
Approximate distance from TTSH
<6 km 1 1 1 1 1 1
6 to 14 km 1.04 1.06 0.81 1.07 1.02 1.07
>14 km 1.13 1.12 0.86 1.34* 0.98 1.46
Numbers presented within the table represent adjusted odds ratios. The baseline comparison group has an odds ratio of 1 and is italicised. Odds ratios significant at a level of p < 0.05 are indicated with an asterisk "*".
Discussion
This study demonstrates that routinely available administrative data can be used to construct a prognostic index for appointment failures. Using a cut-off probability of above 0.24, the model identified defaulters with 80% certainty. Using the same cut-off, 30% of those who actualise their appointments would be wrongly classified. While imperfect, the model enables administrators to predict failed appointments with reasonable certainty for targeted intervention. Interventions have been shown to improve attendances, but certain methods such as personalised phone or postal reminders are manpower intensive [20-22]. With about 1,800 appointments a day in our clinics, the majority of which are actualised without intervention, having such predictions may lead to cost savings by targeting interventions towards patients with higher likelihood of defaulting.
Our analysis concurred with previous studies which showed that long waiting periods, repeat defaulters, and younger age groups are associated with increased likelihood of defaulting [1,10,13]. There are several findings of note that have not previously been reported. We found differences in the odds of attendance amongst different ethnic groups, which may reflect cultural differences that are amenable to interventions. Further studies are needed to explore the reasons for higher failure rates in Malay and Indian patients. More importantly, those who provided a cell phone number had an odds of actualising appointments 6 to 17 times higher than those who did not. This finding may be a conglomeration of various factors. Cell phone ownership may be an indicator of higher socio-economic status, which has been shown to be associated with higher rates of actualisation [10]. The provision of cell phone numbers could also indicate a patient's motivational level to attend appointments. Reasons aside, provision of cell phone numbers is an easily available yet robust predictor for appointment actualisation.
Some variables were less significant predictors than expected. We had expected travel distance to influence appointment failures, but the odds ratios were not as large as other variables. This may be due to convenient transportation and relatively short travel times in a small country like Singapore. Hospitalisation before and during the appointment date also did not contribute significantly, which may signify that hospitalisation itself does not preclude the need to seek treatment for other medical problems.
In the stratified departmental analysis, the effect of predictors, apart from cell phone numbers, was not uniform across departments. For example, the effect of duration from scheduling to appointment varies across specialties. This is to be expected because the duration of symptoms, urgency for treatment, and symptom resolution without treatment are different for conditions consulted at different specialties. The presence of this variation necessitates customised algorithms for individual departments in order for optimal predictions of appointment failure to be made.
There are several limitations to our study. We are uncertain if our findings can be generalised to other settings, as inter-institutional and inter-country differences similar to the observed inter-departmental differences may exist. There may also be differences between time-periods. However, while the predicted probability equation is only relevant for this hospital, the analytic process can be replicated using the methods described, since the study relies only on routinely available administrative data, which can be automatically processed for institutions with computerised appointment systems. Detailed data on failed appointments were unavailable and failed attendances may be reappointed as a new appointment if the patient is contactable. In addition, data before August 2000 is unavailable. Increased data definition may help in increasing the predictive accuracy, but the use of aggregate percentages in this study has produced good results. Our study was also unable to analyse failed appointments by clinical condition and symptoms. Other studies have shown that different clinical conditions and health status may be linked to failed attendances [23,24]. Future studies should include such variables to increase the predictive accuracy, but we note that our methodology already has diagnostic accuracy of more than 80% on the basis of routinely available data alone. This shows that an easily automated and reproducible system can have good predictive ability in spite of not incorporating clinical data, which is not available in most computerised appointment systems.
Our findings can be made operational in several ways. Predictions, based on up-to-date and institutionally relevant data, can be uploaded as automated algorithms into appointment systems. Lists of potential defaulters can then be generated using a desired sensitivity cut-off for targeted interventions to reduce appointment failure. In addition, educational messages can be targeted during prior appointments, based on automated profiling of future failure risk. Another strategy that is commonly used is over-booking to decrease opportunity costs but this can result in increased wait times if overdone. With the forward predictions on the expected appointment failure rate of a future clinic session, over-booking strategies can be optimised.
Conclusion
Failed appointments result in inefficiencies and economic cost and may interrupt continuity of care. We attempted to address the causes in an outpatient clinic and found that a few key routinely available variables could adequately account for appointment failure. The predicted probability model could predict failures with reasonable accuracy. Administrators can use these techniques to uncover factors in their own clinic deserving of further study. In addition, there is potential for incorporating automated algorithms into information systems to achieve better targeting of interventions, as well as to optimise overbooking strategies.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
VJL was involved in all areas including conceiving and designing the study, the data collection, the statistical analysis and writing of the paper. AE was involved in conceiving the study, the data collection, statistical analysis and writing of the paper. MIC was involved in designing the study and writing of the paper. BK involved in conceiving the study and writing of the paper. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors would like to acknowledge Mr R Chan for his help in extracting and processing the data and the Tan Tock Seng Hospital Information Technology Department for assistance in the data extraction.
==== Refs
Deyo RA Inui TS Dropouts and broken appointments. A literature review and agenda for future research Med Care 1980 18 1146 57 7432014
Hamilton W Alison R Sharp D Effect on hospital attendance rates of giving patients a copy of their referral letter: randomized controlled trial BMJ 1999 318 1392 5 10334750
al-Shammari SA Failures to keep primary care appointments in Saudi Arabia Fam Pract Res J 1992 12 171 6 1621537
Gatrad A completed audit to reduce hospital outpatients non-attendance rates Arch Dis Child 2000 82 59 61 10630915 10.1136/adc.82.1.59
Chung JWY Wong TKS Teung ACP Non-attendance at an orthopaedic and trauma specialist outpatient department of a regional hospital Journal of Nursing Management 2004 12 362 15315493 10.1111/j.1365-2834.2004.00484.x
Hermoni D Mankuta D Reis S Failure to keep appointments at a community health centre. Analysis of causes Scand J Prim Health Care 1990 8 151 5 2255819
Macharia WM Leon G Rowe BH Stephenson BJ Haynes RB An overview of interventions to improve compliance with appointment keeping for medical services JAMA 1992 267 1813 7 1532036 10.1001/jama.267.13.1813
Moore CG Wilson-Witherspoon P Probst JC Time and money: effects of failed appointments at a family practice residency clinic Fam Med 2001 33 522 7 11456244
Griffin SJ Lost to follow-up: the problem of defaulters from diabetes clinics Diabet Med 1998 15 S14 24 9829764 10.1002/(SICI)1096-9136(1998110)15:3+<S14::AID-DIA725>3.3.CO;2-9
Oppenhiem GL Bergman JJ English EC Failed appointments: a review J Fam Pract 1979 8 789 96 429996
Simmons AV Atkinson K Atkinson P Crosse B Failure of patients to attend a medical outpatient clinic J R Coll Physicians Lond 1997 31 70 3 9044203
Sanders G Craddock C Wagstaff I Factors influencing default at a hospital colposcopy clinic Qual Health Care 1992 1 236 40 10136870
Dickey W Morrow JI Can outpatient non-attendance be predicted from the referral letter? An audit of default at neurology clinics J R Soc Med 1991 84 662 3 1744872
Frankel S Farrow A West R Non-attendance or non-invitation? A case-control study of failed outpatient appointments BMJ 1989 298 1343 1345 2502248
Pal B Taberner DA Readman LP Jones P Why do outpatients fail to keep their clinic appointments? Results from a survey and recommended remedial actions Int J Clin Pract 1998 52 436 7 9894385
Cosgrove MP Defaulters in general practice: reasons for default and patterns of attendance Br J Gen Pract 1990 40 50 2 2107849
Barron WM Failed appointments. Who misses them, why they are missed, and what can be done Prim Care 1980 7 563 74 7010402
Stata Corp Stata Statistical Software: Release 8.2 2004 Stata Corporation, College Station, Texas
Gruzd DC Shear CL Rodney WM Determinants of failed appointment appointment behavior: the utility of multivariate anlaysis Fam Med 1986 18 217 20 3556868
Patel P Forbes M Gibson J The reduction of broken appointments in general dental practice: an audit and intervention approach Prim Dent Care 2000 7 141 4 11405012 10.1308/135576100322578889
Thomas D Postal reminders can improve attendance at orthodontic clinics Evid Based Dent 2004 5 14 15238969 10.1038/sj.ebd.6400244
Adams LA Pawlik J Forbes GM Nonattendance at outpatient endoscopy Endoscopy 2004 36 402 4 15100947 10.1055/s-2004-814329
Cashman SB Savageau JA Lemay CA Ferguson W Patient health status and appointment keeping in an urban community health center J Health Care Poor Underserved 2004 15 474 88 15453182
Yassin AS Howell RJ Nysenbaum AM Investigating non-attendance at colposcopy clinic J Obstet Gynaecol 2002 22 79 80 12521736 10.1080/01443610120101790
|
16083504
|
PMC1190171
|
CC BY
|
2021-01-04 16:31:53
|
no
|
BMC Health Serv Res. 2005 Aug 6; 5:51
|
utf-8
|
BMC Health Serv Res
| 2,005 |
10.1186/1472-6963-5-51
|
oa_comm
|
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-511608350410.1186/1472-6963-5-51Research ArticlePredictors of failed attendances in a multi-specialty outpatient centre using electronic databases Lee Vernon J [email protected] Arul [email protected] Mark I [email protected] Bala [email protected] Department of Clinical Epidemiology, Tan Tock Seng Hospital, Singapore2 Division of Operations, Tan Tock Seng Hospital, Singapore2005 6 8 2005 5 51 51 5 3 2005 6 8 2005 Copyright © 2005 Lee et al; licensee BioMed Central Ltd.2005Lee et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Failure to keep outpatient medical appointments results in inefficiencies and costs. The objective of this study is to show the factors in an existing electronic database that affect failed appointments and to develop a predictive probability model to increase the effectiveness of interventions.
Methods
A retrospective study was conducted on outpatient clinic attendances at Tan Tock Seng Hospital, Singapore from 2000 to 2004. 22864 patients were randomly sampled for analysis. The outcome measure was failed outpatient appointments according to each patient's latest appointment.
Results
Failures comprised of 21% of all appointments and 39% when using the patients' latest appointment. Using odds ratios from the mutliple logistic regression analysis, age group (0.75 to 0.84 for groups above 40 years compared to below 20 years), race (1.48 for Malays, 1.61 for Indians compared to Chinese), days from scheduling to appointment (2.38 for more than 21 days compared to less than 7 days), previous failed appointments (1.79 for more than 60% failures and 4.38 for no previous appointments, compared with less than 20% failures), provision of cell phone number (0.10 for providing numbers compared to otherwise) and distance from hospital (1.14 for more than 14 km compared to less than 6 km) were significantly associated with failed appointments. The predicted probability model's diagnostic accuracy to predict failures is more than 80%.
Conclusion
A few key variables have shown to adequately account for and predict failed appointments using existing electronic databases. These can be used to develop integrative technological solutions in the outpatient clinic.
==== Body
Background
Failure to comply with outpatient medical appointments is a perennial problem, affecting costs, causing scheduling conflicts, and interrupting continuity of care. Failed appointments in different outpatient settings have ranged from 12% to 42% [1-7]. The resulting economic costs range from £65 per failed appointment in the United Kingdom in 1997 [2] to 3–14% of total outpatient clinic income in the United States [8]. This problem may be compounded if non-compliance with appointments is an indication of poorer clinical outcomes [9]. Most studies on failed appointments focused on the socio-economic and demographic factors that affect failures [1,10-13]. Other factors studied include symptom duration or resolution, illness, long waiting periods, forgotten appointments, and other commitments [13-16]. Successful interventions have included reminders, giving the patient's choice of date, improved communication, and selective overbooking [2,10,17]. However, almost all studies were for specific specialties in small-scaled settings [2,5,8-13].
We wanted to determine the intrinsic and external factors affecting failed outpatient appointments using only routinely available data. Our objective was to examine the factors most associated with failed appointments in Singapore, and to devise a prognostic index that administrators may use to identify potential defaulters. The findings will allow administrators to account for these factors when scheduling attendances, and provide the platform for problem solving. Such a prognostic index will also allow targeting of patients at higher risk of defaulting hence reducing the costs of intervening in patients who do keep their appointment.
Methods
This was a retrospective cohort study on patients attending all outpatient clinics at Tan Tock Seng Hospital, a 1400 bed general hospital in Singapore. Data was obtained from the hospital's appointment systems database and included 3,212,789 outpatient appointments starting from the creation of the electronic database in August 2000, to July 2004. Cancelled or rescheduled appointments were excluded, and a computer generated random sample of 10% of patients was used.
Outcome measures and input factors
The outcome measure was failure of a patient to attend his most recent appointment, analysed for individual patients who had at least one visit from August 2001 to July 2004. This allowed us to have at least one year of appointment history (starting August 2000) for all patients.
A system-unique alphanumeric patient identifier was then used to sort all appointments by individual patients. The most recent appointment was then selected and coded as "actualised" if the patient registered during the scheduled clinic session, or "failure" if the patient did not attend the appointment. The same process was used to identify the appointment history for each patient. To account for the varying frequency and duration of follow-up between patients, we analysed past history of failed appointments as a proportion of all scheduled appointments, hence allowing us to use the entire database for the predicted probability model. Patients with no record of previous appointments within the entire database period starting August 2000 were classified separately. As the maximum inter-appointment duration is usually not longer than a year, we could assume that cases seen after August 2001 with no prior database records were correctly classified as having no prior appointments.
Other factors studied included the patient's gender, race, age-group, days from scheduling to appointment, percentage of previous appointment failures, provision of cell phone numbers, distance from place of residence, and hospital admission during the appointment or between scheduling and appointment. Reasons for failed appointments were not obtained as there was no routine provision for contacting patients who defaulted. Direct distance from the patient's residence to the hospital was computed from the address zip codes and categorised into 3 groups – less than 6 km (1–2 districts away), 6 to 14 km (3–4 districts away), and more than 14 km (outlying districts). The data was stratified by specialties by categorising all 47 sub-specialty departments into 6 functional groups – medical subspecialties, surgical departments, ear, nose, and throat (ENT), ophthalmology, therapy, and others.
Statistical methods
Data extraction and management was done in Microsoft Access and data analysis was performed using Stata [18]. All tests were conducted at the 5% level of significance and we reported the odds ratios and corresponding 95% confidence intervals.
We started with a univariate analysis on all variables by simple regression. As the effect of confounding has been previously shown to be important [19], multivariate analysis with a multiple logistic regression model was also performed starting from the most significant variable in the univariate analysis and adding the next most significant, using the likelihood ratio test to observe improvements in the model's fit. The coefficients from the logistic regression were used to formulate the predicted probability model. For the final model, we used a receiver-operating characteristic (ROC) curve to assess the model's discriminatory ability for appointment actualisation. The data was then stratified by the six specialty functional groups, and the final multiple logistic regression analysis repeated to observe for possible differences across specialty departments.
Results
Failed appointments accounted for 21% of all appointments in the database. From our sampling, a total of 22864 patients were included and of the most recent visit for individual patients, 39% of these appointments resulted in failures. Table 1 gives the characteristics of the study population. 26% had no previous appointment record and more than 40% of appointments were in excess of three weeks after scheduling. Only a small proportion were actually hospitalised prior to, or during the appointment date (2% and 1% respectively). The majority of patients (60%) provided a cell phone number.
Table 1 Demographic characteristics and univariate factors associated with failed appointments, with the corresponding number of subjects (n), odds ratios, confidence intervals, and p-values (overall n = 22864).
Variable n (%) OR 95% CI p-value
Gender
Male 12453 (54%) 1
Female 10411 (46%) 0.94 (0.90, 0.99) 0.035
Race
Chinese 16951 (74%) 1
Malay 2073 (9%) 1.51 (1.38, 1.66) <0.001
Indian 2120 (9%) 1.73 (1.58, 1.90) <0.001
Others 1715 (8%) 1.42 (1.29, 1.57) <0.001
Age group
Up to 20 years 2002 (9%) 1
21 to 30 years 4298 (19%) 0.99 (0.89, 1.10) 0.838
31 to 40 years 4190 (18%) 0.97 (0.87, 1.08) 0.621
41 to 50 years 3992 (17%) 0.76 (0.68, 0.84) <0.001
51 to 60 years 3265 (14%) 0.67 (0.60, 0.75) <0.001
More than 60 years 5137 (22%) 0.75 (0.68, 0.84) <0.001
Days from scheduling to actual appointment
Up to 7 days 5852 (26%) 1
7 to 21 days 7234 (32%) 1.05 (0.98, 1.13) 0.144
More than 21 days 9840 (43%) 1.24 (1.16, 1.33) <0.001
Percentage of previous failed appointments
Up to 20% 5288 (23%) 1
21% to 40% 3584 (16%) 1.14 (1.04, 1.25) 0.007
41% to 60% 3596 (16%) 1.41 (1.29, 1.55) <0.001
More than 60% 4414 (19%) 1.95 (1.79, 2.13) <0.001
No previous appointment 6044 (26%) 4.67 (4.31, 5.06) <0.001
Provided cell phone number 13813 (60%) 0.09 (0.09, 0.10) <0.001
Approximate distance from TTSH
<6 km 8114 (37%) 1
6 to 14 km 8427 (39%) 0.99 (0.93, 1.06) 0.844
>14 km 5248 (24%) 1.11 (1.04, 1.20) 0.003
Admitted
During appointment date 182 (1%) 0.86 (0.63, 1.16) 0.320
Between appointment scheduling date and actual appointment date 423 (2%) 0.87 (0.71, 1.07) 0.183
Department
Surgical 7961 (37%) 1
Medical 6848 (32%) 1.06 (0.99, 1.13) 0.106
ENT 1935 (9%) 1.14 (1.03, 1.27) 0.014
Ophthalmology 3721 (17%) 1.13 (1.04, 1.23) 0.004
Therapy 613 (3%) 2.64 (2.23, 3.11) <0.001
Others 574 (3%) 13.4 (10.49, 17.11) <0.001
Analysis
In the univariate analysis (Table 1), we found that gender, race, age group, days from scheduling to appointment, previous failed appointments, provision of cell phone number, distance from the hospital, and department were all significantly associated with failed appointments.
From the multiple logistic regression analysis (Table 2), age group, days from scheduling to appointment, previous failed appointments, provision of cell phone number, distance from hospital, and department were independently and significantly associated with failed appointments. Those older than 40 years had significantly lower odds of appointment failure than those below 20. Malays and Indians had significantly higher odds ratio (OR 1.48 and 1.61 respectively) compared to Chinese. Scheduling to appointment time was a good predictor, and longer times increased the likelihood of failure (OR 1.29 for 7 to 21 days, and 2.38 for more than 21 days). Prior appointment history was also strongly predictive of failure. Patients with more than 40% failed appointments had significantly higher odds compared to those with less than 20%. Patients without previous appointments had the highest odds ratio of 4.38. Those residing more than 14 km from the hospital had a significant odds of failure 1.14 times that of those residing less than 7 km away. Those providing cell phone numbers were least likely to have failed appointments, with an odds ratio of 0.10 (95% CI: 0.10–0.11). Compared to surgical appointments, ENT, ophthalmology, therapy, and others had significantly higher odds of failure. Variables which did not improve the model's fit were gender, and hospital admission during or prior to appointment.
Table 2 Multivariate factors associated with failed appointments with the corresponding odds ratios, confidence intervals, and p-values.
Variable* OR 95% CI p-value
Age group
Up to 20 years 1
21 to 30 years (x1) 0.96 (0.83, 1.11) 0.575
31 to 40 years (x2) 0.93 (0.81, 1.08) 0.335
41 to 50 years (x3) 0.75 (0.64, 0.86) <0.001
51 to 60 years (x4) 0.66 (0.57, 0.77) <0.001
More than 60 years (x5) 0.84 (0.73, 0.97) 0.019
Race
Chinese 1
Malay (x6) 1.48 (1.31, 1.68) <0.001
Indian (x7) 1.61 (1.42, 1.81) <0.001
Others (x8) 1.03 (0.89, 1.18) 0.716
Days from scheduling to actual appointment
Up to 7 days 1
7 to 21 days (x9) 1.29 (1.16, 1.42) <0.001
More than 21 days (x10) 2.38 (2.16, 2.62) <0.001
Percentage of previous failed appointments
Up to 20% 1
21% to 40% (x11) 0.96 (0.85, 1.09) 0.565
41% to 60% (x12) 1.21 (1.07, 1.36) <0.001
More than 60% (x13) 1.79 (1.60, 2.00) <0.001
No previous appointment (x14) 4.38 (3.95, 4.86) <0.001
Provided cell phone number (x15) 0.10 (0.09, 0.11) <0.001
Approximate distance from TTSH
<6 km 1
6 to 14 km (x16) 1.02 (0.94, 1.11) 0.596
>14 km (x17) 1.14 (1.04, 1.25) <0.001
Department
Surgical 1
Medical (x18) 0.92 (0.84, 0.99) 0.049
ENT (x19) 1.2 (1.05, 1.37) 0.008
Ophthalmology (x20) 1.13 (1.02, 1.26) 0.022
Therapy (x21) 4.73 (3.85, 5.82) <0.001
Others (x22) 20.22 (15.34, 26.65) <0.001
* The indicator variables as used in the predicted probability equation in Figure 1 are shown in brackets () next to the respective variables.
Predicted probability model
Based on the final model, we created a prognostic index to predict failed appointments. The predicted probability of failure (pi) was calculated using the equation shown in Figure 1.
Figure 1 Predicted probability equation for appointment failure derived from the multiple logistic regression model.
From the final model's receiver-operating characteristic curve (Figure 2), the area under the curve of 0.84 (95%CI: 0.83–0.85) indicates that the model's overall diagnostic accuracy in predicting failed appointments is good. Using a cut-off of p = 0.24, the model had a sensitivity of 80%, specificity of 70%, and an accuracy of 73%.
Figure 2 Receiver-operating characteristic curve of the final multiple logistic regression model for failed appointments.
Stratification by department
We also performed a stratified analysis of the final multivariate model for department groups (Table 3). Provision of cell phone numbers was the only factor negatively associated with failed appointments across all departments, while no previous appointments was positively associated throughout. More than 21 days from scheduling to appointment was positively associated for all departments except therapy, where there was an insignificant negative association. Patients older than 40 years were negatively correlated with failed appointments except for elderly ophthalmology patients.
Table 3 Stratified analysis of factors by key departments
Variable Surgical (n = 7961) Medical (n = 6848) ENT (n = 1935) Ophthalmology (n = 3721) Therapy (n = 613) Others (n = 574)
Age group
Up to 20 years 1 1 1 1 1 1
21 to 30 years 0.77* 0.99 0.88 1.36 1.12 0.87
31 to 40 years 0.85 0.81 0.76 1.25 1.12 1.23
41 to 50 years 0.71* 0.7* 0.52* 0.87 0.88 0.81
51 to 60 years 0.71* 0.66* 0.49* 0.62* 0.6 0.42
More than 60 years 0.89 0.69* 0.86 1.1 0.62 0.79
Race
Chinese 1 1 1 1 1 1
Malay 1.69* 1.53* 1.26 1.35 0.98 0.91
Indian 1.44* 1.67* 1.53 1.85* 1.11 5.28*
Others 1.01 0.95 1.19 1.17 0.85 0.7
Days from scheduling to actual appointment
Up to 7 days 1 1 1 1 1 1
7 to 21 days 1.46* 1.36* 1.1 0.95 2.07* 0.78
More than 21 days 2.77* 2.7* 2.28* 1.99* 0.64 2.8*
Percentage of previous failed appointments
Up to 20% 1 1 1 1 1 1
21% to 40% 0.7* 1.37* 0.49* 0.97 1.24 1.18
41% to 60% 0.86 1.7* 1.06 1.21 1.98* 1.06
More than 60% 1.26* 3.1* 1.39 1.41* 1.59 3.25*
No previous appointment 3.84* 5.96* 4.07* 3.24* 1.86 4.37*
Provided cell phone number 0.07* 0.1* 0.06* 0.17* 0.15* 0.15*
Approximate distance from TTSH
<6 km 1 1 1 1 1 1
6 to 14 km 1.04 1.06 0.81 1.07 1.02 1.07
>14 km 1.13 1.12 0.86 1.34* 0.98 1.46
Numbers presented within the table represent adjusted odds ratios. The baseline comparison group has an odds ratio of 1 and is italicised. Odds ratios significant at a level of p < 0.05 are indicated with an asterisk "*".
Discussion
This study demonstrates that routinely available administrative data can be used to construct a prognostic index for appointment failures. Using a cut-off probability of above 0.24, the model identified defaulters with 80% certainty. Using the same cut-off, 30% of those who actualise their appointments would be wrongly classified. While imperfect, the model enables administrators to predict failed appointments with reasonable certainty for targeted intervention. Interventions have been shown to improve attendances, but certain methods such as personalised phone or postal reminders are manpower intensive [20-22]. With about 1,800 appointments a day in our clinics, the majority of which are actualised without intervention, having such predictions may lead to cost savings by targeting interventions towards patients with higher likelihood of defaulting.
Our analysis concurred with previous studies which showed that long waiting periods, repeat defaulters, and younger age groups are associated with increased likelihood of defaulting [1,10,13]. There are several findings of note that have not previously been reported. We found differences in the odds of attendance amongst different ethnic groups, which may reflect cultural differences that are amenable to interventions. Further studies are needed to explore the reasons for higher failure rates in Malay and Indian patients. More importantly, those who provided a cell phone number had an odds of actualising appointments 6 to 17 times higher than those who did not. This finding may be a conglomeration of various factors. Cell phone ownership may be an indicator of higher socio-economic status, which has been shown to be associated with higher rates of actualisation [10]. The provision of cell phone numbers could also indicate a patient's motivational level to attend appointments. Reasons aside, provision of cell phone numbers is an easily available yet robust predictor for appointment actualisation.
Some variables were less significant predictors than expected. We had expected travel distance to influence appointment failures, but the odds ratios were not as large as other variables. This may be due to convenient transportation and relatively short travel times in a small country like Singapore. Hospitalisation before and during the appointment date also did not contribute significantly, which may signify that hospitalisation itself does not preclude the need to seek treatment for other medical problems.
In the stratified departmental analysis, the effect of predictors, apart from cell phone numbers, was not uniform across departments. For example, the effect of duration from scheduling to appointment varies across specialties. This is to be expected because the duration of symptoms, urgency for treatment, and symptom resolution without treatment are different for conditions consulted at different specialties. The presence of this variation necessitates customised algorithms for individual departments in order for optimal predictions of appointment failure to be made.
There are several limitations to our study. We are uncertain if our findings can be generalised to other settings, as inter-institutional and inter-country differences similar to the observed inter-departmental differences may exist. There may also be differences between time-periods. However, while the predicted probability equation is only relevant for this hospital, the analytic process can be replicated using the methods described, since the study relies only on routinely available administrative data, which can be automatically processed for institutions with computerised appointment systems. Detailed data on failed appointments were unavailable and failed attendances may be reappointed as a new appointment if the patient is contactable. In addition, data before August 2000 is unavailable. Increased data definition may help in increasing the predictive accuracy, but the use of aggregate percentages in this study has produced good results. Our study was also unable to analyse failed appointments by clinical condition and symptoms. Other studies have shown that different clinical conditions and health status may be linked to failed attendances [23,24]. Future studies should include such variables to increase the predictive accuracy, but we note that our methodology already has diagnostic accuracy of more than 80% on the basis of routinely available data alone. This shows that an easily automated and reproducible system can have good predictive ability in spite of not incorporating clinical data, which is not available in most computerised appointment systems.
Our findings can be made operational in several ways. Predictions, based on up-to-date and institutionally relevant data, can be uploaded as automated algorithms into appointment systems. Lists of potential defaulters can then be generated using a desired sensitivity cut-off for targeted interventions to reduce appointment failure. In addition, educational messages can be targeted during prior appointments, based on automated profiling of future failure risk. Another strategy that is commonly used is over-booking to decrease opportunity costs but this can result in increased wait times if overdone. With the forward predictions on the expected appointment failure rate of a future clinic session, over-booking strategies can be optimised.
Conclusion
Failed appointments result in inefficiencies and economic cost and may interrupt continuity of care. We attempted to address the causes in an outpatient clinic and found that a few key routinely available variables could adequately account for appointment failure. The predicted probability model could predict failures with reasonable accuracy. Administrators can use these techniques to uncover factors in their own clinic deserving of further study. In addition, there is potential for incorporating automated algorithms into information systems to achieve better targeting of interventions, as well as to optimise overbooking strategies.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
VJL was involved in all areas including conceiving and designing the study, the data collection, the statistical analysis and writing of the paper. AE was involved in conceiving the study, the data collection, statistical analysis and writing of the paper. MIC was involved in designing the study and writing of the paper. BK involved in conceiving the study and writing of the paper. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors would like to acknowledge Mr R Chan for his help in extracting and processing the data and the Tan Tock Seng Hospital Information Technology Department for assistance in the data extraction.
==== Refs
Deyo RA Inui TS Dropouts and broken appointments. A literature review and agenda for future research Med Care 1980 18 1146 57 7432014
Hamilton W Alison R Sharp D Effect on hospital attendance rates of giving patients a copy of their referral letter: randomized controlled trial BMJ 1999 318 1392 5 10334750
al-Shammari SA Failures to keep primary care appointments in Saudi Arabia Fam Pract Res J 1992 12 171 6 1621537
Gatrad A completed audit to reduce hospital outpatients non-attendance rates Arch Dis Child 2000 82 59 61 10630915 10.1136/adc.82.1.59
Chung JWY Wong TKS Teung ACP Non-attendance at an orthopaedic and trauma specialist outpatient department of a regional hospital Journal of Nursing Management 2004 12 362 15315493 10.1111/j.1365-2834.2004.00484.x
Hermoni D Mankuta D Reis S Failure to keep appointments at a community health centre. Analysis of causes Scand J Prim Health Care 1990 8 151 5 2255819
Macharia WM Leon G Rowe BH Stephenson BJ Haynes RB An overview of interventions to improve compliance with appointment keeping for medical services JAMA 1992 267 1813 7 1532036 10.1001/jama.267.13.1813
Moore CG Wilson-Witherspoon P Probst JC Time and money: effects of failed appointments at a family practice residency clinic Fam Med 2001 33 522 7 11456244
Griffin SJ Lost to follow-up: the problem of defaulters from diabetes clinics Diabet Med 1998 15 S14 24 9829764 10.1002/(SICI)1096-9136(1998110)15:3+<S14::AID-DIA725>3.3.CO;2-9
Oppenhiem GL Bergman JJ English EC Failed appointments: a review J Fam Pract 1979 8 789 96 429996
Simmons AV Atkinson K Atkinson P Crosse B Failure of patients to attend a medical outpatient clinic J R Coll Physicians Lond 1997 31 70 3 9044203
Sanders G Craddock C Wagstaff I Factors influencing default at a hospital colposcopy clinic Qual Health Care 1992 1 236 40 10136870
Dickey W Morrow JI Can outpatient non-attendance be predicted from the referral letter? An audit of default at neurology clinics J R Soc Med 1991 84 662 3 1744872
Frankel S Farrow A West R Non-attendance or non-invitation? A case-control study of failed outpatient appointments BMJ 1989 298 1343 1345 2502248
Pal B Taberner DA Readman LP Jones P Why do outpatients fail to keep their clinic appointments? Results from a survey and recommended remedial actions Int J Clin Pract 1998 52 436 7 9894385
Cosgrove MP Defaulters in general practice: reasons for default and patterns of attendance Br J Gen Pract 1990 40 50 2 2107849
Barron WM Failed appointments. Who misses them, why they are missed, and what can be done Prim Care 1980 7 563 74 7010402
Stata Corp Stata Statistical Software: Release 8.2 2004 Stata Corporation, College Station, Texas
Gruzd DC Shear CL Rodney WM Determinants of failed appointment appointment behavior: the utility of multivariate anlaysis Fam Med 1986 18 217 20 3556868
Patel P Forbes M Gibson J The reduction of broken appointments in general dental practice: an audit and intervention approach Prim Dent Care 2000 7 141 4 11405012 10.1308/135576100322578889
Thomas D Postal reminders can improve attendance at orthodontic clinics Evid Based Dent 2004 5 14 15238969 10.1038/sj.ebd.6400244
Adams LA Pawlik J Forbes GM Nonattendance at outpatient endoscopy Endoscopy 2004 36 402 4 15100947 10.1055/s-2004-814329
Cashman SB Savageau JA Lemay CA Ferguson W Patient health status and appointment keeping in an urban community health center J Health Care Poor Underserved 2004 15 474 88 15453182
Yassin AS Howell RJ Nysenbaum AM Investigating non-attendance at colposcopy clinic J Obstet Gynaecol 2002 22 79 80 12521736 10.1080/01443610120101790
|
16091137
|
PMC1190172
|
CC BY
|
2021-01-04 16:31:52
|
no
|
BMC Health Serv Res. 2005 Aug 9; 5:52
|
latin-1
|
BMC Health Serv Res
| 2,005 |
10.1186/1472-6963-5-52
|
oa_comm
|
==== Front
BMC Health Serv ResBMC Health Services Research1472-6963BioMed Central London 1472-6963-5-541610217410.1186/1472-6963-5-54Research ArticleVerbal and physical violence towards hospital- and community-based physicians in the Negev: an observational study Carmi-Iluz Tal [email protected] Roni [email protected] Tami [email protected] Pesach [email protected] The Department of the Family Medicine, Faculty of Health Sciences, Ben-Gurion University of the Negev, Beer-Sheva, Israel2 Sial Research Center for Family Medicine and Primary Care, Faculty of Health Sciences, Ben-Gurion University of the Negev, Beer-Sheva, Israel2005 15 8 2005 5 54 54 20 2 2005 15 8 2005 Copyright © 2005 Carmi-Iluz et al; licensee BioMed Central Ltd.2005Carmi-Iluz et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Over recent years there has been an increasing prevalence of verbal and physical violence in Israel, including in the work place. Physicians are exposed to violence in hospitals and in the community. The objective was to characterize acts of verbal and physical violence towards hospital- and community-based physicians.
Methods
A convenience sample of physicians working in the hospital and community completed an anonymous questionnaire about their experience with violence. Data collection took place between November 2001 and July 2002. One hundred seventy seven physicians participated in the study, 95 from the hospital and 82 from community clinics. The community sample included general physicians, pediatricians, specialists and residents.
Results
Ninety-nine physicians (56%) reported at least one act of verbal violence and 16 physicians (9%) reported exposure to at least one act of physical violence during the previous year. Fifty-one hospital physicians (53.7%) were exposed to verbal violence and 9 (9.5%) to physical violence. Forty-eight community physicians (58.5%) were exposed to verbal violence and 7 (8.5%) to physical violence. Seventeen community physicians (36.2%) compared to eleven hospital physicians (17.2%) said that the violence had a negative impact on their family and on their quality of life (p < 0.05). The most common causes of violence were long waiting time (46.2%), dissatisfaction with treatment (15.4%), and disagreement with the physician (10.3%).
Conclusion
Verbal and/or physical violence against physicians is common in both the hospital and in community clinics. The impatience that accompanies waiting times may have a cultural element. Shortening waiting times and providing more information to patients and families could reduce the rate of violence, but a cultural change may also be required.
==== Body
Background
Over recent years Israel, as well as other countries, has witnessed an increase in the prevalence of acts of violence. This rise is seen in workplaces, in recreation sites, on the roads, within the family and even in schools. The mass media are full of reports on violent acts. Violence does not necessarily involve physical contact; it can be verbal or mental. Sometimes psychological or verbal abuse has more severe consequences than acts of physical violence.
Health service providers in hospitals and community clinics are often exposed to verbal and even physical violence that can engender frustration and despair [1-10]. Violent acts against workers have been defined as "any event that the worker is threatened or attacked by another person due to his job" [3,9]. Many physicians feel threatened by verbal and physical violence at work [4,9]. Physicians in emergency medicine, psychiatrists and primary care physicians are at increased risk of violent acts from patients and families [3,4,10].
Studies from England from 1989, 1991, and 1997 have shown that verbal abuse is the most frequent type of violence reported by British physicians (25–91%) [5-7] compared to physical violence (1–11%). However the latter has significantly affected those physicians who were exposed to it leading in some cases to depression, insomnia, post-traumatic stress disorder, agoraphobia [4] and even a level of fear and/or anxiety that can cause work absenteeism [8]
In the US the rate of violence is even higher. Between the years 1980–1990, 106 healthcare workers died as a result of violence: 27 pharmacists, 26 physicians, and 53 nurses [2]. Another survey of 170 university hospitals in the US showed that 57% of all emergency room employees had been threatened by weapons over the five-year period prior to the survey [10].
Seventy percent of the physicians and 90% of the support staff working in a hospital emergency room in Israel reported violent acts, mostly verbal abuse [1].
The main reasons for these outbursts were long waiting times, dissatisfaction with treatment, something that was said that the patient took exception to, and in some cases the influence of alcohol and/or drugs on the perpetrator of the violence [1]. No other studies of this type have been reported from Israel.
The aim of the present study was to assess violence against physicians in the southern Negev region of Israel, and to compare rates in the hospital with those in community clinics.
Methods
Setting
The study was conducted within the framework of the Southern District of Clalit Health Services, Israel's largest HMO that serves about 60% of the population. The population of the Negev region in southern Israel numbers about 530,000, most of a low to middle socioeconomic level. The largest city in the area is Beer-Sheva with about 200, 000 residents. The Soroka University Medical Center is located in Beer-Sheva. The rest of the Negev's residents live in smaller communities.
The study was conducted among 95 physicians in all major departments of the Soroka Medical Center (internal medicine, surgery and pediatrics) and 82 family physicians and pediatricians working in primary care clinics of the Clalit Health Services in the Negev. Hospital physicians were sampled based on the physician roster of the Soroka University Medical Center and the community-based physicians based on the physician roster of the Southern District of the Clalit Health Services. In the few cases in which the physicians stated that they work both at the hospital and in the community, their primary place of work was used for statistical analyses.
The study instrument
All participants completed an anonymous questionnaire, consisting of 36 items, on their experiences with and attitudes towards violence. The questionnaire included demographic and personal data, reports on exposure to verbal and physical abuse over the previous year, information about how they dealt with the violence and their attitude to it. Most of the items were multiple-choice questions, with one possible answer, but in a small number of questions we used an open format. An example of a closed question is: How did you react to an episode of verbal violence? The response options were: 1) I ignored it; 2) I made a verbal response; 3) I called for security personnel; 4) I called the police; 5) I lodged a complaint with the police; 6) other (with space to write a detailed response).
We developed the questionnaire specifically for the study after a thorough review of the literature on the subject. The questionnaire was revised in light of the results of a pilot study of 15 physicians who did not participate in the study itself. In the explanation that preceded the questionnaire we stated that the term "verbal violence" could be defined as an attempt to attack another person by means of harsh words, cursing, an aggressive manner of speech, threats, or any other manner of speech that it is not acceptable, but does not lead to physical injury. "Physical violence" could be defined as any form of attack that has a physical component.
Data were entered into the Epi-Info 6 program and were analyzed using the SPSS statistical package. T-tests and χ2 tests were used as appropriate. Statistical significance was set at P < 0.05.
Results
The study population was comprised of 177 physicians from a roster of 200 (105 hospital-based and 85 community-based physicians), so the response rate was 88.5%. Ninety five (53.7%) worked in the Soroka Medical Center and 82 (46.3%) worked in the community. Sixty nine (39.7%) were women. The mean age was 41 ± 8.6 years. Eighty physicians (45.5%) were Israeli born and 52 (29.5%) were from the former USSR. The socio-demographic characteristics of the participants are presented in Table 1.
Table 1 Comparison of study groups for sociodemographic and professional data (n = 177; some variables have missing values).
Variable Hospital physicians Community physicians All P
Mean age (years) 40.0 ± 9.4 42.3 ± 7.3 41.0 ± 8.6 NS
Gender [N (%)]
Female 28 (30.1) 41 (50.6) 69 (39.7) <0.01
Male 65 (69.9) 40 (49.4) 105 (60.3)
Country of birth [N (%)]
Israel 53 (56.4) 27 (32.9) 80 (45.5) <0.05
East Europe 9 (9.6) 12 (14.6) 21 (11.9)
West Europe 2 (2.1) 0 2 (1.1)
Former USSR 22 (23.4) 30 (36.6) 52 (29.5)
Other 8 (8.5) 13 (15.9) 21 (12.0)
Professional status [N (%)]
Specialist 44 (46.3) 56 (69.1) 100 (56.8) <0.01
Non-specialist 51 (53.7) 25 (30.9) 76 (43.2)
Mean professional experience (years) 12.3 ± 9.7 14.2 ± 7.3 NS
Verbal violence
Ninety-nine of the 177 participating physicians (56.0%) stated that during the previous year they had been exposed to at least one case of verbal abuse. Fifty-one hospital physicians (53.7%) were exposed to verbal abuse by patients and 69 (72.6%) by family members, compared to 48 (58.5%) and 40 (49.4%), respectively for community-based physicians (p < 0.01 for verbal violence by family members).
Fifty six physicians (73%) reported having experienced 1–3 cases of verbal abuse by their patients over the past year (mean 4.3 ± 6.8). Seventy eight physicians indicated the number of times they had been exposed to verbal violence from family members of patients. Of these, 57 physicians (73%) mentioned that during the last year they had experienced 1–4 cases (mean 3.65 ± 3.39).
Physical violence
Sixteen of the 177 physicians (9%) were exposed to physical violence over the previous year. Nine doctors (5.1%) experienced acts of physical violence by their patients, and 7 doctors (4%) by family members of their patients. Three hospital physicians (3.2%) were exposed to physical violence by patients and 6 (6.3%) by family members, compared to 6 (7.3%) and 1 (1.2%), respectively for community-based physicians.
Nine physicians (5.1%) were punched, 3 (1.7%) had an object thrown at them, 2 (1.1%) were held with excessive pressure, and 1 (0.6%) was punched and had an object thrown at him.
The site of violence
Verbal violence
Forty three of 113 respondents (38.1%) said that the act of verbal violence took place on the ward, 33 (29.2%) in the physician's office, and 29 (25.7%) in the emergency room.
Physical violence
Five respondents (2.8%) said that the act of physical violence took place on the ward, 5 (2.8%) in the physician's office, and 5 (2.8%) in the emergency room.
Physicians' perceptions of the degree of risk from violence
Twenty two of 170 respondents (13%) said that they felt that their health had been endangered as a result of verbal violence the past year, 18 (10.6%) to a minor degree and 2 (1.2%) to a major degree. Thirty nine of 174 respondents (22.4%) said that they felt that their life had been endangered as a result of physical violence the past year.
The impact of violence on physicians (Table 2)
Table 2 The impact of violence on physicians' lives.
Variable Hospital physicians N (%) Community physicians N (%) P
Did violence affect life outside of work?
No 53 (82.8) 30 (63.8) <0.05
Yes 11 (17.2) 17 (36.2)
Length of time violence had this effect?
One day 43 (71.7) 31 (67.4) NS
Up to one month 12 (20.0) 9 (19.6)
More than one month 5 (8.3) 6 (13.0)
One hundred eleven physicians answered a question about the effect of violence on their family life and their overall quality of life. Twenty eight (25.5%) stated that violence had a detrimental effect on their lives. Seventeen of those who said that violence had a detrimental effect on their lives (60.7%) were females compared to 11 males (39.3%) (p < 0.05). Eleven hospital-based physicians (17.2%) and 17 community-based physicians (36.2%) said that violence had a detrimental effect on their life (p < 0.05).
Twenty one physicians (19.8% of 106 who answered the question) said that the detrimental effect lasted for up to one month, 11 (10.4%) said the effect lasted over one month.
The perceived causes of violence
The most frequent causes of violent acts cited by the physicians were: long waiting periods (46.2%), patients' dissatisfaction with the treatment (15.4%), patients' disagreement with the physician (10.3%), no perceivable reason (9.4%), patients' unjustified request for a medical certificate (4.2%), other reasons (14.5%). Thirty two community-based physicians (50%) stated that the reason for violence was long waiting times in their clinic.
Handling of the incident
Seventy three of the 106 physicians (68.9%) who answered this question said that acts of violence weren't treated in any way by the hospital or the community clinic, 19 (17.9%) said that the security staff sent the attacker away, and 10 (9.4%) said that the act was treated through legal channels.
Physicians' satisfaction with the manner in which the authorities handled the act
Of 100 physicians who answered this question, 43 weren't satisfied with the way the incident was handled, 32 were satisfied, and 25 were partially satisfied.
Why did physicians not file a complaint with the local police?
Of 94 physicians who answered this question, 43 (45.7%) felt that the incident did not justify a complaint to the police, 17 (18.1%) were satisfied with the attacker's apology, and 15 (16%) said that they were not prepared to go to court over the incident.
Physicians' attitudes to violence
Sixty percent of the participating physicians, including those who were not personally exposed to acts of violence over the previous year, believed that violence represents a serious threat. Only 6 (3.4%) thought that violence is not a serious problem for physicians today. Eighty three percent stated that they were not trained to prevent or deal with acts of violence.
Discussion
The deterioration of the economic and security situation has led to increasing violence in Israel. Hospital- and community-based physicians, who care for patients affected by physical and mental distress, may be exposed to violent acts at the workplace. Inadequate working conditions, with a small number of physicians caring for a large numbers of patients, overload the health care system causing prolonging waiting times that may trigger violent outbursts on the part of frustrated patients and their families. According to Israeli Ministry of Health data there were 1440 cases of violence in Israeli hospitals in 2001 compared with 675 in 2000. Of these, 432 were physical violence, 739 were verbal violence and 269 involved loss of property. The police intervened in 285 cases of violence in some cases making arrests and filing charges [11].
The aim of this study was to characterize verbal and physical violence against physicians in hospital and in community clinics in the Negev as reported by the physicians themselves and to assess their attitudes to these acts. Ninety nine of 177 participating physicians (56%) reported at least one incidence of verbal violence over the previous year. Sixteen of 177 (9%) reported at least one case of physical violence over the same time period. The response rate of 88.5% is high and was achieved by virtue of a data collection system that included a personal appointment with each physician, use of a relatively short questionnaire that had as its heading the official logo of the Department of Family Medicine, and the fact that the study population was highly motivated and inclined to participate in a study on an issue that is close to their hearts and important to them [12].
In terms of the degree of physical and verbal violence, our findings are similar to those reported from Great Britain [5-7], which show that verbal abuse is the most frequent type of violence, but are better than those reported from a primary care department in a Kuwaiti hospital in 1999 [13]. There, 86% of the physicians surveyed reported being exposed to verbal violence, and 28% to physical violence, with severe injury or even death in 7% of the cases. In the USA 106 health employees died as a result of violence between the years 1980–1990 [2].
In a study conducted in the Barzilai hospital in Ashkelon, Israel 70% of the participating physicians reported acts of verbal violence and 38.5% of the emergency medicine personnel experienced physical violence [1]. It is possible that work in an emergency room entails greater risks of violence than other hospital departments.
Significantly more hospital-based physicians (72.6%) experienced verbal violence by family members than community-based physicians (49.4%). Possible explanations for this finding include:
1. Hospital physicians are exposed more to confrontations with patients' family members who visit their hospitalized relatives and often want to be involved in decisions relating to them, while most patients come to community clinics alone.
2. Family members may feel less constrained to confront hospital physicians whom they don't know than family physicians with whom they may have an ongoing relationship.
Studies from around the world show that acts of violence have a negative effect on the physician's family life and quality of life. In the study conducted in Kuwait 86% of the physicians who experienced violence reported that it caused insomnia, depressions and other effects [14]. In the present study we found that 36.2% of the community-based physicians reported a negative impact on their family life and quality of life, compared to 17.2% of the hospital-based physicians. This may be explained by the long-term professional relationship that the family physician has with the violent patient, compared to the short-term contact that the hospital physician has with the patient.
The most common triggers for violent acts were long waiting time, dissatisfaction with the treatment and/or disagreement with the physician. Similar findings were reported from the study conducted in Barzilai hospital [1]. It is possible that the impatience that accompanies a long waiting time and acts as a trigger for violence is affected by cultural norms as well.
In two thirds of the cases the incident was not handled by the hospital or the community clinic, perhaps because the involved physician may not have perceived it as severe, although even the severe cases were not always handled by the hospital or the police. Some physicians may hesitate to report acts of violence because they do not expect the medical administration to come to their aid. It is also possible that physicians' cultural background may influence their reactions and approach to handling the incident. Most participating physicians, including those who did not experience acts of violence over the previous year, thought that violence is a significant issue for physicians. In Kuwait the overwhelming majority of study physicians said they were concerned about violence in the workplace and felt that physicians should receive formal instruction on how to handle these situations [14]. We also found that most of the physicians felt that they were not trained to prevent or cope with the violence at the workplace, indicating that even though violence exists and is considered a serious problem by physicians, it does not appear to be a major issue for the medical establishment and there are no serious attempts to minimize the phenomenon.
There are some limitations to our study. It was undertaken in major hospital departments, but did not include the emergency department where tension and violence may appear. This may be a reason for the relatively low rate of violence in the hospital. On the other hand it should be noted that many of the physicians who were interviewed as ward physicians actually spend part of their work time in the emergency room. Another limitation is the lack of information regarding the violent patient or his/her family, i.e., do they have a personality disorder or psychiatric problems, or perhaps just a normal person who felt that the waiting time was unreasonable.
The issue of triggers of violence was not dealt with in a way that fully recognizes the complexity of attributions made by people involved in the violent act, since no attempt was made to corroborate the attributes by interviewing the assailants. Although this type of expanded study might have contributed to the discussion of possible triggers, we feel that the methodology used, despite this limitation, is valid.
The results of this study add to our current knowledge because of its comparison of physical and verbal violence on the part of patients and/or their family, and the comparison between hospital- or community-based physicians from the same geographic area.
In order to start to alleviate the problem of violence towards physicians steps have to be taken to address the triggers listed above. Waiting times have to be shortened. This can be accomplished by better planning of medical education and manpower, and by increasing the number of physicians in the clinics to improve the patient-physician ratio. It is also important to change cultural norms that may serve as triggers to violence, but this strategy requires a long period of time and would require further studies before its implementation. In the short term it is important to organize workshops to train physicians to prevent and deal with violent incidents. In these training sessions physicians will be taught to be patient, to provide appropriate and relevant information, and to show respect towards patients and family members. At the same time hospital and clinical security should be increased and enforced; the strict rules for treating violent people should be designed and implemented.
Declaration of competing interests
The author(s) declare that they have no competing interests.
==== Refs
Derazon H Nissimian S Yosefy C Peled R Hay E [Violence in the emergency department] [Hebrew] Harefuah 1999 137 95 101 10959292
Goodman RA Jenkins EL Mercy JA Workplace-related homicide among health care workers in the United States, 1980 through 1990 JAMA 1994 272 1686 1688 7966897 10.1001/jama.272.21.1686
Schnieden V Violence against doctors British Journal of Hospital Medicine 1993 50 6, 9 8364708
Zahid MA al-Sahlawi KS Shahid AA al-Ajmi MT Awadh JA Violence towards doctors: prevalence and effects Hospital Medicine 1999 60 414 418 10492712
Ness GJ House A Ness AR Aggression and violent behaviour in general practice: population based survey in the north of England British Medical Journal 2000 320 1447 1448 10827050
D'Urso P Hobbs R Aggression and the general practitioner British Medical Journal 1989 298 97 98 2493309
Hobbs FD Violence in general practice: a survey of general practitioners' views British Medical Journal 1991 302 329 332 2001509
Hobbs FD Fear of aggression at work among general practitioners who have suffered a previous episode of aggression British Journal of General Practice 1994 44 390 394 8790650
Hobbs FD Keane UM Aggression against doctors: a review Journal of the Royal Society of Medicine 1996 89 69 72 8683503
Kuhn W Violence in the emergency department. Managing aggressive patients in a high-stress environment Postgraduate Medicine 1999 105 143 148, 154 9924500
Friedman M Danger-violence ahead! [Hebrew] Zman Harefuah 2002 1 8 13
Smeeth L Fletcher AE Improving the response rates to questionnaires BMJ 2002 324 1168 1169 12016167 10.1136/bmj.324.7347.1168
Al-Sahlawi KS Zahid MA Shahid AA Hatim M Al-Bader M Violence against doctors: 1. A study of violence against doctors in accident and emergency departments European Journal of Emergency Medicine 1999 6 301 304 10646917
Zahid MA Al-Sahlawi KS Shahid AA Awadh JA Abu-Shammah H Violence against doctors: 2. Effects of violence on doctors working in accident and emergency departments European Journal of Emergency Medicine 1999 6 305 309 10646918
|
16102174
|
PMC1190173
|
CC BY
|
2021-01-04 16:31:53
|
no
|
BMC Health Serv Res. 2005 Aug 15; 5:54
|
utf-8
|
BMC Health Serv Res
| 2,005 |
10.1186/1472-6963-5-54
|
oa_comm
|
==== Front
BMC ImmunolBMC Immunology1471-2172BioMed Central London 1471-2172-6-171602662710.1186/1471-2172-6-17Methodology ArticleImpact of cryopreservation on tetramer, cytokine flow cytometry, and ELISPOT Maecker Holden T [email protected] James [email protected] Sonny [email protected] Smita A [email protected] Vernon C [email protected] Janice K [email protected] Kristine [email protected] Jennie C [email protected] Amanda [email protected] Timothy M [email protected] Michael A [email protected] H Kim [email protected] Corazon [email protected] Donna P [email protected] Mary L [email protected] BD Biosciences, San Jose, USA2 Southwest Oncology Group Statistical Center at Fred Hutchinson Cancer Research Center, Seattle, USA3 Beckman-Coulter, San Diego, USA4 Departments of Surgery, Medicine, Pathology, and Immunology, and Duke Comprehensive Cancer Center, Duke University Medical Center, Durham, USA5 Tumor Vaccine Group, Division of Oncology, University of Washington, Seattle, USA2005 18 7 2005 6 17 17 26 1 2005 18 7 2005 Copyright © 2005 Maecker et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Cryopreservation of PBMC and/or overnight shipping of samples are required for many clinical trials, despite their potentially adverse effects upon immune monitoring assays such as MHC-peptide tetramer staining, cytokine flow cytometry (CFC), and ELISPOT. In this study, we compared the performance of these assays on leukapheresed PBMC shipped overnight in medium versus cryopreserved PBMC from matched donors.
Results
Using CMV pp65 peptide pool stimulation or pp65 HLA-A2 tetramer staining, there was significant correlation between shipped and cryopreserved samples for each assay (p ≤ 0.001). The differences in response magnitude between cryopreserved and shipped PBMC specimens were not significant for most antigens and assays. There was significant correlation between CFC and ELISPOT assay using pp65 peptide pool stimulation, in both shipped and cryopreserved samples (p ≤ 0.001). Strong correlation was observed between CFC (using HLA-A2-restricted pp65 peptide stimulation) and tetramer staining (p < 0.001). Roughly similar sensitivity and specificity were observed between the three assays and between shipped and cryopreserved samples for each assay.
Conclusion
We conclude that all three assays show concordant results on shipped versus cryopreserved specimens, when using a peptide-based readout. The assays are also concordant with each other in pair wise comparisons using equivalent antigen systems.
==== Body
Background
An increasing number of experimental vaccines are being developed for diseases in which cellular immunity is likely to be required for protection [1]. These include HIV [2], hepatitis C [3], malaria [4], and cancer [5]. In these settings, immunological monitoring of antigen-specific T cell responses is likely to be an important part of vaccine assessment [6-10]. Unfortunately, surrogate markers of protection have not been established for vaccines that induce cellular immunity, although correlations between T cell IFNγ production and clinical responses have been reported in small numbers of patients [11,12] and in mouse models [13-17].
Even in the absence of surrogate markers of protection, the degree to which vaccines can induce T cell responses can be taken as a measure of immunogenicity, or potency, and as such can be used to compare different vaccine candidates. It is likely that inducing a particular level of antigen-specific T cell response to a vaccine will be necessary, though perhaps not sufficient, for vaccine efficacy of either prophylactic or therapeutic vaccines [9,10].
Traditional assays of cellular immunity have been bulk assays, including proliferation assays measuring 3H-thymidine incorporation [18] or cytotoxicity assays measuring 51Cr release [19]. These are increasingly being replaced by single-cell assays, such as MHC-peptide tetramer staining [20], cytokine flow cytometry (CFC) – also known as intracellular cytokine staining (ICS) [21,22], and enzyme-linked immunospot (ELISPOT) [23]. These assays tend to be more quantitative, in that they report a fraction of T cells or PBMC that bind a particular MHC-peptide combination or that produce a particular cytokine in response to an antigen.
Comparisons of tetramer, CFC, and/or ELISPOT have been published [12,24-31], but the relative effect of PBMC cryopreservation on each of these assays has not been examined in a side-by-side fashion. It is known that cryopreservation can negatively impact functional responses [32-35], particularly to nominal antigens. One effect of cryopreservation appears to be the loss of antigen processing capability, as measured by a disproportionate loss of responses to protein antigens, as compared to peptides [36]. There is also a relationship between viability post-thawing and capacity for functional responses [37].
The method of cryopreservation can have a tremendous impact upon viability and function [37-39]. Nevertheless, some authors have reported equivalent results in ELISPOT for fresh and frozen samples, when using an optimized protocol [39-42]. For the present study, an optimized cryopreservation protocol was developed (Disis et al., manuscript submitted) [see Additional file 1]. Two factors that had a positive impact upon cell viability and recovery, and which were incorporated into this optimized protocol, included: (1) the use of human serum albumin as a protein source in the freezing medium, and (2) the use of warmed medium for initial dilution of cells after thawing.
Using this optimized cryopreservation protocol, the present study was conducted in order to determine: (1) the correlation of fresh and cryopreserved results for tetramer, CFC, and ELISPOT assays; (2) the sensitivity and specificity of each assay using CMV seropositive and CMV seronegative donors; and (3) the inter-assay correlations using fresh versus optimally cryopreserved cells. Fresh and cryopreserved PBMC from leukapheresed healthy donors were overnight shipped, in blinded fashion, to three different laboratories, each of which was an experienced practitioner of one of the three assays. These laboratories reported results for their assay to a statistical core, where the results were compiled and statistical analyses carried out.
Results
1. Fresh versus cryopreserved assay correlations
PBMC derived from leukapheresis of 20 CMV seropositive and 21 CMV seronegative donors were analyzed as fresh samples (overnight shipped) or frozen samples (cryopreserved then shipped on dry ice). Cells were processed and analyzed in a blinded fashion by tetramer staining, CFC, and ELISPOT. Representative data for each assay is shown in Figure 1. Correlations of fresh versus frozen data were performed, and are shown in Figure 2.
Fresh versus frozen tetramer results
20 HLA-A0201+ donors (12 CMV seropositive and 8 CMV seronegative) were analyzed by tetramer staining, using an HLA-A0201 tetramer loaded with CMV pp65495–503 peptide. Both seropositive and seronegative donors were analyzed together to determine whether putative negative results were similar in fresh versus frozen assays, as well as putative positive results. Frequencies of tetramer+ cells in fresh versus frozen samples from these donors were significantly correlated (r = 0.9, 95% C.I.: 0.9–1.0; p < 0.001) (Figure 2A). To be sure that this was not simply based upon the correlation of the negative results alone, the statistics were recalculated on CMV seropositive data only, and the correlation coefficient remained 0.9 (95% C.I.: 0.7–1.0).
To determine the bias between fresh and frozen samples from the CMV seropositive donors, the frequency of tetramer+ cells in frozen samples was subtracted from that of the respective fresh samples (Figure 2B). The median difference in responses was -0.008 (95% C.I.: -0.2 to 0.07). The Wilcoxon signed-rank test was unable to detect a significant bias towards either sample type.
Fresh versus frozen CFC
CD8+ T cell IFNγ production was measured by CFC using two different CMV antigens for stimulation: (1) a mixture of overlapping peptides corresponding to the CMV pp65 protein; and (2) the pp65495–503 peptide. The former stimulus was used for comparison to ELISPOT, while the latter allowed comparison to tetramer staining. CFC responses in fresh and frozen PBMC samples from all donors correlated significantly with both antigens. The estimated correlation coefficient for CMV pp65 peptide mix stimulated samples was 0.6 (95% C.I.: 0.4–0.8) (p < 0.001), and that for CMV-A2 peptide stimulated samples was 0.8, (95% C.I.: 0.6–0.8) (p < 0.001) (Figure 2C). These coefficients did not change noticeably when only CMV seropositive data points were considered (r = 0.7 (95% C.I.: 0.3–0.9) and 0.8 (95% C.I.: 0.4–0.9), respectively). Positive correlation of fresh and frozen samples was also observed for SEB-stimulated samples (r = 0.5, p = 0.002; data not shown).
To determine a bias towards a particular sample type, the difference between responses of fresh and frozen samples from CMV seropositive donors were calculated (Figure 2D). The median difference in responses to CMV-A2 peptide was -0.04 (95% C.I.: -0.22 to 0.00) and the median difference to CMV pp65 peptide mix was 0.13 (95% C.I.: 0.04–0.34). The Wilcoxon signed-rank test was unable to detect a significant bias towards either sample type stimulated with either of the two CMV antigens. However, there was a trend toward higher fresh responses to CMV pp65 peptide mix (p = 0.06).
CFC responses to pp65 peptide mix and SEB in CD4+ T cells were also measured, and showed similar correlation between fresh and frozen samples (r = 0.6 and 0.4, respectively, with p < 0.001 for both stimuli) (data not shown).
Fresh versus frozen ELISPOT
IFNγ producing cells were measured by ELISPOT using either CMV pp65 peptide mix or SEB, and enumerated in replicates of 6 wells. The responses of fresh and frozen samples from all donors correlated significantly for pp65 peptide mix stimulation (r = 0.9, 95% C.I.: 0.8–0.9; p < 0.001) (Figure 2E). The correlation coefficient was somewhat lower when only CMV seropositive data points were considered (r = 0.7, 95% C.I.: 0.4–0.9). Interestingly, no significant correlation was observed for SEB activated samples (data not shown).
Fresh versus frozen bias was determined by subtracting pp65 peptide mix responses of frozen samples from their respective fresh samples, for CMV seropositive donors (Figure 2F). The median difference in responses was -12.5, (95% C.I.: -2 to -34). There was a significant bias toward higher responses in frozen samples (p = 0.04).
2. Sensitivity and specificity measurements
To further characterize the relative performance of the three assays, they were examined for their ability to predict CMV serostatus, using fresh and frozen PBMC. Side-by-side dot plots (Figure 3) show the results on CMV seronegative versus seropositive donors for each assay.
Sensitivity was defined as the proportion of CMV seropositive donors correctly identified using a particular cutoff, while specificity was defined as 1 – the false positive rate at that cutoff. These values vary inversely with each other, depending upon the cutoff value used to classify results as positive or negative. To quantitatively compare assay performance, the highest attainable specificity for a sensitivity of ≥90% (if achievable) was reported for each assay (Table 1).
Sensitivity and specificity of tetramer staining
For tetramer staining on fresh samples (Figure 3A), the median number of tetramer+ cells was 0.02% for CMV seronegative donors, and 0.26% for CMV seropositive donors (n = 8 and 12, respectively). For frozen samples (Figure 3B), these medians were 0.01% and 0.26%, respectively.
A positive/negative cutoff of 0.03% for fresh samples and 0.02% for cryopreserved samples was determined as described above. Neither sample type yielded 100% sensitivity and specificity (Table 1). Additionally, there was no consistent change in sensitivity and specificity between fresh and cryopreserved data sets.
Sensitivity and specificity of CFC
For fresh CFC samples (Figure 3C), the median response was 0.07% for seronegative donors and 0.71% for seropositive donors (n = 20 and 21, respectively). For frozen samples (Figure 3D), these medians were 0.01% and 0.58%, respectively.
For CMV pp65 peptide mix stimulation, positive/negative cutoff values of 0.13% for fresh samples and 0.05% for frozen samples were calculated as described above. For CMV pp65495–503 peptide stimulation, the cutoffs were 0.08% for both fresh and frozen samples. Again, none of these assays reached 100% sensitivity and specificity (Table 1). There was also no consistent change in sensitivity and specificity between fresh and cryopreserved data sets.
Sensitivity and specificity of ELISPOT
For fresh ELISPOT samples (Figure 3E), the median number of SFC was 0 for seronegative donors and 34 for seropositive donors. For frozen samples (Figure 3F), these medians were 1 and 67, respectively. These numbers reiterate the bias in ELISPOT toward higher responses in frozen versus fresh samples in this study.
When positive/negative cutoff values were calculated as above, these were 4 SFC for fresh samples and 16 SFC for frozen samples. As with the other assays, 100% sensitivity and specificity was not reached (Table 1). Also, there was no consistent change in sensitivity and specificity between fresh and cryopreserved data sets.
Another way to compare sensitivity and specificity between assays and formats is to plot the sensitivity versus false positive rate in an ROC plot, and then calculate the area under this curve. The greater the area, the greater the ability of the assay to discriminate positive from negative samples. In comparing these areas (Table 1), no statistically significant differences were found between fresh and frozen samples for any of the assays, or between CFC and ELISPOT, or CFC and tetramer (all p values >0.05). This is despite the fact that tetramer and CFC results were calculated only for defined subsets of cells, whereas all PBMC were used in ELISPOT. Better discrimination might be obtained for CFC if CD4+ T cell responses were also included, or for tetramer if class II and/or additional class I epitope tetramers had been used. In fact, when CFC responses for pp65 peptide mix were recalculated to include all IFNγ+ PBMC, not just CD8+ T cells, the area under the ROC curve was, on average, higher (0.890 for fresh, 0.915 for frozen PBMC), but still not significantly different from ELISPOT.
3. Inter-assay correlations
Two-way correlations were performed between tetramer staining and CFC, and between CFC and ELISPOT (Figure 4). CMV pp65495–503-specific responses obtained using CFC were compared to HLA-A0201 pp65495–503 tetramer staining for all 20 HLA-A0201+ donors in the study (12 CMV seropositive, 8 CMV seronegative). CMV pp65 peptide mix-specific responses obtained by CFC and ELISPOT were compared for all 20 CMV seropositive and 21 CMV seronegative donors in the study.
Tetramer staining versus CFC
These assays correlated significantly with each other for both fresh and frozen samples (Figures 4A and 4B). The estimated correlation coefficient was 0.9 for both fresh and frozen samples (95% C.I.: 0.8–1.0 for both) (p < 0.001 for both). These were the strongest inter-assay correlations observed in this study.
CFC versus ELISPOT
A significant correlation between these two assays was observed using both fresh and frozen PBMC samples (data not shown). The estimated correlation coefficient was 0.5 (95% C.I.: 0.2–0.7) (p = 0.001) for fresh samples, and 0.7 (95% C.I.: 0.5–0.8) (p < 0.001) for frozen samples. However, when only CMV seropositive donors were considered, the p values became non-significant for both fresh and frozen samples. Thus, CFC and ELISPOT were less tightly correlated than tetramer and CFC. This is in agreement with another published report comparing ELISPOT and CFC [29].
The scales and readouts of CFC and ELISPOT are very different (%IFNγ+ CD8+ T cells versus SFC per 105 PBMC). To correct for these differences, the CFC results were expressed as the total number of IFNγ+ cells per 105 PBMC, without gating on CD3 or CD8. This mimicked the readout of the ELISPOT assay, which reported positive events per 105 PBMC (the number of cells plated in each ELISPOT well). The estimated correlation coefficients under these conditions were 0.7 for both fresh (95% CI: 0.4–0.8) and frozen (95% CI: 0.5–0.8) samples (p < 0.001 both) (Figure 4C and 4D). When only CMV seropositive samples were considered, the correlation coefficients became 0.8 (95% CI: 0.5–0.9) and 0.4 (95% CI: 0.0–0.7), respectively (p < 0.001 and p = 0.06, respectively). Thus, expression of the CFC results in this format appeared to improve the correlation. This may reflect the response of non-CD8+ cells in these assays. Note also that when CFC results were plotted in this way, there were on average several-fold higher responses with CFC versus ELISPOT (as represented by most points being above the diagonal in Figures 4C and 4D).
Discussion
This study represents a comprehensive analysis of the effect of cryopreservation (using an optimized cryopreservation protocol) on tetramer, CFC, and ELISPOT assays using peptide-based antigens. Each assay was individually optimized by a laboratory experienced in that technique, to ensure the best possible performance for each. For example, costimulatory antibodies (CD28 and CD49d) were used in CFC but not ELISPOT, despite the fact that addition of CD28 antibody can improve ELISPOT sensitivity [43]. In our hands, occasional donors developed unacceptably high ELISPOT backgrounds (>100 spots per 106 PBMC) with the use of CD28 costimulation (data not shown), so it was not used for that assay, although it was used for CFC. While there are shortcomings of this study design, we felt that this provided the fairest and most robust way to compare these assays. The assays were compared by correlation of results for fresh and frozen samples; by analyzing sensitivity and specificity of each assay on fresh and frozen samples; and by determining inter-assay correlations using fresh and frozen samples. The overall findings are summarized in Table 2.
None of the three assays showed a significant reduction of signal in frozen cells relative to fresh cells. The fresh-to-frozen correlation was strongest for tetramer staining, which does not rely on cell function, then CFC and ELISPOT. Compared to CMV responses, SEB responses were less well correlated in fresh and frozen samples using CFC, and not at all correlated using ELISPOT. The reasons for this are not clear; however, they may stem from the relative affinities of T cells responding to CMV peptides versus SEB. It is known that T cells bearing a number of different TCR Vβ sequences can participate, to varying degrees, in the SEB response. The participation of different Vβ families is related to their affinity for SEB. Low affinity Vβ responses may be preferentially lost upon cryopreservation. The differential representation of these Vβ subsets in different donors may thus lead to inconsistencies in the correlation of fresh to cryopreserved responses for SEB. CMV responses may be of generally uniform and higher affinity, as suggested by their typically bright, clustered IFNγ staining (compared to SEB responses, where IFNγ staining tends to be more of a smear). If this is true, it may also suggest that other low-affinity responses, such as those to tumor antigens, may be more susceptible to loss upon cryopreservation; this needs to be tested.
Rather than compare cryopreserved PBMC to fresh, same-day activated PBMC, the former were compared to fresh, overnight-shipped PBMC from leukapheresed donors. Since some functional degradation undoubtedly occurs with overnight shipping, this is not ideal in terms of assessing the total signal loss due to cryopreservation. However, the reality of large clinical studies is that PBMC will in all likelihood be cryopreserved and/or shipped overnight to a laboratory that does immunological monitoring. Thus, comparison of these two conditions represents a comparison of two likely scenarios for handling of PBMC samples in clinical trials. The current results imply that there is unlikely to be a pronounced difference in results with any of the three assays when using either of these conditions. It is unknown whether a detectable loss of signal would be observed if PBMC were subjected to both overnight shipping and then cryopreservation. Of course, all of these data assume that reasonable care is taken in sample cryopreservation and shipping, which were optimized for this study (Disis et al., manuscript submitted). Also, the stability of healthy donor cells, as used in this study, may be superior to those from certain disease states, e.g., HIV or tumor patients. Finally, the use of shipped PBMC (rather than whole blood) may have improved the results seen in this study; thus our results should not be taken as indicative of what would be achieved if whole blood samples were shipped overnight.
Because this study used peptide-based antigens (and SEB), drastic loss of functional responses as reported for whole-protein antigens [36] were not seen. The current results should not be interpreted to apply to non-peptide antigens, since antigen-processing capabilities are preferentially lost upon cryopreservation.
We defined sensitivity and specificity on the basis of reactivity with donors who were CMV seropositive, and lack of reactivity with donors who were CMV seronegative. This comparison is not ideal, since there are reports of non-correlation of serological and T cell responses to CMV [44]. In particular, assays examining a single epitope response (e.g., tetramer and CFC for pp65495–503) may underestimate CMV responders, due to immunodominance hierarchies [45]. HLA-A2-restricted responses to pp65495–503 are known to be suppressed in individuals co-expressing HLA-B7 [46], and such individuals were not excluded from this study. However, CFC assays using pp65 peptide mix were found in at least one previous study to differentiate CMV seropositive and seronegative donors with 100% sensitivity and specificity, although the sample size was small (15 seropositive and 14 seronegative donors) [47]. All three assays were recently compared in a study of fresh PBMC from 21 CMV seropositive and 20 CMV seronegative healthy donors, with a sensitivity of 87.5% and specificity of 100% for each assay [48].
None of the three assays attained 100% sensitivity and specificity using either fresh or cryopreserved PBMC in the present study. ELISPOT, especially on cryopreserved PBMC, showed slightly greater sensitivity (for specificity ≥90%) than did CFC or tetramer (although the difference was not statistically significant). This was partly due to the reactivity of one or two seronegative donors in the CFC and tetramer assays. This reactivity was more pronounced in the assays with fresh PBMC, but was still present in the cryopreserved PBMC from the same donors. In the case of CFC, the cryopreserved PBMC assay was repeated on additional cells and still showed a similar level of reactivity, suggesting that it was not due to a technical error. It is possible that these donors represented true discordant responses between serological and cellular assays.
The quantitative comparison of tetramer and CFC resulted in the tightest correlation. This could be related to the fact that both of these assays use the same readout platform (flow cytometry). The correlation of CFC and ELISPOT was less precise, and was not statistically significant when CMV seronegative donors were excluded. The correlation appeared tighter when CFC results were expressed as a proportion of all PBMC. This could be due to variable proportions of CD4+ T cells contributing to the response to pp65 peptide mix. It is also possible that non-T cells contribute to this response, although this was not directly assessed in this study.
Conclusion
We conclude that tetramer, CFC, and ELISPOT assays can be performed on optimally cryopreserved PBMC with minimal or no loss of signal when compared to fresh, overnight-shipped PBMC. The assays correlate significantly in direct comparisons using the same antigen systems, whether fresh or cryopreserved PBMC are used. The strongest correlations of fresh and cryopreserved PBMC are seen with tetramer and CFC assays; and these two assays also correlate most strongly with each other. All three assays showed roughly similar sensitivity and specificity in discriminating CMV seropositive from seronegative donors. The strong correlation of tetramer and CFC assays in fresh and cryopreserved cells, along with their multiparameter information content, make them ideal choices for immune monitoring assays.
Methods
PBMC isolation and processing
PBMC from leukapheresis (obtained from healthy donors without cytokine mobilization) were isolated using Ficoll gradient separation. Briefly, 5 ml of leukapheresis product were aliquoted into 50 ml conical tubes (BD Falcon, Franklin Lakes, NJ) washed once by adding HBSS (Gibco Invitrogen Corporation, Grand Island, NY) and centrifuged for 10 minutes at 280 × G. The pelleted cells were resuspended and 40 ml of HBSS were added. Ten ml of Ficoll Paque (Amersham Biosciences, Piscataway, NJ) were carefully underlayed and the tubes centrifuged at 400 × G for 40 minutes. The buffy coat was collected and washed twice with HBSS. Viability was assessed using 0.4% Trypan blue (Sigma, St. Louis, MO). For fresh-shipped specimens, 2 × 107 viable lymphocytes were resuspended in 50 ml RPMI+10% fetal bovine serum and shipped overnight at ambient temperature in a 50 ml conical tube packed in an insulated foam container. Fresh-shipped PBMC were centrifuged as soon as received and the assays set up as described below.
Cryopreservation [see Additional file 1]
To cryopreserve PBMC, 2X freezing media was first prepared, containing 20% DMSO in RPMI (Sigma Chemical Co., St. Louis, MO) containing 12.5% human serum albumin (HSA) (Gemini Bioproducts, Woodland, CA), and cooled on ice for a minimum of 30 minutes. Ficolled PBMC at 2 × 107 viable lymphocytes/ml were resuspended in cooled RPMI+12.5% HSA with no DMSO. An equal volume of chilled 2X freezing media was added to the cell suspension dropwise, while gently swirling the tube. One ml of this cell suspension was aliquoted into each cryovial (Sarstedt, Inc., Newton, NC). Once aliquoted, cryovials were placed on ice and then transferred into a freezing container (Nalgene, Rochester, NY), and stored at -80°C for 24 hours. Cryovials were then transferred into liquid nitrogen for long-term storage. After 30 days, cryovials were overnight shipped on dry ice to the recipient laboratories.
Thawing [see Additional file 1]
Cryopreserved PBMC were stored at -80°C until thawing to set up the assays. Cryopreserved cells were thawed and slowly diluted with 8 ml of warm RPMI+10% fetal bovine serum+antibiotics (cRPMI-10, all components from Sigma). The cells were centrifuged for 7 minutes at 250 × G, then resuspended as described below for each assay. Viability and recovery were checked using Trypan blue, and were >80% and >50%, respectively, in all samples.
MHC class I-peptide tetramer staining
Fresh and frozen PBMC from 12 CMV seropositive and 8 CMV seronegative patients were screened with MHC tetramer composed of HLA-A*0201 monomers carrying the CMV pp65495–503 peptide epitope (NLVPMVATV). For flow cytometry analysis, the Multiple Antibody Single Color protocol (iMASC, Beckman Coulter Inc., Fullerton, CA) was used. Briefly, ten μl each of CD4, CD13, and CD19 antibodies conjugated to PE-Cy5 (PC5) were added to sample tubes in order to exclude CD4 T cells, granulocytes, and B cells from analysis. In addition, 10 μl of CD8 FITC and 10 μl tetramer PE were added, followed by 1 × 106 PBMC in 100 μl of flow cytometry buffer (HBSS containing 0.1% bovine serum albumin, 0.02% sodium azide). Samples were incubated for 30 minutes at room temperature followed by a wash with flow cytometry buffer and fixation in 1% formaldehyde. Samples were run on a BD FACSCalibur flow cytometer that was set to acquire 30,000 CD8+ T cells. Analysis was performed using CellQuest software (BD Biosciences, San Jose, CA) and gating was done to accept CD8+ /tetramer+ cells and to exclude PC5-positive cells.
Cytokine flow cytometry
CFC assays were performed according to a previously published method [49,50]. 200 μl containing 2 × 106 PBMC in cRPMI-10 medium were plated per well in 96-well round-bottom plates. For cryopreserved PBMC, the thawed cells were then rested at 37°C, 7% CO2 overnight. For both fresh and cryopreserved PBMC, activation reagents (stimulus + brefeldin A) were added in a volume of 20 μl per 200 μl of cell suspension per well and the cells were then incubated at 37°C for 6 hours. Stimuli included CMV pp65 peptide mix (BD Biosciences; used at a final concentration of 1.7 μg/ml/peptide); CMV pp65495–503 peptide (SynPep Corp., Dublin, CA; used at a final concentration of 10 μg/ml); and SEB (Sigma; used at a final concentration of 1 μg/ml). All samples received a final concentration of 1 μg/ml each of CD28+CD49d costimulatory antibodies and 10 μg/ml of brefeldin A (both from BD Biosciences). After 6 hours incubation, the cells were treated with 2 mM final concentration of EDTA for 15 minutes at room temperature, then fixed with FACS Lysing Solution (BD Biosciences) and stored at -80°C. When ready to stain, the frozen plates were thawed at 37°C and processed further with FACS Permeabilizing Solution 2 (BD Biosciences) followed by staining with IFNγ FITC/CD69 PE/CD8 PerCPCy5.5/CD3 APC (BD Biosciences) for 1 hour at room temperature. Plates were washed and cells resuspended in 1% paraformaldehyde in PBS.
Samples were acquired within 24 hours of staining using a FACSCalibur flow cytometer with a Multiwell Autosampler, using Multiwell Plate Manager and CellQuest Pro software (BD Biosciences). 40,000 CD3+CD8+ lymphocytes were collected per sample. A "response region" was set around double-positive cells in a gated dot plot displaying CD69 versus IFNγ staining from an SEB-stimulated sample. This response region was then applied to all samples to determine the percentage of cytokine-positive cells. Data were reported as the net percent of CD3+CD8+ lymphocytes that were IFNγ+ after subtracting the response of unstimulated samples.
ELISPOT
PBMC were assayed for IFNγ production in the presence of CMV pp65 peptide mix (BD Biosciences), SEB, and media in replicates of 6. Multiscreen-HA 96-well plates (Millipore, Bedford, MA) were coated overnight at 4°C with 100 μl/well of 10 μg/ml mouse anti-human IFNγ mAb 7-D1K (diaPharma Group, Inc., West Chester, OH) in Dulbecco's Phosphate Buffered Saline (DPBS) (Gibco Invitrogen). The plates were washed 3 times for 5 minutes each with 150 μl DPBS/well and blocked with 150 μl/well of RPMI-1640, 10% human AB serum, 25 mM HEPES, 100 U/ml penicillin, 100 μg/ml streptomycin, and 2 mM L-glutamine for 1 hour at 37°C in 5% CO2. PBMC were plated at 100,000 per well with 1:800 of CMV pp65 peptide mix (approximately 1.75 μg/ml of each peptide), 100 ng/ml of SEB, or media in a total volume of 200 μl/well for 18–24 hours at 37°C in 5% CO2.
The plates were washed with 0.05% Tween/DPBS using a Tecan 96PW plate washer (Tecan, Research Triangle Park, NC). A solution of 100 μl of mouse anti-human IFNγ biotinylated mAb 7-B6-1 (diaPharma) at 1 μg/ml in DPBS was added to each well and the plates incubated for 2 hours at 37°C, 5% CO2. Vectastain ABC Peroxidase (Vector Labs, Inc., Burlingame, CA) was added at 100 μl/well for 1 hour at room temperature after washing with 0.05% Tween/DPBS using the Tecan plate washer. The plates were washed for the last time with 0.05% Tween/DPBS followed by DPBS. Color was developed using 100 μl/well of 3-amino-9-ethyl-carbazole [AEC] (Sigma) reconstituted in an acetate buffer for 4 minutes at room temperature in the dark. Color development was stopped with deionized water. Basins were removed and the membranes dried overnight in the dark. Membranes were attached to sealing tape (Millipore, Bedford, MA) and the number of spots per well was determined using a KS ELISPOT Automated Reader System with KS ELISPOT 4.2 Software (Carl Zeiss, Inc., Thornwood, NY). The mean number of spots from the six replicate wells at each dilution was reported for each antigen. The analyses in this paper were based on the wells containing 1 × 105 responder PBMC, which is the dilution that yielded the highest ratio of spots/PBMC (data not shown).
Statistical analyses
All samples tested were included in the analysis, as no attempt was made to exclude outliers. Tests of correlations between fresh and frozen samples were performed using the Pearson Product Moment Correlation Coefficient on the natural logarithm of responses. This transformation was performed to correct for observed skewness in the data, since the statistical test assumes a bivariate normal distribution. However, the significance of the correlations was largely unchanged when untransformed data was used. If the net response equaled zero, a small constant was added (0.004 for % positive and 1 for cells/105 PBMC) prior to computing the natural log. The Wilcoxon Signed-Rank test was performed for differences between paired fresh and frozen samples. Tests of correlations between assays were performed using the Pearson Product Moment Correlation Coefficient after natural logarithmic transformation and adjustment for zeroes.
For each assay and antigen combination, the operating characteristics were summarized in terms of the sensitivity and false positive rate (1-specificity) for cut-off values of net response of both fresh and frozen samples. The sensitivity and false positive rate were defined as the proportion of correctly identified CMV seropositive samples and incorrectly identified seronegative samples, respectively, with an assay response exceeding each cut-off value. The ROC curve was constructed as a plot of the false positive test rate versus sensitivity for all cut-off values in the range of assay responses observed. For each ROC curve, the sensitivity and specificity was reported for the cut-off that maximized specificity subject to the constraint that sensitivity ≥90%; or alternatively, if the sensitivity was bounded below 90%, at the specificity corresponding to the maximum sensitivity. Tests for differences in the areas under the ROC curves were performed using the nonparametric test for correlated data of Delong et al [51]. The ROC analysis was performed with Stata version 7 software (StataCorp LP, College Station, TX). All other statistical analyses were performed with SAS version 9 software (SAS Institute, Cary, NC).
Authors' contributions
HTM wrote the manuscript with input from SAG, DPA, JM, TMC, and TKR. JM and DPA compiled the data and did statistical analysis. SB, JKP, AS, JCC, and CD processed and analyzed the samples. HTM, SAG, VCM, TMC, MAM, HKL, TKR, and MLD planned and supervised the study and all authors reviewed and edited the manuscript
Supplementary Material
Additional File 1
Protocol for Isolation, Cryopreservation, and Thawing of PBMC Optimized protocol used in this study for cryopreservation and thawing of PBMC
Click here for file
Acknowledgements
This work was supported by Public Health Service grant #U54 CA90818 from the National Cancer Institute.
Figures and Tables
Figure 1 Representative data from the three assays on the same CMV seropositive donor PBMC. (A) Tetramer staining after gating as described in the Methods. Background seen in staining of fresh cells with negative tetramer was caused by streptavidin-PE. Frozen cells were stained with a different lot of negative tetramer. (B) CFC staining after gating as described in the Methods. (C) ELISPOT data as analyzed on the Automated Reader System as described in the Methods. Error bars represent the S.D. of 6 replicates.
Figure 2 Correlation of fresh-shipped and cryopreserved samples in all three assays. (A, C, E) Correlation graphs of fresh-shipped versus cryopreserved PBMC samples for tetramer, CFC, and ELISPOT, respectively. Correlation coefficients (r) are shown for all data as well as for CMV seropositive donors only. Open symbols represent CMV seronegative donors; closed symbols, seropositive donors (*, CMV serostatus unknown). The diagonal line represents the line of perfect agreement between the assays. (B, D, F) Within-donor differences are shown for fresh-shipped versus cryopreserved responses. Bars represent the median difference of all donors. All statistics are based on a natural logarithm transformation, which was done to better approximate a bivariate normal distribution.
Figure 3 Results of the three assays on CMV seronegative versus seropositive donors. (A and B) Tetramer results on fresh-shipped and cryopreserved PBMC samples, respectively. (C and D) CFC results on fresh-shipped and cryopreserved PBMC samples, respectively. (E and F) ELISPOT results on fresh-shipped and cryopreserved PBMC samples, respectively. The dotted line represents the suggested cutoff based upon maximum specificity for sensitivity ≥90%, or that which obtains maximum sensitivity if maximum sensitivity <90%.
Figure 4 Inter-assay correlations. (A and B) Correlation of tetramer staining and CFC in fresh-shipped and cryopreserved PBMC samples, respectively. (C and D) Correlation of ELISPOT and CFC in fresh-shipped and cryopreserved PBMC samples, respectively, with CFC results reported as number of IFNγ+ cells per 105 PBMC. In panels C-D, the mean of 6 replicates is shown for all ELISPOT data. Correlation coefficients (r) are shown for all data as well as for CMV seropositive donors only. Open symbols represent CMV seronegative donors; closed symbols, seropositive donors (*, CMV serostatus unknown). The diagonal line represents the line of perfect agreement between the assays.
Table 1 Cutoffs, sensitivity, and specificity from ROC curves
Assay Antigen Sample Type Cutoff Point1 Sensitivity Specificity Area Under ROC Curve2
Tetramer (% +) CMV pp65 A2 peptide Fresh 0.05 92% 88% 0.906+/-0.076
Cryo. 0.02 83% 100% 0.875+/-0.086
CMV pp65 peptide mix Fresh 0.13 90% 71% 0.819+/-0.075
CFC (% +) Cryo. 0.05 90% 76% 0.920+/-0.046
CMV pp65 A2 peptide Fresh 0.08 75% 100% 0.844+/-0.084
Cryo. 0.08 83% 88% 0.891+/-0.074
ELISPOT (# SFC) CMV pp65 peptide mix Fresh 4 95% 94% 0.982+/-0.018
Cryo. 16 90% 100% 0.985+/-0.015
1Cut-off is that which achieves maximum specificity for sensitivity ≥90%, or that which obtains maximum sensitivity if maximum sensitivity <90%.)
2Values are given +/- standard error.
Table 2 Summary of assay characteristics
Assay Type of assay Readout Fresh to frozen correlation Sensitivity and specificity1 Inter-assay correlations
Tetramer Phenotypic 4-color flow cytometry r = 0.9 >70%2 r = 0.9
CFC Functional 4-color flow cytometry r = 0.6 (CMV)
r = 0.4 (SEB) >70%2 r = 0.9
r = 0.7
ELISPOT Functional Plate reader r = 0.6 (CMV)
NS (SEB) >90% r = 0.7
1For both fresh and frozen samples with all peptide-based stimuli.
2Potentially higher if multiple tetramers or CD4 and CD8 responses are considered.
==== Refs
Zinkernagel RM Hengartner H On immunity against infections and vaccines: credo 2004 Scand J Immunol 2004 60 9 13 15238068
McMichael AJ Hanke T HIV vaccines 1983-2003 Nat Med 2003 9 874 880 12835708
Torresi J Bharadwaj M Jackson DC Gowans EJ Neutralising antibody, CTL and dendritic cell responses to hepatitis C virus: a preventative vaccine strategy Curr Drug Targets 2004 5 41 56 14738217
Moorthy VS Good MF Hill AV Malaria vaccine developments Lancet 2004 363 150 156 14726170
Gilboa E The promise of cancer vaccines Nat Rev Cancer 2004 4 401 411 15122211
Disis ML Maecker HT Clay TM Lyerly HK Chang JCC Lotze MT, Thompson AW Immunologic monitoring to assess immunity to solid tumors Measuring immunity: the immunoloic surrogates handbook 2003 London , Academic Press
Keilholz U Weber J Finke JH Gabrilovich DI Kast WM Disis ML Kirkwood JM Scheibenbogen C Schlom J Maino VC Lyerly HK Lee PP Storkus W Marincola F Worobec A Atkins MB Immunologic monitoring of cancer vaccine therapy: results of a workshop sponsored by the Society for Biological Therapy J Immunother 2002 25 97 138 12074049
Coulie PG van der Bruggen P T-cell responses of vaccinated cancer patients Curr Opin Immunol 2003 15 131 137 12633661
Maecker HT Disis ML The role of immune monitoring in evaluating cancer immunotherapy Immunotherapy of Cancer Cancer Drug Discovery and Development 2005 Totowa, NJ , Humana Press
Maecker HT Maino VC T cell immunity to HIV: defining parameters of protection Current HIV Research 2003 1 249 259 15043207
Belli F Testori A Rivoltini L Maio M Andreola G Sertoli MR Gallino G Piris A Cattelan A Lazzari I Carrabba M Scita G Santantonio C Pilla L Tragni G Lombardo C Arienti F Marchiano A Queirolo P Bertolini F Cova A Lamaj E Ascani L Camerini R Corsi M Cascinelli N Lewis JJ Srivastava P Parmiani G Vaccination of metastatic melanoma patients with autologous tumor-derived heat shock protein gp96-peptide complexes: clinical and immunologic findings J Clin Oncol 2002 20 4169 4180 12377960
Sun P Schwenk R White K Stoute JA Cohen J Ballou WR Voss G Kester KE Heppner DG Krzych U Protective immunity induced with malaria vaccine, RTS,S, is linked to Plasmodium falciparum circumsporozoite protein-specific CD4(+) and CD8(+) T cells producing IFN-gamma J Immunol 2003 171 6961 6967 14662904
Perez-Diez A Spiess PJ Restifo NP Matzinger P Marincola FM Intensity of the vaccine-elicited immune response determines tumor clearance J Immunol 2002 168 338 347 11751979
Gurunathan S Prussin C Sacks DL Seder RA Vaccine requirements for sustained cellular immunity to an intracellular parasitic infection Nat Med 1998 4 1409 1415 9846579
Gurunathan S Stobie L Prussin C Sacks DL Glaichenhaus N Fowell DJ Locksley RM Chang JT Wu CY Seder RA Requirements for the maintenance of Th1 immunity in vivo following DNA vaccination: a potential immunoregulatory role for CD8+ T cells J Immunol 2000 165 915 924 10878366
Stobie L Gurunathan S Prussin C Sacks DL Glaichenhaus N Wu CY Seder RA The role of antigen and IL-12 in sustaining Th1 memory cells in vivo: IL-12 is required to maintain memory/effector Th1 cells sufficient to mediate protection to an infectious parasite challenge Proc Natl Acad Sci U S A 2000 97 8427 8432 10890924
Mendez S Gurunathan S Kamhawi S Belkaid Y Moga MA Skeiky YA Campos-Neto A Reed S Seder RA Sacks D The potency and durability of DNA- and protein-based vaccines against Leishmania major evaluated using low-dose, intradermal challenge J Immunol 2001 166 5122 5128 11290794
Gehrz RC Knorr SO Characterization of the role of mononuclear cell subpopulations in the in vitro lymphocyte proliferation assay Clin Exp Immunol 1979 37 551 557 315848
McCoy JL Herberman RB Rosenberg EB Donnelly FC Levine PH Alford C 51 Chromium-release assay for cell-mediated cytotoxicity of human leukemia and lymphoid tissue-culture cells Natl Cancer Inst Monogr 1973 37 59 67 4123455
Altman JD Moss PAH Goulder PJR Barouch DH McHeyzer-Williams MG Bell JI McMichael AJ Davis MM Phenotypic analysis of antigen-specific T lymphocytes Science 1996 274 UNITED STATES 94 96 8810254
Waldrop SL Pitcher CJ Peterson DM Maino VC Picker LJ Determination of antigen-specific memory/effector CD4+ T cell frequencies by flow cytometry: evidence for a novel, antigen-specific homeostatic mechanism in HIV-associated immunodeficiency J Clin Invest 1997 99 UNITED STATES 1739 1750 9120019
Suni MA Picker LJ Maino VC Detection of antigen-specific T cell cytokine expression in whole blood by flow cytometry J Immunol Methods 1998 212 NETHERLANDS 89 98 9671156
Czerkinsky C Andersson G Ekre HP Nilsson LA Klareskog L Ouchterlony O Reverse ELISPOT assay for clonal analysis of cytokine production. I. Enumeration of gamma-interferon-secreting cells J Immunol Methods 1988 110 29 36 3131436
Kuzushima K Hoshino Y Fujii K Yokoyama N Fujita M Kiyono T Kimura H Morishima T Morishima Y Tsurumi T Rapid determination of Epstein-Barr virus-specific CD8(+) T-cell frequencies by flow cytometry Blood 1999 94 3094 3100 10556194
Moretto WJ Drohan LA Nixon DF Rapid quantification of SIV-specific CD8 T cell responses with recombinant vaccinia virus ELISPOT or cytokine flow cytometry Aids 2000 14 2625 2627 11101083
Asemissen AM Nagorsen D Keilholz U Letsch A Schmittel A Thiel E Scheibenbogen C Flow cytometric determination of intracellular or secreted IFNgamma for the quantification of antigen reactive T cells J Immunol Methods 2001 251 101 108 11292486
Pahar B Li J Rourke T Miller CJ McChesney MB Detection of antigen-specific T cell interferon gamma expression by ELISPOT and cytokine flow cytometry assays in rhesus macaques J Immunol Methods 2003 282 103 115 14604545
Whiteside TL Zhao Y Tsukishiro T Elder EM Gooding W Baar J Enzyme-linked immunospot, cytokine flow cytometry, and tetramers in the detection of T-cell responses to a dendritic cell-based multipeptide vaccine in patients with melanoma Clin Cancer Res 2003 9 641 649 12576430
Karlsson AC Martin JN Younger SR Bredt BM Epling L Ronquillo R Varma A Deeks SG McCune JM Nixon DF Sinclair E Comparison of the ELISPOT and cytokine flow cytometry assays for the enumeration of antigen-specific T cells J Immunol Methods 2003 283 141 153 14659906
Heeger PS Greenspan NS Kuhlenschmidt S Dejelo C Hricik DE Schulak JA Tary-Lehmann M Pretransplant frequency of donor-specific, IFN-gamma-producing lymphocytes is a manifestation of immunologic memory and correlates with the risk of posttransplant rejection episodes J Immunol 1999 163 2267 2275 10438971
Helms T Boehm BO Asaad RJ Trezza RP Lehmann PV Tary-Lehmann M Direct visualization of cytokine-producing recall antigen-specific CD4 memory T cells in healthy individuals and HIV patients J Immunol 2000 164 3723 3732 10725731
Oldham RK Dean JH Cannon GB Ortaldo JR Dunston G Applebaum F McCoy JL Djeu J Herberman RB Cryopreservation of human lymphocyte function as measured by in vitro assays Int J Cancer 1976 18 145 155 133996
Weinberg A Wohl DA Brown DG Pott GB Zhang L Ray MG van der Horst C Effect of cryopreservation on measurement of cytomegalovirus-specific cellular immune responses in HIV-infected patients J Acquir Immune Defic Syndr 2000 25 109 114 11103040
Costantini A Mancini S Giuliodoro S Butini L Regnery CM Silvestri G Montroni M Effects of cryopreservation on lymphocyte immunophenotype and function J Immunol Methods 2003 278 145 155 12957403
Weinberg A Betensky RA Zhang L Ray G Effect of shipment, storage, anticoagulant, and cell separation on lymphocyte proliferation assays for human immunodeficiency virus-infected patients Clin Diagn Lab Immunol 1998 5 804 807 9801338
Maecker HT Dunn HS Suni MA Khatamzas E Pitcher CJ Bunde T Persaud N Trigona W Fu TM Sinclair E Bredt BM McCune JM Maino VC Kern F Picker LJ Use of overlapping peptide mixtures as antigens for cytokine flow cytometry J Immunol Methods 2001 255 27 40 11470284
Weinberg A Zhang L Brown D Erice A Polsky B Hirsch MS Owens S Lamb K Viability and functional activity of cryopreserved mononuclear cells Clin Diagn Lab Immunol 2000 7 714 716 10882680
Kleeberger CA Lyles RH Margolick JB Rinaldo CR Phair JP Giorgi JV Viability and recovery of peripheral blood mononuclear cells cryopreserved for up to 12 years in a multicenter study Clin Diagn Lab Immunol 1999 6 14 19 9874657
Kreher CR Dittrich MT Guerkov R Boehm BO Tary-Lehmann M CD4+ and CD8+ cells in cryopreserved human PBMC maintain full functionality in cytokine ELISPOT assays J Immunol Methods 2003 278 79 93 12957398
Smith JG Liu X Kaufhold RM Clair J Caulfield MJ Development and validation of a gamma interferon ELISPOT assay for quantitation of cellular immune responses to varicella-zoster virus Clin Diagn Lab Immunol 2001 8 871 879 11527795
Gebauer BS Hricik DE Atallah A Bryan K Riley J Tary-Lehmann M Greenspan NS Dejelo C Boehm BO Hering BJ Heeger PS Evolution of the enzyme-linked immunosorbent spot assay for post-transplant alloreactivity as a potentially useful immune monitoring tool Am J Transplant 2002 2 857 866 12392292
Sobota V Bubenik J Indrova M Vlk V Jakoubkova J Use of cryopreserved lymphocytes for assessment of the immunological effects of interferon therapy in renal cell carcinoma patients J Immunol Methods 1997 203 1 10 9134025
Ott PA Berner BR Herzog BA Guerkov R Yonkers NL Durinovic-Bello I Tary-Lehmann M Lehmann PV Anthony DD CD28 costimulation enhances the sensitivity of the ELISPOT assay for detection of antigen-specific memory effector CD4 and CD8 cell populations in human diseases J Immunol Methods 2004 285 223 235 14980436
Zhu J Shearer GM Marincola FM Norman JE Rott D Zou JP Epstein SE Discordant cellular and humoral immune responses to cytomegalovirus infection in healthy blood donors: existence of a Th1-type dominant response Int Immunol 2001 13 785 790 11369706
Betts MR Casazza JP Patterson BA Waldrop S Trigona W Fu TM Kern F Picker LJ Koup RA Putative immunodominant human immunodeficiency virus-specific CD8(+) T- cell responses cannot be predicted by major histocompatibility complex class I haplotype J Virol 2000 74 9144 9151 10982361
Ghanekar SA Nomura LE Suni MA Picker LJ Maecker HT Maino VC Gamma interferon expression in CD8(+) T cells is a marker for circulating cytotoxic T lymphocytes that recognize an HLA A2-restricted epitope of human cytomegalovirus phosphoprotein pp65 Clin Diagn Lab Immunol 2001 8 628 631 11329470
Dunn HS Haney DJ Ghanekar SA Stepick-Biek P Lewis DB Maecker HT Dynamics of CD4 and CD8 T cell responses to cytomegalovirus in healthy human donors J Infect Dis 2002 186 15 22 12089657
Hobeika AC Morse MA Osada T Ghanayem M Niedzwiecki D Barrier R Lyerly HK Clay TM Enumerating antigen-specific T-cell responses in peripheral blood: a comparison of peptide MHC Tetramer, ELISpot, and intracellular cytokine analysis J Immunother 2005 28 63 72 15614046
Suni MA Dunn HS Orr PL deLaat R Sinclair E Ghanekar SA Bredt BM Dunne JF Maino VC Maecker HT Performance of plate-based cytokine flow cytometry with automated data analysis BMC Immunology 2003 4 9 12952557
Maecker HT Hawley TS, Hawley RG Cytokine flow cytometry Flow Cytometry Protocols Meth Molec Biol Walker J 2004 2nd Totowa, NJ , Humana Press 95 107
DeLong ER DeLong DM Clarke-Pearson DL Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach Biometrics 1988 44 837 845 3203132
|
16026627
|
PMC1190174
|
CC BY
|
2021-01-04 16:03:32
|
no
|
BMC Immunol. 2005 Jul 18; 6:17
|
utf-8
|
BMC Immunol
| 2,005 |
10.1186/1471-2172-6-17
|
oa_comm
|
==== Front
BMC ImmunolBMC Immunology1471-2172BioMed Central London 1471-2172-6-181604277610.1186/1471-2172-6-18Research ArticleInhibition of chemokine expression in rat inflamed paws by systemic use of the antihyperalgesic oxidized ATP Fulgenzi Alessandro [email protected]'Antonio Giacomo [email protected] Chiara [email protected] Elena Dal [email protected] Paolo [email protected] Josè S [email protected] Maria Elena [email protected] Università degli Studi di Milano, Istituto di Patologia Generale, via Mangiagalli 31, 20133, Milano, Italy2 Ospedale S. Raffaele, via Olgettina 60, 20100, Milano, Italy3 Medestea Research and Production, via Magenta 43, 10128, Torino, Italy2005 22 7 2005 6 18 18 22 3 2005 22 7 2005 Copyright © 2005 Fulgenzi et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
We previously showed that local use of periodate oxidized ATP (oATP, a selective inhibitor of P2X7 receptors for ATP) in rat paw treated with Freund's adjuvant induced a significant reduction of hyperalgesia Herein we investigate the role of oATP, in the rat paws inflamed by carrageenan, which mimics acute inflammation in humans.
Results
Local, oral or intravenous administration of a single dose of oATP significantly reduced thermal hyperalgesia in hind paws of rats for 24 hours, and such effect was greater than that induced by diclofenac or indomethacin. Following oATP treatment, the expression of the pro-inflammatory chemokines interferon-gamma-inducible protein-10 (IP-10), mon ocyte chemoattractant protein-1 (MCP-1) and interleukin-8 (IL-8) within the inflamed tissues markedly decreased on vessels and infiltrated cells. In parallel, the immunohistochemical findings showed an impairment, with respect to the untreated rats, in P2X7 expression, mainly on nerves and vessels close to the site of inflammation. Finally, oATP treatment significantly reduced the presence of infiltrating inflammatory macrophages in the paw tissue.
Conclusion
Taken together these results clearly show that oATP reduces carrageenan-induced inflammation in rats.
==== Body
Background
We previously showed the ability of the local treatment with periodated oxidized ATP (oATP), an inhibitor of the P2X7 ATP (adenosine 5'-triphosphate) receptor [1], to relieve inflammatory pain in the rat paw, in which chronic inflammation was induced by Freund's complete adjuvant (CFA) injection [2,3]. There are two classes of ATP-receptors, the ionotropic P2X receptors and the G-protein-coupled P2Y receptors. Recently, these purinoceptors have been extensively studied because of their important roles in several ATP-mediated cellular functions [4]. In particular, there are seven subunits of P2X receptors (P2X1–7), which are differently expressed by many cell types [4]. Some P2X receptors are expressed in DRG neurons [5].
ATP, released by neuronal and non-neuronal cells, is able to initiate a pronociceptive signal through different P2X subtypes of P2 purinoceptors [6]. The expression of P2X by subsets of primary afferent neurons plays a role in the generation of pain from the periphery to the spinal cord [7,8]. ATP seems to be involved in the initiation of impulses in some sensory fibers [5]. In fact, excitation of sensory neurons by ATP evokes a sensation of pain in humans [9]. In rats, P2X receptor-mediated excitation of nociceptive afferents in normal and arthritic knee joints has been demonstrated [10]. Extracellular ATP exerts its activity during inflammatory processes. In fact, ATP is released by sensory nerves and by damaged cells during inflammation and acts by exciting primary sensory neurons [11,12].
P2X7 receptors for ATP have been demonstrated on monocytes/macrophages, dendritic cells, mast cells, fibroblasts and lymphocytes [13]. Such receptors mediate the ATP cytolytic activity on macrophages [14,15]. We showed the presence of P2X7 receptors also on subcutaneous sensory nerve terminals [3]. Thus, we hypothesized that local antihyperalgesic effect of oATP was related to the inhibition of pain transmission through the block of P2X7 receptors, due to oATP.
In the present work we investigated the effect of oATP treatment in the acutely carrageenan-inflamed rat hind paws. The treatment was performed locally, orally, or intravenously. The expression of the pro-inflammatory chemokines interferon-gamma-inducible protein-10 (IP-10), monocyte chemoattractant protein-1 (MCP-1) and interleukin-8 (IL-8) in the paws was assessed. IP-10 is a CXC-chemokine which regulates the recruitment of T cells [16,17]. It has inhibitory activity on fibroblast migration [18] and has been recently demonstrated to be a potent angiostatic protein in vivo [19]. MCP-1 is a CC-chemokine involved in the recruitment and activation of monocyte and macrophage lineage cell; it is high in serum of patients affected by both relapsing polychondritis and rheumatoid arthritis [20]. It has been also recognized as an angiogenic chemokine through the induction of VEGF (vascular endothelial growth factor) -A gene expression [21]. IL-8 is a CXC-chemokine which exerts potent chemotacting and stimulating activities on neutrophils. In fact, it causes the neutrophil degranulation and the generation of reactive oxygen intermediates, which are able to cause tissue damage and to amplify the inflammatory response [22]. We choosed to examine the expression of IL-8 because it is modulated during acute, carrageenan-induced inflammation, characterized by neutrophil recruitment. However, we also examined the expression of IP-10 and MCP-1, which recruit T cells and monocytes/macrophages, respectively, because such chemokines are released by mastocytes, macrophages and fibroblasts resident in the inflamed tissue. Moreover, some inflammatory mediators (histamine, thrombin, and cytokines as TNFα) produced during acute inflammation can activate endothelial cells, thus enabling them to produce and release MCP-1 [23]. In parallel with chemokine expression, we investigated the expression of P2X7 receptors and the presence of macrophage infiltrate in the paw tissue. Finally, we evaluated the antinociceptive effect of oATP in comparison with that obtained by two known antiinflammatory drugs, e.g. diclofenac and indomethacin. Our results clearly show that oATP exerts both antinflammatory and antihyperalgesic effects.
Results
Therapy with oATP reduces thermal hyperalgesia in rats
To test the potential role of oATP in reducing thermosensation in vivo, we subjected rats to the paw withdrawal latencies assay, and measured the intervals expressed in second. In untreated rats the values of paw withdrawal latencies averaged 12.0 ± 2.0, and were completely similar to those obtained in paws locally treated with saline (11.9 ± 1.0). oATP, administered using three different routes, did not significantly influence such data (12.0 ± 0.8, when administered locally; 11.9 ± 0.7, orally; 12.2 ± 0.9, intravenously; n = 7).
Following oATP administration, the rats were fully awake and responsive to stimuli.
oATP administration reduced hind paw thermal hyperalgesia. Different doses of oATP dissolved in 0.15 ml saline (50, 100, 200 μM) were administered either intraplantarly, orally or intravenously 3 hours after carragenaan injection into right hind paws. Paw withdrawal latencies were measured 3 hours after oATP or saline (vehicle) administration. oATP treatments all significantly increased the antinociceptive score as revealed by an increase in the withdrawal latency compared to basal measurement (= 0 dose) (Fig. 1). Local oATP treatment was significantly less efficient than both oral and intravenous treatments in reducing hyperalgesia (Fig. 1). We decided to use the minimal dose displaying the maximal intravenous effect (e.g. 100 μM) of oATP in the following experiments. Time 0 corresponds to oATP injection. Local intraplantar injection of oATP in rat inflamed paws (3 hours after carrageenan injection) induced a significant increase in paw withdrawal latencies after 1 hour of treatment. Such increase was maintained in time (e.g. at 3, 6, 12 and 24 hours from treatment) compared with time 0 (Fig. 2a). Oral and intravenous administration of oATP induced significantly higher antihyperalgesic effect than local injection. Such effect was very evident after 3 hours of treatment, improved in the following 9 hours and slightly decreased after 24 hours (Fig. 2b and 2c, respectively)
In Fig. 3 the antihyperalgesic effect exerted by oATP (orally or intravenously injected) is compared to the effect exerted by diclofenac or indomethacin used at the same dose. Data show that the antihyperalgesic activity of oATP is significantly more elevated than that of diclofenac and indomethacin.
Histologic expression of P2X7 receptors and HIS36
Each data is representative of 5 different performed experiments for each treatment. As reported in Fig. 4, an intense expression for P2X7 receptors was observed in nerve endings and peripheral nerves of control saline-treated rat hind paws. A similar result, but seldom and focally less intense, was observed in peripheral nerves in hind paws treated with carrageenan. In all specimens treated with carrageenan, an irregularly diffuse inflammatory infiltrate was present, with a prevalence of granulocytes and a variable amount of lympho-histiocytic cells displaying a weak and focal P2X7 positivity.
Hind paws treated with intravenous oATP, alone or after inflammatory reaction induced by carrageenan, presented a reduced expression of P2X7, either in nerves, or close to epidermidis, in dermis or intramuscolar. Few arterial vessels had a weak P2X7 positivity with a more prominent reduction in P2X7 expression in areas close to the inflammation site. Focally groups of cells with lympho-histiocytic cell morphology showed a strong cytoplasmatic granular positivity for P2X7.
Treatment with local oATP, alone or after inflammatory reaction induced by carrageenan, showed a focal, irregular and mild reduction of P2X7 expression, often more prominent in areas with phlogosis both in nerves and in vessels. In all specimens sweat glands revealed a faint but fairly diffuse P2X7 positivity.
Medium size isolated round cells with strong positivity for HIS36 and identified as macrophages were often present in a perivascular pattern in the dermis and subcutis of hind paws. They were more frequent in tissues with carrageenan treatment only (as reported in Fig. 5), few in rat hind paws with oATP local or intravenous administration and very few in those with saline only (data not shown). No differences in intensity or cellular immunostaining distribution were observed among the experiments. The summary of the immunohistochemical results are reported in table 1.
Chemokine immunofluorescence expression
Clinical findings about the development of a local inflammation in hind paws by carrageenan and the antiinflammatory effects of oATP were supported by immunofluorescence assessment of the presence of three inflammatory chemokines, commonly involved in inflammation-dependent immune response phenomena: MCP-1, IL-8 and IP-10. In all the processed rat paws, cytokine expression was confirmed in dermis and always absent in epidermidis. Any relevant expression of MCP-1 was not observed by confocal microscopy in the analyzed bioptic samples, excluding a weak label of arteriolar endothelium and isolated cells in reticular dermis in carrageenan inflamed specimen sections (data not shown). The control, submitted to saline only, failed to show fluorescent signals for all the three cytokines on both papillary and reticular layers of the dermis (Fig. 6a). Conversely, carrageenan-treated rat hind paw section samples expressed both IL-8 and IP-10 on dermis infiltrating cells, but not on vessel walls (Fig. 6b). A low expression of IP-10 on papillary dermis cells was observed on hind paws from rats treated with oATP alone, following either local (Fig. 6e) or intravenous injection (Fig. 6f). The IP-10 appeared co-localized with IL-8 on small vessel wall of dermis reticular layer in specimens from rat paws carrageenan-injected, then submitted to oATP local treatment (Fig. 6c). No significant chemokine labeling was assessed on specimens upon local treatment of rat paws with carrageenan and intravenous injection of oATP (Fig. 6d). The intravenous administration of oATP seemed more effective then the local injection in reducing the expression of chemokines in carrageenan-inflamed hind paws, according to clinical findings.
Discussion
The hypothesis that ATP is a pain mediator [28] has supported the studies on its P2X receptors. Recently, ATP-mediated mechanical hyperalgesia has been shown to be decreased by the use of P2X3 receptor antagonists [29].
Based on our previous reports of the antinociceptive activity of oATP at the level of the hind paws in rats treated intraplantarly with CFA [3], we investigated the therapeutic effect of oATP in another model of rat paw inflammation, carrageenan-induced. The extracellular ATP released in inflamed tissues has pronociceptive activity [30,31]: it possibly acts by binding and activating the P2X receptors for ATP present on pain-sensing neurons. In addition, since P2X7 receptors are localized on many tissue components involved in inflammation, the binding and activation of such receptors by ATP could enable the same structures to release proinflammatory and pronociceptive mediators. Local treatment with oATP abrogates inflammatory pain possibly by inhibiting P2X7 receptors for ATP which are localized on nerve terminals and on vessels [2,3].
Paw inflammation is a multifactorial response due to proinflammatory chemokines which are responsible for the infiltration and activation of various leukocyte population in joint tissue. We analysed in parallel the expression of some chemokines (MCP-1, IP 10 and IL 8) and the expression of P2X7 receptors in the same inflamed tissues and their modifications due to oATP treatment. The analysis of the chemokine presence had the aim to evaluate wheter the use of oATP was able to influence the recruitment of circulating leukocytes. This feature was not shown by IL-lβ, a cytokine produced by activated macrophages, which fails to display chemoattractant properties, indicating only the state of macrophage activation. The three selected chemokines are described to be able to recruit the principal inflammatory cells: granulocytes (IL-8), monocytes/macrophages (MCP-1) and T lymphocytes (IP-10).
Our clinical results indicate that: 1) local treatment with oATP significantly relieves inflammatory pain in carrageenan-induced paw inflammation model, as in previously reported CFA model [3]; 2) oral and intravenous treatments with oATP are more efficient than local treatment in reducing the inflammatory pain; 3) intravenous treatment is more efficient than oral treatment; 4) the antinociceptive activity of a single dose of oATP lasts 24 hours; 5) the antinociceptive activity of oATP is more evident than that due to diclofenac or indomethacin in our experimental model. Based on the increased antinociceptive effect due to oATP intravenously administered with respect to that orally administered, we studied the expression of P2X7 receptors and of chemokines in inflamed tissues following local or intravenous oATP administration, e.g. in conditions of minimal or maximal antihyperalgesic oATP effect.
P2X7 receptors are involved in several processes concerning immunomodulation and inflammation. Previous reports show that the absence of P2X7 receptors alters cytokine production and leukocyte function and attenuates the inflammatory response [32,33]. In fact, high ATP concentrations were unable to promote IL-lβ extracellular accumulation from lypopolysaccharide (LPS)-activated blood samples derived from P2X7 receptor-deficient mice, in contrast with blood samples obtained from wild-type. In addition, P2X7 receptor-deficient mice were protected against clinical signs of mAb-induced arthritis [33]. Recent data indicate that, in response to ATP binding, the P2X7 receptors facilitate cation channel activation, non-specific pore formation, rapid changes in plasma membrane morphology (blebbing) and secretion of IL-1β from LPS-primed macrophages [34]. The "blebbing" induced by ATP was blocked by oATP. Blebbing was abrogated in the presence of Rho-effector kinase inhibitors, whereas ATP-induced IL-lβ release was unaffected, suggesting different signalling pathways for P2X7 receptors. In addition, the P2X7 receptor-protein complex comprises several distinct identified proteins, with different regulatory capacities [35], indicating a complexity for the P2X7 receptor functions.
We showed the antinociceptive activity of oATP, which is able to block the P2X7 receptor [1], in rats bearing acute paw inflammation. The question arises about the mechanisms by which the oATP-induced inhibition of P2X7 receptor activation are able to relieve inflammatory pain. Our previous data suggested the possibility that the block of P2X7 receptors expressed by nerve terminals could be useful to inhibit pain transmission [3]. In inflamed paws the expression of P2X7 receptors was less evident in oATP-treated than in untreated rats, independently on the route of oATP administration. Noteworthly, such expression was reduced at the level of nerve terminals and of vessels close to inflammation site, more than at the level of the immune cells (see Fig. 4 and 5). The expression of MCP-1 was weakly evident only in carrageenan-inflamed rat paws at the level of arteriolar endothelium and of some dermal cells; in the same structures we observed by immunohistochemistry an increased number of macrophages (HIS36 positive cells). MCP-1 was absent and only few macrophages were present in paws from oATP-treated rats.
These data confirm the strict correlation between P2X7 expression by leukocytes and relative inflammatory response and underline that the reduction of inflammatory effects due to oATP administration is similar to that observed in P2X7 receptor deficient mice [33]. Carrageenan-treated rat paws expressed IL-8 and IP-10 on dermal infiltrating cells, whereas a low expression of IP-10 on dermal cells followed oATP treatment alone (both local and intravenous). IP-10 and IL-8 were co-localized on the wall of small vessels in carrageenan-inflamed paws of intraplantarly oATP-treated rats. Moreover, no chemokine label was observed in carrageenan-inflamed paws of intravenously-oATP-treated rats. Such results indicate that oATP treatment significantly reduced chemokine expression at vascular rather than at infiltrating cell level. In fact oATP local treatment of carrageenan-inflamed paws completely abrogated MCP-1 expression, whereas the intravenous treatment with oATP abrogated the expression of all the three chemokines in inflamed paws. In particular, the abrogation of MCP-1 production by oATP could be related to the oATP-induced block of P2X7 function on activation of endothelial and immune cells, which can become ineffective to produce MCP-1.
In our results we underline the relation between the absence of recruited macrophages and the absence of MCP-1 expression in the inflamed paws of intravenously oATP-treated rats.
Other studies demonstrate the presence of P2X7 receptors on nervous and vascular structures. In fact, P2X7 immunoreactivity has been shown in rats at the level of glial cells of gastrointestinal musculature, in mienteric and submucosal ganglia (where perineuronal nerve endings appeared brightly labeled) [36]. In addition, pharmacological identification of P2X1, P2X4 and P2X7 nucleotide receptors has been evidenced in the smooth muscle of human umbilical cord and chorionic blood vessels [37]. Moreover, LPS-activated endothelial cells have been shown able to secrete ATP, via P2X7 receptors, and to release IL-lα [38].
During carrageenan-induced acute inflammation ATP is released by damaged cells. When ATP activates cells involved in the inflammatory process and bearing P2X7 receptors, other mediators and ATP molecules are released by these cells. Indeed, oATP could block such amplifying mechanism of inflammation and also downregulate P2X7 receptor expression, especially in vessels and nerves close to the site of inflammation (see Fig 4). This hypothesis is supported by the fact that chemokine expression is reduced by oATP treatment. We hypothesize that oATP, by blocking the P2X7 receptors present on nerves and endothelial cells, could regulate some effects of ATP (having pronociceptive functions when released) on these structures. The effect of oATP was greater when given orally or intravenously instead of locally, possibly because the oral and intravenous administrations of oATP could permit its systemic diffusion. Thus oATP could reach sites distal to the site of inflammation and so enhance the contact with circulating leukocytes bearing P2X7 receptors. Such contact favours the inactivation of immune cells, which are able to produce inflammatory pronociceptive mediators. In addition, the most antinociceptive activity obtained when oATP was intravenously injected (instead of locally or orally) could indicate that the molecule crosses the blood-brain barrier and that it may also have central effects and not only peripheral effects. It is also possible that oATP blocks other receptors than P2X7, e.g. P2X1 and P2X2 [4]. Effectively, human P2X1 and P2X7 receptors are co-expressed in several cell types such as lymphocytes or epithelial cells [39], and oATP was found to be ineffective in separating P2X1 receptor current from the P2X7 current. In addition, none oATP activity on P2X4 has been reported. However, the involvement of P2X7 is also assured by the fact that P2X3-/- mice did not display a difference in thermal hyperalgesia to wild type controls in the carrageenan model of inflammation [40]. Our data suggest that the potential analgesic and also the anti-inflammatory properties of oATP might have significant therapeutic potential in reducing inflammatory pain. In addition, the more efficient antinociceptive activity of oATP with respect to that of some known NSAIDs is helpful, in consideration of NSAID direct cytotoxic effects [41].
Recently, it has been demonstrated that oATP can reduce pro-inflammatory TNFα-induced signalling in HEK293 cells, that lack of P2X7 expression. Such and other reported results possibly indicate that oATP interfere with the activation of signalling pathways involved in the inflammatory response, independently on oATP action on P2X receptors [42], thus showing a mechanism independent of the expression or activation of known P2 receptor subtypes. However, the involvement of the P2X7 receptors in the inflammatory pain has been suggested by other results [43]. In fact, mice lacking P2X7 receptors did not display inflammatory and neuropathic hypersensitivity to both mechanical and thermal stimuli and were impaired in the ability to release IL-1 beta, IL-10 and MCP-1, the latter in agreement with our findings, obtained with the administration of oATP in rats bearing paw inflammation. All data suggest that P2X7 receptors exert a role in both acute and chronic inflammations.
Conclusion
Our results show that oATP systemically administered removes hyperalgesia induced by carrageenan in rat paws. Tha data suggest the possible use of oATP to reduce the inflammatory pain.
Methods
Animals
The procedures followed the guidelines of the International association for the Study of Pain [24]. Male Wistar rats from Harlan Italy (Corezzana, Milano, Italy) weighing about 250 g were used. They were housed in groups of three per cage, allowed free access to water and food and exposed to a 12/12-hour light/dark cycle, and acclimatised to the laboratory at least 5 days prior to the experiments.
Induction of inflammation
Following brief exposure to halothane (5%) anesthesia (Hoechst, Milano, Italy), inflammation was induced by the injection of 0.1 ml 1% carrageenan (λ-carrageenan, Sigma-Aldrich, Milan, Italy) in sterile saline in the right hind paw of the rats. Maximal incidence of inflammation (e.g. the hind paw edema) occurred 3 hours after carrageenan injection, as previously reported [25].
Administration of oATP
Three hours following carrageenan injection, the rats were treated with oATP (Sigma-Aldrich) dissolved in 0.15 ml sterile saline, whereas control rats received the drug vehichle alone (0.15 ml sterile saline). Firstly, 3 different doses oATP (50 μM, 100 μM, 200 μM, in 0.15 ml, respectively) were administered: 1) locally, by intraplantar injection in the right hind paw; 2) orally, by using an orogastric catheter; 3) intravenously, through the left femoral vein. The injections were performed under brief halothane anestesia and the antinociceptive effect of oATP was measured 3 hours following its administration. Since a more significant antihyperalgesic effect was observed by using 100 μM oATP (see Results section), such dose was selected among the tested doses to perform further experiments. To perform histologic evaluations (by immunohistochemistry or by confocal analysis), oATP was administered locally or intravenously; in fact, as shown in the Results section, these routes of administration respectively showed minimal or maximal antihyperalgesic effect (as clinically evidenced).
Reduced thermal hyperalgesia measure
The observers of the thermal hyperalgesia measurements were blinded. The method of Hargreaves was used to assess the hind paw nociceptive thresholds to thermal stimuli [26]. We used a plantar test apparatus (Ugo Basile, Comerio, Italy). Briefly, the rats were placed in a clear plastic chamber and left to acclimatize for 5 minutes before testing. Radiant heat stimulus was induced by light from a 8 V-50 W halogen bulb (64607 OSRAM), delivered to the plantar surface of the rat's hind paw through the base of the plastic box; the beam was about 12 mm in diameter. The rats were familiarized with the handling procedure 3 days before the antinociceptive test. Time taken by the animal to withdraw its right hind paw was measured before and after oATP injection. Reduced thermal hyperalgesia was defined as a significant increase in the withdrawal latency compared to the basal measurement. The values of thermal hyperalgesia were determined at 1, 3, 6, 12 and 24 hours after oATP injection. The same procedure was used to test the antihyperalgesic effect of diclofenac and indomethacin orally or intravenously administered at the same dose of oATP (e.g. l00 μM in 0.15 ml saline)
Confocal analysis
Rat right hind paws were sampled and processed for cryostatic inclusions. For any treatment, 5 different experiments were performed. The treatments included: saline, carrageenan, oATP alone, and carrageenan ± local or intravenous oATP administration. The same hind paws were also used for immunohistochemistry.
Biopsies were fixed in 4% paraformaldehyde in Dulbecco's PBS (DPBS) then cryoprotected in 20% sucrose in DPBS, embedded in Tissue Tek medium and snap-frozen by immersion in liquid nitrogen. Pseudo-seriated l0 μm thick sections were submitted to indirect immunofluorescence staining by using mouse monoclonal antibodies (mAbs) against mouse α MCP-1 (B-B4 RPE, Groningen, The Netherlands)ÙIP-10 (Peprotech EC Ltd., London, UK), and IL-8 (Pharmingen, San Diego, CA). The sections were incubated in blocking solution (1% BSA in DPBS, 30 minutes, RT), then in primary Ab diluted in blocking solution (2 hours, RT), finally, after 30 minutes of DPBS washing, in conjugated secondary Ab (Rabbit-anti-Mouse Ig-FITC conjugated, DAKO, Milano, Italy) in DPBS (30 minutes, RT). Double staining was realized by superposing the same procedure with a different primary Ab and a TRITC-labelled secondary Ab. The 3 possible couplings of primary Abs were tested and single staining served as controls. DAPI nuclear staining (0.2 nM, 20 minutes, RT) followed the immunostaining steps.
The sections, mounted with Fluorsave (Calbiochem, Merck Eurolab Srl, Milano, Italy) were analysed with the support of a confocal microscope Leica TCS SP2 (Leica Microsystems, Heidelberg, GmbH); 3D maximum projections were obtained from single channel-collected series of images, subsequently they were superposed by Adobe Photoshop software.
Immunohistochemistry
Materials and methods for immunostaining have been previously described [3]. Immunostain was perfomed on all specimens by using two different procedures. Sections (4 μm-thick) were prepared on slides pretreated with poly-L-lysine (Sigma Diagnostic Inc., St Louis, USA). A double staining method was applied to examine the relationship between P2X7 and cutaneous nerve fibers using anti-P2X7 receptor purified polyclonal antibody (Chemicon International, Inc., Temecula, CA) and the monoclonal antibody Protein Gene Products 9.5 (Novocastra Laboratories Ltd, Newcastle, UK). The anti-P2X7 polyclonal antibody was directed against a c-term antibody (epitope 567–595), which has been previously shown highly selective [27]. Binding was revealed by 3',3'-diaminobenzidine (Liquid Dab, BioGenex, San Ramon, CA) for P2X7 receptor and by 5-bromochloroindoxyl phosphate and nitro blue tetrazolium chloride (DAKO, Copenhagen, Denmark) for PGP 9.5. The two procedures were performed alternatively, on both antibodies. Sections were counterstained with Green Light. To obtain a negative control, the primary antibodies were routinely omitted.
In a separate setting of immunostainings we used a mouse monoclonal antibody HIS36 (Santa Cruz, California, USA) specific for rat macrophage subset. All the immunostained slides were viewed in a blinded fashion.
Statistical analysis
The withdrawal latency data (expressed as mean ± SEM) were analyzed using a two-way analysis of variance followed by Dunnett's test. Significance was assumed when P < 0.05.
Abbreviations
oATP, periodate oxidized ATP; MCP-1, monocyte chemoattractant protein-1; VEGF, vascular endothelial growth factor, IP-10, interferon gamma-inducible protein-10; IL8 interleukin 8
Authors' contributions
AF carried out clinical experiments and statistical analysis, GdA and EDC carried out the immunohisochemical studies, CF and PT carried out the confocal analysis, JSF participated in the sequence alignment and MEF performed the design of the study and has written the manuscript.
Acknowledgements
We thank Professor M Tiengo for his helpful suggestion.
The manuscript was supported in part by grants obtained from PhD "Scienze Neurologiche e del Dolore"
Figures and Tables
Figure 1 Reduction by oATP (used at the doses of 50, 100 and 200 μM, dissolved in 0.15 ml saline) of inflammatory pain in rat paws, as measured by paw withdrawal latency to a radiant heat stimulus, following the administration of carrageenan. Results are expressed as the mean ± SEM with n = 7 rats per group. * p < 0.05 compared with dose 0; ^ p < 0.05 compared with local administration at the same dose. All values were significantly higher in oATP-treated paws (circle) than in saline-treated paws (triangle).
Figure 2 a)Withdrawal latencies measured 3 hours after intraplantar carrageenan injection (basal measurement = time 0) and at different hours following oATP (100 ( μM in 0.15 ml saline, continuous line) or 0.9% saline (0.15 ml, hatched line) local intraplantar injection (circle). Mean ± SEM of 7 rats. * p < 0.05 compared to saline at the same time. b) Withdrawal latencies measured 3 hours after intraplantar carrageenan injection (basal measurement = time 0) and at different hours following oATP (100 μM in 0.15 ml saline, continuous line) or saline (0.15 ml, hatched line) oral administration (square). Mean ± SEM of 7 rats. * p < 0.05 compared to saline at the same time. c) Withdrawal latencies measured as previously (3 hours after local carrageenan injection = time 0) and at different hours of oATP (100 μM in 0.15 ml saline, continuous line) or saline (0.15 ml, hatched line) intravenous administration (triangle). Mean ± SEM of 7 rats. * p < 0.05 compared to saline at the same time.
Figure 3 a) Withdrawal latencies measured 24 hours after intraplantar carrageenan injection (basal measurement = time 0) and after different hours from oATP (circle), or diclofenac (square), or indomethacin (triangle) oral administration (100 μM in 0.15 ml saline). Mean ± SEM of 7 rats. *p < 0.05 compared to diclofenac or indomethacin at the same time. b) Withdrawal latencies measured 24 hours after intraplantar carrageenan injection (basal measurement = time 0) and after different hours from oATP (circle), or diclofenac (square), or indomethacin (triangle) intravenous injection (100 μM in 0.15 ml saline). Mean ± SEM of 7 rats. *p < 0.05 compared to diclofenac or indomethacin at the same time.
Figure 4 Immunohistochemistry of P2X7 receptors. Hind paw sections from control rats treated with saline (a) or oATP locally (c) or intravenously (e). After carrageenan induced inflammation, hind paws were submitted, to either saline (b), oATP local (d) or oATP intravenous (f) treatment. Notice the strong labelling on nerves and vessels in specimens treated only with saline, with variable presence of P2X7 which shows a general reduction after oATP local treatment, more evident after intravenous oATP treatment (original magnification 200x, P2X7 evidenced in brown, 3,3 diaminobenzidine, PGP 9.5 evidenced in blue, nithroblue tetrazolium). Data are representative of 5 experiments.
Figure 5 Presence of macrophages in rat inflamed paw section. In dermis and subcutis of carrageenan-treated hind paw a discrete, focal macrophagic (HIS36positive cells) infiltration is present (original magnification 250x; HIS36 evidenced with Dab). Dataare representative of 5 experiments.
Figure 6 Confocal microscopy on rat paws. The sections are doubly labeled with anti- IP-10 and IL-8 mAbs revealed by FITC- (green) or TRITC- (red) conjugated secondary Abs respectively. Merged Free Projection Max of images seriesshows cytokine expression after saline (a), carrageenan (b), carrageenan+oATP local treatment (c), carrageenan+oATP intravenous treatment (d), local oATP treatment (e), intravenous oATPtreatment (f). Nuclei are stained with DAPI (blue) (original magnification 400x). Notice thepresence of activated cells in the derma in b and f, the endoluminal signal in dermal small vessels(a, c, e) and the absence of cytokine labeling in d (original magnification 400x).
Table 1 Immunohistochemical results in rat hind paws after induction of inflammation by carrageenan
Treatment P2X7 receptor nerve expression P2X7 receptor vascular expression P2X7 receptor expression in lympho-histiocytic cells Quantity of macrophages (HIS36+)
Saline only Strong Strong Weak Very few
Carrageenan only Strong, locally moderate Strong, focally moderate Weak Focally in discrete amount
Carrageenan + OATP intraplantar injection Moderate to weak Moderate to weak Weak to moderate Few
Carrageenan + oATP intravenous administration Weak to absent Very weak (absent in area close to inflammatory infiltrate) Moderate, focally strong Few
Note. A discrete amount of granulocytes are present in carrageenan-treated hind paws.
==== Refs
Murgia M Hanau S Pizzo P Rippa M Di Virgilio F Oxidized ATP: an irreversible inhibitor of the macrophage purinergic P2Z receptor J Biol Chem 1993 268 8199 8203 8463330
Dell'Antonio G Quattrini A Dal Cin E Fulgenzi A Ferrero ME Antinociceptive effect of a new P2Z/P2X7 antagonist, oATP, in arthritic rats Neurosci Lett 2002 327 87 90 12098642 10.1016/S0304-3940(02)00385-3
Dell'Antonio G Quattrini A Dal Cin E Fulgenzi A Ferrero ME Relief of inflammatory pain in rats by local use of the selective P2X7 ATP receptor inhibitor, oxidized ATP Arthritis Rheum 2002 46 3378 3385 12483745 10.1002/art.10678
North A Molecular physiology of P2X receptors Physiol Rev 2002 82 1013 1067 12270951
Dunn PM Zhong Y Burnstock G P2X receptors in peripheral neurons Prog Neurobiol 2001 65 107 134 11403876 10.1016/S0301-0082(01)00005-3
Kennedy C Leff P Painful connection for ATP Nature 1995 377 385 386 7566110 10.1038/377385a0
Burnstock G P2X receptors in sensory neurones Br J Anaesth 2000 84 476 488 10823099
Burnstock G Purine-mediated signalling in pain and visceral perception Trends Pharmacol Sci 2001 22 182 188 11282418 10.1016/S0165-6147(00)01643-6
Hamilton SG Warburton J Bhattacharjee A Ward J McMahon SB ATP in human skin elicits a dose-related pain response which is potentiated under conditions of hyperalgesia Brain 2000 123 1238 1246 10825361 10.1093/brain/123.6.1238
Dowd E McQueen DS Chessell DP Humphrey PP P2X receptor-mediated excitation of nociceptive afferents in the normal and arthritic rat knee joint Br J Pharmacol 1998 125 341 346 9786507 10.1038/sj.bjp.0702080
Burnstock G Autonomic neuroeffector junctions: reflex vasodilatation of the skin J Invest Dermatol 1977 69 47 57 17640 10.1111/1523-1747.ep12497872
Krishtal OA Marchenko SM Obukhov AG Cationic channels activated by extracellular ATP in rat sensory neurons Neuroscience 1988 27 995 1000 2855265 10.1016/0306-4522(88)90203-5
Gu BJ Zhang WY Bendall LJ Chessell IP Buell GN Wiley JS Expression of P2X(7) purinoceptors on human lymphocytes and monocytes: evidence for nonfunctional P2X(7) receptors Am J Physiol Cell Physiol 2000 279 C1189 C1197 11003599
Surprenant A Rassendren F Kawashima E North RA Buell G The cytolytic P2Z receptor for extracellular ATP identified as a P2X receptor (P2X7) Science 1996 272 735 738 8614837
Ferrari D Chiozzi P Falzoni S Hanau S Di Virgilio F Purinergic modulation of interleukin-1 release from microglial cells stimulated with bacterial endotoxin J Exp Med 1997 185 579 582 9053458 10.1084/jem.185.3.579
Solomon MF Kuziel WA Simeonovic CJ The contribution of chemokines and chemokine receptors to the rejection of fetal proislet allografts Cell Transplant 2004 13 503 514 15565863
Christen U von Herrath MG IP-10 and type 1 diabetes: a question of time and location Autoimmunity 2004 37 273 282 10.1080/08916930410001713124
Tager AM Kradin RL LaCamera P Bercury SD Campanella GS Leary CP Polosukhin V Zhao LH Sakamoto H Blackwell TS Luster AD Inhibition of pulmonary fibrosis by the chemokine IP-10/CXCL10 Am J Respir Cell Mol Biol 2004 31 395 404 15205180 10.1165/rcmb.2004-0175OC
Yang J Richmond A The angiostatic activity of interferon-inducible protein-10/CXCL10 in human melanoma depends on binding to CXCR3 but not to glycosaminoglycan Mol Ther 2004 9 846 855 15194051 10.1016/j.ymthe.2004.01.010
Stabler T Piette JC Chevalier X Marini-Portugal A Kraus VB Serum cytokine profiles in relapsing polychondritis suggest monocyte/macrophage activation Arthritis Rheum 2004 50 3663 3667 15529362 10.1002/art.20613
Hong KH Ryu J Han KH Monocyte chemoattractant protein-1-induced angiogenesis is mediated by vascular endothelial growth factor-A Blood 2005 105 1405 1407 Epub 2004 Oct 21 15498848 10.1182/blood-2004-08-3178
Gessler P Pretre R Hohl V Rousson V Fischer J Dahinden C CXC-chemokine stimulation of neutrophils correlates with plasma levels of myeloperoxidase and lactoferrin and contributes to clinical outcome after pediatric cardiac surgery Shock 2004 22 513 520 15545821 10.1097/01.shk.0000145939.54838.51
Luster AD Chemokines-chemotactic cytokines that mediate inflammation N Engl J Med 1998 338 436 445 9459648 10.1056/NEJM199802123380706
Zimmermann M Ethical guidelines for investigations of experimental pain in conscious animals Pain 1983 16 109 110 6877845 10.1016/0304-3959(83)90201-4
Bianchi M Panerai AE Effects of lornoxicam, piroxicam, and meloxicam in a model of thermal hindpaw hyperalgesia induced by formalin injection in rat tail Pharmacol Res 2002 45 101 105 11846620 10.1006/phrs.2001.0921
Hargreaves K Dubner R Brown F Flores C Joris J A new and sensitive method for measuring thermal nociception in cutaneous hyperalgesia Pain 1988 32 77 88 3340425 10.1016/0304-3959(88)90026-7
Sim JA Young MT Sung HY North NA Surprenant A Reanalysis of P2X7 receptor expression in rodent brain J Neurosci 2004 24 6307 6314 15254086 10.1523/JNEUROSCI.1469-04.2004
Hamilton SG ATP and pain Pain and practice 2002 2 289 294 10.1046/j.1533-2500.2002.02041.x
Dai Y Fukuoka T Wang H Yamanaka H Obata K Tokunaga A Noguchi K Contribution of sensitized P2X receptors in inflamed tissue to the mechanical hypersensitivity revealed by phosphorylated ERK in DRG neurons Pain 2004 108 258 266 15030945 10.1016/j.pain.2003.12.034
Bland-Ward PA Humphrey PP Acute nociception mediated by hindpaw P2X receptor activation in the rat Br J Pharmacol 1997 122 365 371 9313948 10.1038/sj.bjp.0701371
Hamilton SG Wade A McMahon SB The effects of inflammatory mediators on nociceptive behaviour induced by ATP analogues in the rat Br J Pharmacol 1999 126 326 332 10051152 10.1038/sj.bjp.0702258
Solle M Labasi J Perregaux DG Stam E Petrushova N Koller BH Altered cytokine production in mice lacking P2X7 receptors J Biol Chem 2001 276 125 132 11016935 10.1074/jbc.M006781200
Labasi JM Petrushova N Donovan C McCurdy S Lira P Payette MM Absence of the P2X7 receptor alters leukocyte function and attenuates an inflammatory response J Immunol 2002 168 6436 6445 12055263
Verhoef PA Estacion M Schilling W Dubyak GR P2X7 receptor-dependent blebbing and the activation of Rho-effector kinases, caspases, and IL-1 beta release J Immunol 2003 170 5728 5738 12759456
Adinolfi E Kim M Young MT Di Virgilio F Surprenant A Tyrosine phosphorylation of HSP90 within the P2X7 receptor complex negatively regulates P2X7 receptors J Biol Chem 2003 278 37344 37351 12869560 10.1074/jbc.M301508200
Vanderwinden JM Timmermans JP Schiffmann SN Glial cells, but not interstitial cells, express P2X7, an ionotropic purinergic receptor, in rat gastrointestinal musculature Cell Tissue Res 2003 312 149 154 12684872
Valdecantos P Briones R Moya P Germain A Huidobro-Toro JP Pharmacological identification of P2X1, P2X4 and P2X7 nucleotide receptors in the smooth muscles of human umbilical cord and chorionic blood vessels Placenta 2003 24 17 26 12495655 10.1053/plac.2002.0862
Imai M Goepfter C Kaczmarek E Robson SC CD39 modulates IL-1 release from activated endothelial cells Biochem Biophys Res Commun 2000 270 272 278 10733939 10.1006/bbrc.2000.2410
Seyffert C Schmalzing G Markwardt F Dissecting individual current components of co-expressed human P2X1 and P2X7 receptors Curr Top Med Chem 2004 4 1719 30 15579104 10.2174/1568026043387160
Souslova V Cesare P Ding Y Akopian AN Stanfa L Suzuki R Carpenter K Dickenson A Boyce S Hill R Nebenuis-Oosthuizen D Smith AJ Kidd EJ Wood JN Warm-coding deficits and aberrant inflammatory pain in mice lacking P2X3 receptors Nature 2000 407 1015 1017 11069182 10.1038/35039526
Tomisato W Tsutsumi S Hoshino T Hwang HJ Moi M Tsuchiya T Nizushima T Role of direct cytotoxic effects of NSAIDs in the induction of gastric lesions Biochem Pharmacol 2004 67 575 585 15037209 10.1016/j.bcp.2003.09.020
Beigi RD Kertesy SB Aquilina G Dubyak GR Oxidized ATP (oATP) attenuates proinflammatory signaling via P2 receptor-independent mechanisms Br J Pharmacol 2003 140 507 519 14522842 10.1038/sj.bjp.0705470
Chessell IP Hatcher JP Bountra C Michel AD Hughes JP Green P Egerton J Murfin M Richardson J Peck WL Grahames CB Casula MA Yiangou Y Birch R Anand P Buell GN Disruption of the P2X7 purinoceptor gene abolishes chronic inflammatory and neuropathic pain Pain 2005 114 386 396 15777864 10.1016/j.pain.2005.01.002
|
16042776
|
PMC1190175
|
CC BY
|
2021-01-04 16:03:32
|
no
|
BMC Immunol. 2005 Jul 22; 6:18
|
utf-8
|
BMC Immunol
| 2,005 |
10.1186/1471-2172-6-18
|
oa_comm
|
==== Front
BMC Mol BiolBMC Molecular Biology1471-2199BioMed Central London 1471-2199-6-181605351810.1186/1471-2199-6-18Research ArticleAnalysis of the function of E. coli 23S rRNA helix-loop 69 by mutagenesis Liiv Aivar [email protected] Diana [email protected]äli Ülo [email protected] Jaanus [email protected] Estonian Biocentre, Riia 23, 51010 Tartu, Estonia2 Institute of Molecular Biology and Cell Biology, Tartu University, Riia 23, 51010 Tartu, Estonia3 Clinic for Neurology, Leipziger Str. 44, D-39120 Magdeburg, Germany2005 29 7 2005 6 18 18 6 1 2005 29 7 2005 Copyright © 2005 Liiv et al; licensee BioMed Central Ltd.2005Liiv et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The ribosome is a two-subunit enzyme known to exhibit structural dynamism during protein synthesis. The intersubunit bridges have been proposed to play important roles in decoding, translocation, and the peptidyl transferase reaction; yet the physical nature of their contributions is ill understood. An intriguing intersubunit bridge, B2a, which contains 23S rRNA helix 69 as a major component, has been implicated by proximity in a number of catalytically important regions. In addition to contacting the small ribosomal subunit, helix 69 contacts both the A and P site tRNAs and several translation factors.
Results
We scanned the loop of helix 69 by mutagenesis and analyzed the mutant ribosomes using a plasmid-borne IPTG-inducible expression system. We assayed the effects of 23S rRNA mutations on cell growth, contribution of mutant ribosomes to cellular polysome pools and the ability of mutant ribosomes to function in cell-free translation. Mutations A1912G, and A1919G have very strong growth phenotypes, are inactive during in vitro protein synthesis, and under-represented in the polysomes. Mutation Ψ1917C has a very strong growth phenotype and leads to a general depletion of the cellular polysome pool. Mutation A1916G, having a modest growth phenotype, is apparently defective in the assembly of the 70S ribosome.
Conclusion
Mutations A1912G, A1919G, and Ψ1917C of 23S rRNA strongly inhibit translation. Mutation A1916G causes a defect in the 50S subunit or 70S formation. Mutations Ψ1911C, A1913G, C1914A, Ψ1915C, and A1918G lack clear phenotypes.
==== Body
Background
High-to-medium-resolution structures of the ribosome have by their ability to generate structure-based functional hypotheses radically changed the way the ribosome is studied. One of the more intriguing results that has come from structural studies is the extraordinary number of roles attributed to a single 19 nt helix-loop, H69 of 23S rRNA (Fig 1). Crystallographic studies of the Thermus thermophilus ribosome [1] and cryo-EM studies of E. coli ribosomes [2,3] have made it evident that H69 is a component of both the A and P sites with an ability to simultaneously contact two tRNAs. It contacts the D-stem and D-stem junction of the A site tRNA by the loop residues 1913–1915 and the same parts of the P site tRNA by backbone-backbone interactions with stem nucleotides 1908, 1909, 1922 and 1923 [1] (Fig 1). In addition, H69 loop residues 1912, 1913, 1914 and 1918 contact 16S rRNA H44, thus forming the intersubunit bridge B2a [1,2]. Chemical cross-linking and footprinting data further corroborates the close proximity of H69 to intersubunit contact area [4,5]. The importance of H69 in subunit association is emphasized by the recent finding that DMS-modifications of A1912 or A1918 (but not of A1913) abolish 70S formation in an in vitro test system [6]. Also, hydroxyl-radical footprinting of the anti-subunit-association factor IF3 on the 30S subunit implicates IF3 binding to the region that is occupied by the loop of H69 in 70S ribosomes [7], suggesting that disallowing of the bridge B2a may be important for keeping the subunits separate before correct initiation of translation.
Since H69 adopts a different conformation in 50S subunits and 70S ribosomes, it has to change conformation upon 30S binding [1,8]. Conformational flexibility of H69 may also be important in translocation since it is hard to imagine tRNA movement from A to P site with H69 stuck in its path. An active role for H69 in translocation has been proposed [9] but has not yet been experimentally tested. In addition to interactions with tRNAs and the 30S subunit, contacts of H69 with various A site substrates have been proposed. Based on cryo-EM reconstitution of the ribosome with bound aa-tRNA-EF-Tu-GDP-kirromycin, Valle et al. speculate that tRNA contacts with H69 might actively promote the observed kink in tRNA structure [10]. Cryo-EM studies have also led to proposals of H69 contacts with eEF2 [11], RF2 [12], RF3 [13], RRF [14] and SmpB in the Ala-tmRNA-SmpB-EF-Tu-kirromycin complex [15].
Another interesting feature of the H69 is its three pseudouridines at positions 1911, 1915 and 1917 [16]. They are synthesized by a single synthase, RluD, which is the only pseudouridine synthase in E. coli whose deletion leads to a strong growth defect [17]. Defective RluD function leads to impaired ribosome assembly [18]. This observation suggests that H69 actively promotes the process of ribosomal large subunit assembly.
O'Connor and Dahlberg selected three mutations in H69 (ΔA1916, insertion of two adenosines after A1916, and C1914U) that cause increased +1 and -1 frameshifting and read-through of all three stop codons [19]. Here we mutate each residue in the loop of H69 and analyze the growth phenotypes, assembly of the mutant ribosomes, their incorporation into polysomes and activities in poly-uridine-directed poly-phenylalanine synthesis. The results obtained in this work point to residues A1912, A1916, Ψ 1917 and A1919 as important for correct functioning of the E. coli ribosome.
Results
Experimental design
Because of the crucial nature of protein synthesis for cellular viability and the perceived importance of 23S rRNA helix 69 for correct functioning of the ribosome, its mutations are likely to be lethal. Therefore, we used inducible expression to study in vivo phenotypes of mutations in 23S rRNA. Mutated 23S rRNA genes were expressed from the plasmid ptBsB under the control of IPTG-inducible tac promoter [20]. In order to be able to quantify the fraction of mutant 23S rRNA in ribosomes and to functionally differentiate between plasmid-borne and chromosomally encoded ribosomes during in vitro translation, the single mutations were combined with the second site mutation A1067U in the plasmid ptBsB1067T. This mutation confers resistance to thiostrepton during cell free translation [21]. Therefore, A1067U enables to discriminate the activity of chromosomally encoded wild-type ribosomes from mutant ribosomes containing plasmid-encoded 23S rRNA. 30–40% of the cellular ribosome pool contain mutant 23S rRNA [22]. We constructed the following mutations in the loop of H69: Ψ1911C, A1912G, A1913G, C1914A, Ψ1915C, A1916G, Ψ1917C, A1918G and A1919G.
Effect of mutations on cell growth
Mutant 23S rRNA expression was induced with 1 mM IPTG in XL-1 cells growing in rich liquid media at 37°C in the early exponential growth phase. The plasmid, carrying 23S rRNA gene with the single mutation A1067U, was used as a control. The expression of 23S rRNA genes containing mutations A1912G, Ψ1917C and A1919G resulted in very strong growth inhibition 2–3 hrs post-induction leading to a complete cessation of cell growth well below the cell densities, which were reached by uninduced cultures (Fig 2). Induction of 23S rRNA variants carrying mutations Ψ1915C and A1916G resulted in modestly increased doubling times but nevertheless allowed the cultures to reach maximal cell densities similar to uninduced cultures (Fig 2). Mutations Ψ1911C, A1913G, C1914A and A1918G had little or no effect on growth (Fig 2).
Incorporation of mutant 23S rRNA into poly-ribosomes
Quantitative monitoring of the proportion of mutant 23S rRNAs in translating polysomes, initiating 70S ribosomes and translationally idle 50S subunits enables us to assign defects in translation to the initiation phase (when mutant rRNA is under-represented in 70S and over-represented in 50S fractions) or the elongation phase (when mutant rRNA is under-represented in polysomes) of translation. Furthermore, possible changes in relative sizes of the polysome, 70S or 50S fractions can provide useful information on translational competency of the mutant cells. In addition, changes in the overall shape and number of the peaks can point to defects in ribosome assembly.
The marker-mutation A1067U was used to determine the fraction of mutant 23S rRNA. Ribosomes were fractionated by 15%-40% sucrose gradient centrifugation and rRNA was extracted from tetrasome, trisome, disome, 70S and 50S fractions. Relative proportions of plasmid-encoded and chromosomally encoded rRNAs were determined by a standard primer extension assay [23].
Expression of the single mutation A1067U 23S rRNA variant results in around 45% of the polysome and 70S pools consisting of mutant ribosomes (Fig 3). 23S rRNA variant A1912G is under-represented in polysomes, but not in the 70S ribosomes (Fig 3). Its progressively larger deprivation in larger polysomes (30% presence in disomes, 15% in trisomes, and 10% in tetrasomes) is consistent with defects in the elongation phase of translation of the mutant ribosomes rather than, for example, in a late initiation step.
Mutation A1913G exhibits a modest counter-selection of mutant ribosomes in the polysome fractions reaching approximately two-fold deprivation in the tetrasomes (Fig 3).
23S rRNA variant Ψ1917C exhibits a relatively modest fractional deprivation in polysomes and in 70S ribosomes (Fig 3). Interestingly, it leads to a large reduction in the amount of polysomes; indeed so large that we were unable to collect tetrasomes for analysis. Therefore, expression of this mutant must reduce the ability of the wild-type ribosomes to engage in translation.
Mutation A1919G leads to a modest deprivation of plasmid borne 23S rRNA in the 70S ribosomes, in addition to a large progressive deprivation in the polysomes fractions (Fig 3). This is, once again, suggestive of a mostly elongation-level defect in translation by the mutant ribosomes. In addition, expression of the 23S rRNA variant A1919G led to appearance of an extra gradient peak, corresponding to particles sedimenting approximately as 40S (Fig 3). However, the amount of the 40S particles varied widely between experiments from nonexistent to nearly the levels of the 50S subunits (data not shown). In spite of the variable results we tentatively suggest that the mutation A1919G affects 50S subunit assembly.
23S rRNA variant A1916G exhibits a large deprivation in 70S ribosomes and is nearly absent in polysomes (Fig 3). In addition, 23S rRNA variant A1916G reproducibly exhibited enlarged and widened 50S peak, suggestive of conformational heterogeneity in the 50S population (Fig 3). Conformational heterogeneity can be caused by a defect in ribosomal large subunit assembly. It is possible that the transition A1916G confers a defect of ribosome large subunit assembly. On the other hand, lack of 23S rRNA variant A1916G in the 70S and polysome fractions can be caused by a defect in association with the 30S subunit, which in turn can cause an initiation defect.
Expression of mutations Ψ1911C, C1914A, Ψ1915C or A1918G did not lead to significant changes in the fraction of mutant ribosomes or in the appearances of the gradient profiles (Fig 3).
Cell-free translation of poly(U) by mutant ribosomes
For in vitro translation, tight-couple 70S ribosomes were isolated from induced XL1 lysates by sucrose gradient ultracentrifugation. The second site mutation, A1067U, confers thiostrepton resistance to plasmid-borne 50S ribosomes enabling studies of the cell free translation of mutant ribosomes through inactivation of the wild-type ribosomes by thiostrepton. In the presence of five-fold molar excess of thiostrepton, poly(U)-directed translation of wild-type ribosomes was inhibited by 97–99% [22]. The ribosomes isolated from induced cells that harbor plasmid-encoded A1067U mutant 23S rRNA exhibited 30% thiostrepton resistance during poly(U) translation (Fig 4). If A1912G, A1916G or A1919G mutation was added as the second mutation to the A1067U, thiostrepton-resistance dropped to nearly zero (Fig 4). This means that ribosomes harboring mutations at positions 1912, 1916 and 1919 are completely inactive in the poly(U)-directed translation system. Mutation Ψ1917C causes a three-fold reduction in poly(Phe) synthesis of the mutant ribosomes (Fig 4). Mutation Ψ1915C cause a two-fold reduction in cell-free translation capability of the mutant ribosomes (Fig 4). Mutation A1913G exhibits slightly reduced levels in poly(U) translation. Mutations Ψ1911C, C1914A and A1918G have no effect on poly(U) translation.
Discussion
We have dissected the function of the helix-loop 69 of 23S rRNA by subjecting mutants of its loop residues to various functional tests. The results of assays of cell growth, polysome incorporation, and cell-free translation pointed to the same mutations (A1912G, A1916G, Ψ1917C and A1919G) as seriously compromised. However, the phenotypes of the aforementioned mutations fall into three distinct types.
First, mutations at positions 1912 and 1919 are defective in the elongation phase of protein synthesis, both having reduced amounts of mutant ribosomes in the polysomes, but are abundant in the 70S ribosomes. The gradual decrease of the A1912G and A1919G ribosomes in successive polysomal fractions is consistent with placement of one mutant ribosome per mRNA. This can happen when the mutant ribosomes are able to initiate translation but cannot undergo into elongation phase. The mutant ribosomes are also completely inactive during in vitro translation, showing that the under-representation of mutant ribosomes in cellular polysomes is a direct consequence of the inactivity of the mutant ribosomes. Although mutant ribosomes are largely excluded from the polysomes, the total amount of polysomes in the relevant sucrose gradient fractions is not appreciably reduced (Fig 3). It should be noted that expression of 30–40% of mutant ribosomes in the background of wild-type ribosomes rapidly leads to a complete cessation of cell growth. It is likely that this pseudo-dominant phenotype is caused by "choking" of the translation by freezing the ribosomes (both mutant and wt) on the cellular mRNAs.
Second, as indicated by the widened 50S peak in the sucrose gradient, ribosomes carrying the mutation A1916G are structurally heterogeneous and thus apparently defective in the assembly of mutant 50S subunits. This conjecture is further supported by the strongly reduced activity of the mutant ribosomes in cell free translation and by their strong counter-selection in the 70S ribosome pool. Yet this mutation leads only to a modest retardation in cell growth. Therefore, complete inactivation of the mutant ribosomes in the wild-type background is not in itself detrimental to cell growth.
Third, the Ψ1917C mutation leads to relatively modest effects in cell free translation and in the fractions of mutant ribosomes in the polysomes. Therefore, unlike for A1912G and A1919G, the induction of Ψ1917C-carrying 23S rRNA does not freeze the ribosomes (mutant or wild-type) on mRNAs. Yet it leads to a complete stop in cell growth 3–4 hours after the induction of mutant 23S rRNA synthesis. The Ψ1917C is also the only mutation that leads to a clearly reduced polysome pool 2 hours after induction. We believe that this reduction in the ability of both mutant and wild-type ribosomes to enter the polysome pool could explain the observed drastic growth defect by strong reduction in the cellular protein synthesis levels. Such an effect could, in principle, be achieved by sequestering of an essential factor for translation by the mutant ribosomes.
A number of nucleotides in the loop of helix 69 have been implicated in binding of the A site tRNA (A1913-Ψ1915) and as components of the intersubunit bridge B2a (A1912-C1914, A1918) based on structural [1,3] and modification interference [6] studies (Fig 1). Surprisingly, of the aforementioned five nucleotides, only the mutation A1912G exhibited a strong phenotype. The DMS-modification of the N1 position of A1912 was previously shown to be detrimental to 70S ribosome formation in vitro [6]. However, the A1912G mutation seems to exert its strong effect on translation at the level of elongation, rather than by inhibiting 70S ribosome formation (Fig 3). Therefore, we failed to confirm the functional importance of any of the proposed H69 contacts with tRNA or the SSU by the mutagenesis approach. This is reminiscent of the results of O'Connor and Dahlberg who disrupted the stem of H69 by introducing a C1909:C1921 mismatch and found no growth effect or defects in translation [19]. Yet, this disruption should fall squarely in the middle of the H69 backbone-to-backbone contact area with the P site bound tRNA [1] (Fig 1). Also, of the three conserved pseudouridines in H69, only mutation of the Ψ1917 has a strong effect. Notably, out of three pseudouridine residues in this region, only the Ψ1917 is universally conserved [16]. Thus, Ψ1917 is likely to have a unique function.
Sucrose gradient pattern suggested that the mutation A1916G affects 50S assembly. As the H69 forms a spindle-like structure on the interface side of the large subunit [1,8], it is difficult to imagine how the mutation in the loop region could affect 50S structure, and thereby its assembly. On the other hand, A1916G transition can affect ribosomal subunit interaction. It is possible that association of the subunits is important for final maturation of the 50S subunits.
Conclusion
We scanned the loop of helix 69 by mutagenesis and analyzed the mutant ribosomes using a plasmid-borne IPTG-inducible expression system. We assayed the effects of 23S rRNA mutations on cell growth, contribution of mutant ribosomes to cellular polysome pools and the ability of mutant ribosomes to engage in cell-free translation. Mutations A1912G, and A1919G have very strong growth phenotypes, are inactive during in vitro protein synthesis and are under-represented in the polysomes. Mutation Ψ1917C has a very strong growth phenotype and leads to a general depletion of the cellular polysome pool. Mutation A1916G, having a modest growth phenotype, is apparently defective in the assembly of the mutant 50S subunits or in the 70S formation.
Methods
Plasmids, strains and mutagenesis
The host strain for plasmids was E. coli XL1-Blue (supE44 hsdR17 recA1 endA1 gyrA46 thi relA1 lac- F' [proAB+ lacIq lacZΔ M15 Tn10(tetr). Plasmid ptBsB1067T [20,24] containing the BstE II-BamH I fragment of the rrnB operon (tRNAGlu2-23 S rRNA and the 5 S rRNA genes) under the control of the inducible tac promoter was used to construct the mutations. A single point mutation at position A1067 to T confers the thiostrepton resistance of plasmid borne ribosomes [21].
Site-directed mutagenesis was performed by the PCR-based approach of Mikaelian [25]. All PCR fragments were fully sequenced after cloning into the 23S rRNA gene in ptBsB1067U (the SalI-SacII fragment was replaced).
Measurement of cell growth
Cell growth was measured at OD600 using E. coli strain XL-1 Blue containing mutant 23S rRNA genes in the plasmid pBsB1067U and a low copy-number plasmid pREP4, which expresses additional lac repressor protein [26]. Cells were grown at 37°C in rich liquid media (2xYT) with ampicillin (100 μg/ml) and kanamycin (50 μg/ml) and mutant 23S rRNA expression was induced with 1 mM IPTG. Culture densities were monitored for 12 hrs after induction.
Preparation of the ribosomes and analysis of mutant rRNA content in polysomes
E. coli strain XL1-Blue transformed with the ptBsB1067T derivative plasmids were grown at 37° C in 2xYT medium (16 g/l tryptone, 10 g/l yeast extract, 5 g/l NaCl) supplemented with ampicillin (100 μg/ml). Ribosomes were isolated from cells after induction with IPTG (1 mM) at A600 = 0.2 for 2 hours. Bacteria were collected by low-speed centrifugation and resuspended in lysis buffer (16% sucrose (w/v), 6 mM MgCl2, 60 mM NH4Cl, 60 mM KCl, 50 mM Tris-HCl pH-8.0, 6 mM β-mercaptoethanol). After addition of lysozyme (0.5 mg/ml final concentration) the cells were lysed by freezing and thawing 3 times. S-30 lysate was prepared by centrifugation at 12.000 g for 30 min in an SS34 rotor (Sorvall). The volume of the lysate was doubled with the buffer LLP (12 mM MgCl2, 60 mM NH4Cl, 60 mM KCl, 20 mM Tris-HCl pH-8.0, 6 mM β-mercaptoethanol), loaded onto a 5 ml sucrose cushion (20% sucrose, 12 mM MgCl2, 500 mM NH4Cl, 50 mM Tris-HCl pH-8.0, 6 mM β-mercaptoethanol) followed by the centrifugation for ω2t = 5.0 × 1011 using a Beckman SW-41 rotor. Crude ribosomes were dissolved in buffer LLP and stored in small aliquots at -80°C.
For preparation of polysomes, 70 S ribosomes and 50 S ribosomes, the cell lysates were diluted 2 times with LLP buffer and loaded onto a 15–40% sucrose gradient in LLP buffer and centrifuged for ω2t = 3.5 × 1011 in a Beckman SW-28 rotor. Polysomal, 70 S, 50 S, and 30 S gradient fractions were collected and precipitated with 2.5 volumes of ice-cold ethanol. rRNA was prepared using modified protocol of [27] For the extraction of rRNA ribosomes were dissolved in 200 μl water and 1 ml of PN solution (Qiagen, Cat. No. 19071) was added. Ribosomal proteins were extracted by vigorous shaking for 20 min at room temperature. 20 μl 50% silica suspension in water was added and RNA was bound for additional 10 min at room temperature with gentle mixing. Silica was pelleted by centrifugation at 6000 rpm for 30 sec and washed twice with 70% ethanol. RNA was eluted with 50 μl of water (10 min at room temperature). The proportion of plasmid-encoded 23S rRNA was determined by the modified primer extension protocol of Sigmund et al. [23] using the A1067T as the marker-mutation [22]. The resulting DNA fragments were resolved in 12% polyacrylamide-urea gel. Autoradiograms were digitalized using PhosphoImager (Molecular Dynamics) and quantified using the ImageQuant software (Molecular Dynamics).
Poly(U)-directed protein synthesis
Poly(U) translation was performed essentially as described in Saarma and Remme [20]. Thiostrepton (Calbiochem) was dissolved in dimethylsulfoxside (DMSO) to 1 mM and used for inhibiting wild-type ribosomes. 0.5 A260 units of ribosomes were preincubated at 37°C for 15 min in the presence or absence of 7.5 μM thiostrepton and 0.02 mg poly(U) in 50 μl buffer LLP followed by the addition of 50 μl of factor mix containing 0.02 mg bulk tRNA (Boehringer Mannheim), 2 mM ATP, 0.5 mM GTP, 8 mM phosphoenolpyruvate (PEP), 2 μM pyruvate kinase, 0,01 mM [14C]Phe (150 cpm/pmol, Amersham) and 0.2 mg S-100 enzymes. After 30 min incubation at 37°C, reactions were stopped by addition of 1 ml 5% trichloroacetic acid (TCA) and heated for 20 min at 95° C. Precipitates were collected onto GF/A filters (Whatman) and counted for radioactivity. Thiostrepton resistance of the ribosomes was calculated by dividing TCA-insoluble radioactivity obtained in the presence of thiostrepton to that obtained in the absence of the drug.
Authors' contributions
AL and DK did most of the experimental work and participated in planning and design of the experiments. ÜM participated in the analysis of mutant ribosomes and writing of the manuscript. JR conceived the study, participated in its design, and helped to write the manuscript.
Acknowledgements
We thank Tanel Tenson and Silja Kuusk (both Tartu University) for critically reading the manuscript. This work was supported by the Howard Hughes Medical Institute International Research grant No. 55000332 and Estonian Science Foundation Grant No. 5822
Figures and Tables
Figure 1 A scheme of E. coli 23S rRNA helix 69. Nucleotides implicated in contacts with 30S subunit are in red [1,3,6]. Proposed contact areas with A and P site tRNAs [1] are shown in yellow and green boxes. Numbers of the pseudouridine residues are indicated according to standard E. coli 23S rRNA numeration.
Figure 2 Growth of XL-1 cells expressing mutant 23S rRNA. The expression of the mutant rRNA was induced at O.D.600 = 0.05 (0 min time point) by addition of IPTG. Y-axis shows the optical densities of the cultures at time points. Control plasmid is ptBsB1067T. Density of the bacterial culture grown in the absence of IPTG is shown by blue circles and in the presence of IPTG is shown by red squares.
Figure 3 Distribution of plasmid encoded 23S rRNA in ribosomal fractions. Mutant plasmids were expressed in the strain XL1-Blue and ribosomes were fractionated by sucrose gradient centrifugation. The percentage of plasmid-encoded 23S rRNA in the ribosomal fractions was determined by RNA sequencing. The gradient profile and the percentages of the plasmid-encoded 23S rRNA in the corresponding gradient fractions is shown for each mutant. The arrow shows the direction of sedimentation. TET- tetrasomes, TRI- trisomes, DI- disomes. Control plasmid is ptBsB1067T.
Figure 4 Cell-free Poly(U)-dependent protein synthesis activity in the presence of thiostrepton. Ratio of protein synthesis in the presence and absence of the drug is given as resistance. Standard deviations from 2–5 independent experiments are shown. Control plasmid is ptBsB1067T.
==== Refs
Yusupov MM Yusupova GZ Baucom A Lieberman K Earnest TN Cate JH Noller HF Crystal structure of the ribosome at 5.5 A resolution Science 2001 292 883 896 11283358 10.1126/science.1060089
Gabashvili I Agrawal R Spahn CMT Grassucci RA Svergun DI Frank J Penczek P Solution Structure of the E. coli70S Ribosome at 11.5 A Resolution Cell 2000 100 537 549 10721991 10.1016/S0092-8674(00)80690-X
Gao H Sengupta J Valle M Korostelev A Eswar N Stagg SM Roey PV Agrawal RK Harvey SC Sali A Chapman MS Frank J Study of the Structural Dynamics of the E. coli70S Ribosome Using Real-Space Refinement Cell 2003 113 789 801 12809609 10.1016/S0092-8674(03)00427-6
Merryman C Moazed D Daubresse G Noller HF Nucleotides in 23 S rRNA protected by the association of 30 S and 50 S ribosomal subunits Journal of Molecular Biology 1999 285 107 113 9878392 10.1006/jmbi.1998.2243
Mitchell P Osswald M Brimacombe R Identification of Intermolecular RNA Cross-Links at the Subunit Interface of the Escherichia-Coli Ribosome Biochemistry 1992 31 3004 3011 1372517 10.1021/bi00126a023
Maivali U Remme J Definition of bases in 23S rRNA essential for ribosomal subunit association RNA 2004 10 600 604 15037769 10.1261/rna.5220504
Dallas A Noller HF Interaction of Translation Initiation Factor 3with the 30S Ribosomal Subunit Molecular Cell 2001 8 855 864 11684020 10.1016/S1097-2765(01)00356-2
Harms J Schluenzen F Zarivach R Bashan A Gat S Agmon I Bartels H Franceschi F Yonath A High Resolution Structure of the Large RibosomalSubunit from a Mesophilic Eubacterium Cell 2001 107 679 688 11733066 10.1016/S0092-8674(01)00546-3
Bashan A Zarivach R Schluenzen F Agmon I Harms J Auerbach T Baram D Berisio R Bartels H Hansen HAS Fucini P Wilson D Peretz M Kessler M Yonath A Ribosomal Crystallography:Peptide Bond Formation andIts Inhibition Biopolymers 2003 70 19 41 12925991 10.1002/bip.10412
Valle M Zavialov A Li W Stagg SM Sengupta J Nielsen NC Nissen P Harvey SC Ehrenberg M Frank J Incorporation of aminoacyl-tRNA into the ribosome as seen by cryo-electron microscopy Nature Structural Biology 2003 10 899 1074 14566331 10.1038/nsb1003
Spahn CMT Gomez-Lorenzo M Grassucci RA Jorgensen R Andersen GR Beckmann R Penczek P Ballesta JPG Frank J Domain movements of elongation factor eEF2 and the eukaryotic 80S ribosome faciliate tRNA translocation EMBO J 2004 23 1008 1019 14976550 10.1038/sj.emboj.7600102
Rawat UB Zavialov A Sengupta J Valle M Grassucci RA Linde J Vestergaard B Ehrenberg M Frank J A cryo-electron microscopic study of ribosome-bound termination factor RF2 Nature 2003 421 87 90 12511960 10.1038/nature01224
Klaholz BP Myasnikov AG Van Heel M Visualization of release factor 3 on the ribosome during termination of protein synthesis Nature 2004 427 862 865 14985767 10.1038/nature02332
Agrawal RK Sharma MR Kiel MC Hirokawa G Booth TM Spahn CM Grassucci RA Kaji A Frank J Visualization of ribosome-recycling factor on the Escherichia coli 70S ribosome: functional implications Proc Natl Acad Sci U S A 2004 101 8900 8905 15178758 10.1073/pnas.0401904101
Valle M Gillet R Kaur S Henne A Ramakrishnan V Frank J Visualizing tmRNA entry into a stalled ribosome Science 2003 300 127 130 12677067 10.1126/science.1081798
Ofengand J Ribosomal RNA pseudouridines and pseudouridine synthases FEBS Letters 2002 514 17 25 11904174 10.1016/S0014-5793(02)02305-0
Raychaudhuri S Conrad J Hall BG Ofengand J A pseudouridine synthase required for the formation of two universally conserved pseudouridines in ribosomal RNA is essential for normal growth of Escherichia coli RNA 1998 4 1407 1417 9814761 10.1017/S1355838298981146
Ofengand J Malhotra A Remme J Gutgsell NS Del Campo M Jean-Charles S Peil L Kaya Y Pseudouridines and pseudouridine synthases of the ribosome Cold Spring Harb Symp Quant Biol 2001 66 147 159 12762017 10.1101/sqb.2001.66.147
O´Connor M Dahlberg AE The Involvement of Two Distinct Regions of 23 SRibosomal RNA in tRNA Selection Journal of Molecular Biology 1995 254 838 847 7500354 10.1006/jmbi.1995.0659
Saarma U Remme J Novel mutants of 23S RNA: characterization of functional properties Nucleic Acids Res 1992 20 3147 3152 1377819
Thompson J Cundliffe E Dahlberg AE Site-directed mutagenesis of Escherichia coli 23 S ribosomal RNA at position 1067 within the GTP hydrolysis centre J Mol Biol 1988 203 457 465 2462056 10.1016/0022-2836(88)90012-5
Liiv A Remme J Base-pairing of 23 S rRNA ends is essential for ribosomal large subunit assembly J Mol Biol 1998 276 537 545 9551095 10.1006/jmbi.1997.1532
Sigmund CD Ettayebi M Borden A Morgan EA Antibiotic resistance mutations in ribosomal RNA genes of Escherichia coli Methods Enzymol 1988 164 673 690 3071688
Lewicki BT Margus T Remme J Nierhaus KH Coupling of rRNA transcription and ribosomal assembly in vivo. Formation of active ribosomal subunits in Escherichia coli requires transcription of rRNA genes by host RNA polymerase which cannot be replaced by bacteriophage T7 RNA polymerase J Mol Biol 1993 231 581 593 8515441 10.1006/jmbi.1993.1311
Mikaelian I Sergeant A A general and fast method to generate multiple site directed mutations Nucleic Acids Res 1992 20 376 1311073
Farabaugh PJ Schmeissner U Hofer M Miller JH Genetic studies of the lac repressor. VII. On the molecular nature of spontaneous hotspots in the lacI gene of Escherichia coli J Mol Biol 1978 126 847 857 370408 10.1016/0022-2836(78)90023-2
Boom R Sol C Beld M Weel J Goudsmit J Wertheim-van Dillen P Improved silica-guanidiniumthiocyanate DNA isolation procedure based on selective binding of bovine alpha-casein to silica particles J Clin Microbiol 1999 37 615 619 9986822
|
16053518
|
PMC1190176
|
CC BY
|
2021-01-04 16:22:25
|
no
|
BMC Mol Biol. 2005 Jul 29; 6:18
|
utf-8
|
BMC Mol Biol
| 2,005 |
10.1186/1471-2199-6-18
|
oa_comm
|
==== Front
BMC Med GenetBMC Medical Genetics1471-2350BioMed Central London 1471-2350-6-281604280410.1186/1471-2350-6-28Case ReportCardiac conduction abnormalities and congenital immunodeficiency in a child with Kabuki syndrome: Case report Shah Maulik [email protected] Brian [email protected] Melissa [email protected] Daphne E [email protected] Alan [email protected] Division of Medical Genetics, Department of Pediatrics, Saint Louis University, 1465 South Grand Blvd., Saint Louis, MO, 63104-1095, USA2 Saint Louis University Cancer Center, 3655 Vista Ave., Saint Louis, MO, 63110, USA3 Department of Pathology, Saint Louis University, 1402 South Grand Blvd., Saint Louis, MO, 63104, USA4 School of Medicine, Saint Louis University, 1402 South Grand Blvd., Saint Louis, MO, 63104, USA5 Division of Allergy and Immunology, Department of Pediatrics, Saint Louis University, 1565 South Grand Blvd., Saint Louis, MO, 63104-1095, USA2005 25 7 2005 6 28 28 10 12 2004 25 7 2005 Copyright © 2005 Shah et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Since it's recognition in 1981, a more complete phenotype of Kabuki syndrome is becoming evident as additional cases are identified. Congenital heart defects and a number of visceral abnormalities have been added to the typical dysmorphic features originally described.
Case Report
In this report we describe the clinical course of a child diagnosed with Kabuki syndrome based on characteristic clinical, radiological and morphologic features who died of a cardiac arrhythmia at 11-months of age. This infant, however, had abnormal pulmonary architecture and alterations in his cardiac conduction system resulting in episodes of bradycardia and asystole. This child also had an immunological phenotype consistent with common variable immunodeficiency. His clinical course consisted of numerous hospitalizations for recurrent bacterial infections and congenital hypogammaglobulinemia characterized by low serum IgG and IgA but normal IgM levels, and decreased antibody levels to immunizations. T-, B- and NK lymphocyte subpopulations and T-cell function studies were normal.
Conclusion
This child may represent a more severe phenotype of Kabuki syndrome. Recurrent infections in a child should prompt a thorough immunological evaluation. Additionally, electrophysiology testing may be indicated if cardiopulmonary events occur which are not explained by anatomic defects.
==== Body
Background
Kabuki syndrome (KS) is a multiple congenital anomalies/mental retardation (MCA/MR) syndrome of unknown cause. It was first described by Niikawa et al. [1] in 1981 and is also known as Niikawa-Kuroki syndrome. In the interim, more than 300 patients of both Asian and non-Asian heritage have been reported[2]. Although a few associated cytogenetic abnormalities have been reported[3], the majority of patients have normal chromosomes and no genetic basis has to date been identified. The inheritance pattern of this disorder has not been established. Most cases are sporadic, but a few families with multiple generations of affected individuals have been reported suggesting autosomal dominant inheritance[4]. Diagnosis is based on characteristic dysmorphic features, visceral abnormalities and clinical features. Common dysmorphic features include arched eyebrows, eversion of the lateral lower eyelid, cleft palate, bifid uvula, and persistent fetal finger tip pads[5]. Visceral abnormalities often include structural heart defects[6] and abdominal wall defects [7]. Other clinical features may include microcephaly and post-natal growth retardation. In addition, recurrent infections, principally otitis media, as well as upper respiratory tract infections and pneumonia occur in patients with KS[8]. However, antibody immune defects have been described in only isolated patients[9].
Structural heart defects are encountered in 32% – 58% of children and no specific congenital heart defect predominates[2]. Although heart defects in the ventricular septum may cause cardiac rhythm disturbances, in our patient, bradyarrythmia progressing to asystole and death was the primary clinical problem resulting from an abnormal conduction system.
Case presentation
J.R. was a 3210 gram child born to a 21-year-old gravida 1, para 1 mother by vaginal delivery at 36 weeks of gestation. The mother received appropriate prenatal care and the pregnancy was uncomplicated. There was no gestational history of tobacco, alcohol, illicit substance or medication use. A cardiac echogenic focus and possible anatomic heart defect was noted on routine prenatal ultrasound examination and the child was transferred to the neonatal intensive care unit for management at birth. A thoracic echocardiogram showed mitral stenosis, aortic stenosis and coarctation of the aorta. Surgical repair of the coarctation was conducted. A right diaphragmatic eventration was simultaneously repaired. During this initial hospitalization no cardiac arrhythmias were monitored on telemetry. Due to the structural heart defects a chromosome analysis including fluorescence in situ hybridization for both DiGeorge loci was performed and revealed no abnormalities. A diagnosis of Kabuki syndrome was made by two independent geneticists based on phenotypic features. Clinical features in this child consistent with the diagnosis of Kabuki syndrome included arched eyebrows with sparse hair laterally, eversion of the lateral lower eyelid, a cleft palate, bifid uvula, broad nasal root with depressed nasal tip, and persistent fetal finger tip pads with otherwise normal dermatoglyphics. Neuroimaging revealed severe corpus callosum hypoplasia and ophthalmologic examination showed optic nerve atrophy bilaterally. Congenital hypothyroidism was also detected. Further cytogenetic analysis using telomere probes was conducted without detection of abnormalities.
The subsequent hospital course was complicated by numerous infections secondary to hypogammaglobulinemia and he was finally discharged to home at 3 months of age. After one week at home, he was noted to be cyanotic and non-responsive and was re-admitted to the hospital where the patient was considered to be septic and placed on broad spectrum antibiotics.
At 6.5 months of age, he was admitted to the hospital with a urinary tract infection. Sinus bradyarrythmia was recorded associated with hypoxemia with prolonged pauses progressing to asystole. Resuscitation with epinephrine was successful. Bronchoscopy was performed and showed normal airways. A cardiac echocardiogram showed diastolic dysfunction. He was started on supplemental O2 at 1/8 L by nasal canula and received furosemide, aldactazide and verapamil. No other episodes of cardiac rhythm disturbance were noted on telemetry during the remaining hospital stay.
At 9 months of age, he was admitted to the hospital in respiratory distress. He was ventilated mechanically and monitored in the ICU. While on cardiac monitoring he developed a tachyarrhythmia without 1:1 conduction then prolongation of the QT interval followed by an idioventricular rhythm which slowed to asystole. External pacing failed. After alternate doses of epinephrine and atropine the patient returned to sinus rhythm. Thoracic echocardiogram showed normal ventricular function with normal velocities across inlet and outlet valves. Troponin levels remained less than 1 and there was no other evidence of ischemia. Because of the previous episodes of bradyarrythmia, a dual chamber epicardial pacemaker was placed without complications. His ventricular and diastolic function normalized and diuretics were discontinued. He remained hemodynamically stable but continued hospitalization was required for various nosocomial infections. The family elected to make an advance directive to prevent further resuscitative efforts. Approximately one month later during hospitalization for respiratory bronchiolitis, he had another episode of bradyarrythmia which progressed to asystole and death. The pacemaker was queried and found to have functioned within normal parameters. The family consented for a limited autopsy.
Cardiovascular autopsy findings
The heart weighed twice the normal expected weight and there was biventricular hypertrophy, with the right ventricular wall thickness being about four times the normal thickness and the left ventricular wall thickness about twice normal (Figure 1). The right atrium and the coronary sinus were dilated and the left atrium was small with endocardial fibroelastosis (Figure 2). The mitral valve was stenotic, the circumference being about 3/4ths the normal circumference. The valve leaflets were thick and myxoid, and the chordae tendinae were shortened. The posterior leaflet of the mitral valve was directly inserted into the papillary muscle. There was mild aortic stenosis and the aortic valve leaflets were thick and dysplastic. The pulmonary valve circumference was about one-third greater than normal. The membranous portion of the interventricular septum was about three times the normal length and mapping of the junctional tissue revealed that the atrioventricular node was displaced caudally and the bifurcation of the Bundle of His was likewise displaced caudally (Figure 3). The valve leaflet was made up of primarily mucopolysaccharide material (blue stain). The placement of the pacemaker was verified to be appropriate and dense adhesions were present between the pacer wires and the abdominal wall and bowel loops.
Postmortem pulmonary angiogram revealed severe pruning of the pulmonary vascular tree (Figure 4) and absence of the background 'blush' produced by filling of intracinar arteries. Microscopic examination revealed marked luminal narrowing or fibrous occlusion of intra-acinar arteries (Figure 5). In addition there was marked dilatation of lymphatics within pulmonary septa.
Immunological phenotype
The patient exhibited recurrent infections requiring numerous hospitalizations. These included Klebsiella pneumonia, RSV pneumonia, Enterococcus sepsis, Candida albicans urinary tract infection, Enterobacter urosepsis, sinusitis and otitis media.
Immune evaluation revealed persistent lymphopenia; however, the percentages of T-, B-, and NK-cell subpopulations were normal (Table 1). Furthermore, lymphoproliferative responses to mitogens PHA, PWM and alloantigens were normal. However, lymphoproliferative response to Concavalin A stimulation was absent. Serum IgG and IgA levels were markedly decreased, and antibody responses to tetanus toxoid and Hemophilus influenzae type B (HiB) were decreased. Following immunization with conjugated pneumococcal vaccine (Prevnar), antibody responses were decreased to 4 of 7 serotypes. Serum albumin levels were normal, and there was no evidence of protein loss through the gastrointestinal or urinary systems, and there was no evidence of chylous thorax. Thus, IVIG therapy was initiated. However, the patient shortly succumbed from cardiac arrhythmia. Autopsy revealed severe thymic involution (the thymus weighed 1 g.) and lymphoid depletion in the spleen.
Additional autopsy findings
Other findings included nesidioblastosis, vacuolation of the adrenal cortex, undescended testes, contraction band necrosis of the muscularis propria of the gastro-intestinal tract and growth retardation with growth parameters being < 3rd percentile.
Conclusion
Since its original description in 1981, there are now numerous reports in the literature across ethnic lines defining the phenotype of Kabuki syndrome. However, the etiology and genetics of Kabuki syndrome are poorly understood. Although a few cytogenetic abnormalities in patients have been reported including X chromosome rings [10], translocations[11], inversions[11] and duplications[12] as well as a variety of autosomal chromosomal defects[8], the majority of patients have normal chromosomes and no specific genes to date have been identified. The majority of cases are sporadic, however, a few families with multiple generations of affected persons suggests autosomal dominant inheritance [4]. In the absence of molecular diagnosis, a confirmatory diagnosis is based on clinical judgment and the reported phenotypic abnormalities associated with this syndrome have been expanded since the initial case reports. Our patient had characteristic dysmorphology, visceral abnormalities and clinical features of Kabuki syndrome. The diagnosis was determined by independent evaluation by two separate geneticists. In addition, he had cognitive delay, microcephaly, growth retardation, cleft palate, hypothyroidism, coarctation of the aorta, and diaphragmatic eventration.
Of the visceral abnormalities associated with Kabuki syndrome, congenital heart disease appears to be the most common with rates ranging from 32% to as high as 58%[2]. In regards to the specific anatomic abnormalities, there is some dispute in the literature. The earlier studies were not conclusive for specific cardiac defects while the later studies show a greater association of coarctation of the aorta. In none of these reviews or in isolated case reports has there been a report of an abnormal cardiac conduction system or reports of arrhythmia. This child had abnormalities in his cardiac conduction system that eventually lead to numerous episodes of bradycardia and eventually to asystole. Although he had structural cardiac abnormalities which can often result in alterations in chamber size and predispose to arrhythmia, his surgical correction was appropriate and there does not appear to have been secondary strain on the ventricles. A noted abnormality on evaluation of his AV node and Purkinje tracts was their altered placement. After his asystolic episode leading to death, his pacemaker was queried and showed normal functioning. During the time of bradyarrythmia it appeared to fire appropriate without normal capture. The abnormal placement of his conduction system likely contributed to the lack of pacemaker pickup.
Increased susceptibility to infections has been reported as a frequent complication in KS. Recurrent otitis media has been reported in 63% of patients with KS. This has often been attributed to anatomic reasons secondary to the cleft palate; however, this occurs in only 35% of patients with clefting of the palate not associated with a syndromic diagnosis. In addition, some of the patients have also had bacterial pneumonia and one patient had Aspergillus fumigatus pneumonia. Hypogammaglobulinemia with low serum IgG and IgA levels but normal IgM level has been previously reported in four patients with KS[13,14]. These patients were older than our patient when diagnosed with hypogammaglobulinemia. We believe this is the first report of an infant with KS diagnosed with hypogammaglobulinemia. Furthermore, our patient displayed the same pattern of hypogammaglobulinemia, namely hypogammaglobulinemia with normal IgM. Chrzanowska et al. [9] also diagnosed an associated T-cell defect with the hypogammaglobulinemia in a 10-year-old boy. Though our patient did have lymphopenia (1818 cells/mm3), percentages of T-, B- and NK-cell populations were normal. Furthermore, naïve T-cells were normal, CD3+CD45RA+, 79%. This is contrast to decreased naïve CD4+ T-cells reported by Chrzanowska [9]. Importantly, lymphoproliferative responses to PHA, PWM and alloantigens were normal. However, response to Concavalin A stimulation was absent, perhaps indicating a T suppressor defect.
Immune cytopenias have been previously reported in KS. Niikawa et al. [13] reported hemolytic anemia and Watanabe et al. [14] reported idiopathic thrombocytopenia. Autoimmune cytopenias have been associated with common variable immunodeficiency (CVID) [15] and hypogammaglobulinemia with normal IgM deficiency (hyper-IgM syndrome) [16]. The pattern of low IgG and IgA with normal IgM concentrations may be seen in both CVID and hyper-IgM syndromes. In KS, the hypogammaglobulinemia has generally been described as acquired hypogammaglobulinemia, occurring in older children. Our patient is the first description of a probable congenital diagnosis of hypogammaglobulinemia most likely from CVID.
Herein we describe a male infant with Kabuki syndrome presenting with cardiac arrhythmia and congenital immunodeficiency. Based on our single case report, we do not advocate changes in management of patients with Kabuki syndrome. However, those with cardiac abnormalities should be monitored closely during times of hospitalization for cardiac arrhythmias. Those children presenting with arrhythmias may warrant electrophysiological evaluation. Additionally, it may be beneficial for children with recurrent infections associated with a diagnosis of Kabuki syndrome to have a thorough immunologic evaluation. The association of hypogammaglobulinemia and KS supports a genetic etiology. Future studies of the hypogammaglobulinemia B-cell subsets, expression of IgG and IgA surface B-cells, and IgM to IgG isotype switching would better characterize the immunologic phenotype in this syndrome. Appropriate prevention strategies should be implemented to decrease the likelihood of a catastrophic infection.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
B.B. and M.M were responsible for chart review and organization of pertinent material and contributed to the writing and editing of the manuscript. B.B. and D.D were responsible for procurement and analysis of pathologic specimens. A.K. was a significant contributor in writing this manuscript and was responsible for the conduction and interpretation of immunologic studies. M.S. initially diagnosed this child and was responsible for the final writing, editing, organization and submission of this manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors express their appreciation to Theresa Forsythe for secretarial support.
Figures and Tables
Figure 1 Anatomy of right ventricle and atrium. A. Dilated coronary sinus. B. Dysplastic tricuspid valve. C. Short thickened chordae tendinae almost implanted into papillary muscle. D. Right ventricular hypertrophy with ventricular wall thickness of 8 mm. E. Pacer wire.
Figure 2 Anatomy of left ventricle and atrium. A. Left atrium with endocardial fibroelastosis. B. Short thickened chordae tendinae with direct insertion of posterior mitral valve leaflet into papillary muscle. C. Thickened dysplastic mitral valve with stenosis.
Figure 3 Histology of Conduction system. VVG stain.
Figure 4 Pulmonary arteriograms. A. Age-matched normal child. B. Kabuki syndrome patient.
Figure 5 Histology of the lung. Movat pentachrome stain.
Table 1 Comparison of immunophenotypes.
Study Patient Normal for Age
Phenotype Analysis
Lymphocytes/mm3 1818 6000 ± 1500
CD2, % 70 73 ± 8
CD3, % 62 66 ± 13
CD4, % 42 43 ± 12
CD8, %, 19 25 ± 9
CD45RA+CD3, % 79 64 – 93
CD45RO+CD3, % 11
CD25+CD4, % 14 <3
CD20, % 14 8 ± 3
smIgM, % 11 4 – 16
smIgM, % 10 3 – 15
CD56, % 10 13 ± 7
Lymphoproliferative Responses
PHA, cpm 164,082 100,530 – 657,376
%NR 65 >50
Con A, cpm 10 53,173 – 502,758
%NR 0 >50
PWM, cpm 121,168 40,305 – 337,597
%NR 96 >50
MLC, cpm 118,219 43,801 – 328,175
SI 23.2 >3.0
Immunoglobulins
IgG, mg/dl 113 399–1068
IgA, mg/dl 11 15–95
IgM, mg/dl 97 49–202
IgE, IU/ml <2 3–29
anti-HiB, μg/ml <0.5 >1.0
anti-Diphtheria toxoid, IU/ml 1.61 >0.05
anti-Tetanus toxoid, IU/ml 0.2 >0.5
anti-Streptococcus, μg/ml
Serotype 4 4.6 >2.0
Serotype 6 2.2 >2.0
Serotype 9 0.3 >2.0
Serotype 14 1.0 >2.0
Serotype 18 12.2 >2.0
Serotype 19 0.9 >2.0
Serotype 23 0.3 >2.0
CH50, U/ml 64 31 – 64
Immunological studies in our 10 month old child. PHA, phytohemagglutinin; Con A, concanavalin A; PWM, pokeweed mitogen; MLC, mixed lymphocyte culture to B-cell alloantigens; %NR, percent normal response; SI, stimulation index; HiB, Hemophilus influenzae type B.
==== Refs
Niikawa N Matsuura N Fukushima Y Ohsawa T Kajii T Kabuki make-up syndrome: a syndrome of mental retardation, unusual facies, large and protruding ears, and postnatal growth deficiency J Pediatr 1981 99 565 569 7277096
Wessels MW Brooks AS Hoogeboom J Niermeijer MF Willems PJ Kabuki syndrome: a review study of three hundred patients Clin Dysmorphol 2002 11 95 102 12002156 10.1097/00019605-200204000-00004
Lynch SA Ashcroft KA Zwolinski S Clarke C Burn J Kabuki syndrome-like features in monozygotic twin boys with a pseudodicentric chromosome 13 J Med Genet 1995 32 227 230 7783176
Halal F Gledhill R Dudkiewicz A Autosomal dominant inheritance of the Kabuki make-up (Niikawa-Kuroki) syndrome Am J Med Genet 1989 33 376 381 2801772 10.1002/ajmg.1320330317
Kawame H Hannibal MC Hudgins L Pagon RA Phenotypic spectrum and management issues in Kabuki syndrome J Pediatr 1999 134 480 485 10190924
Digilio MC Marino B Toscano A Giannotti A Dallapiccola B Congenital heart defects in Kabuki syndrome Am J Med Genet 2001 100 269 274 11343317 10.1002/ajmg.1265
Donadio A Garavelli L Banchini G Neri G Kabuki syndrome and diaphragmatic defects: a frequent association in non-Asian patients? Am J Med Genet 2000 91 164 165 10748421 10.1002/(SICI)1096-8628(20000313)91:2<164::AID-AJMG19>3.0.CO;2-E
Matsumoto N Niikawa N Kabuki make-up syndrome: a review Am J Med Genet C Semin Med Genet 2003 117 57 65 12561059 10.1002/ajmg.c.10020
Chrzanowska KH Krajewska-Walasek M Kus J Michalkiewicz J Maziarka D Wolski JK Brecevic L Madalinski K Kabuki (Niikawa-Kuroki) syndrome associated with immunodeficiency Clin Genet 1998 53 308 312 9650771
McGinniss MJ Brown DH Burke LW Mascarello JT Jones MC Ring chromosome X in a child with manifestations of Kabuki syndrome Am J Med Genet 1997 70 37 42 9129739 10.1002/(SICI)1096-8628(19970502)70:1<37::AID-AJMG8>3.0.CO;2-O
Prasad C Chudley AE Genetics and cardiac anomalies: the heart of the matter Indian J Pediatr 2002 69 321 332 12019554
Lo IF Cheung LY Ng AY Lam ST Interstitial Dup(1p) with findings of Kabuki make-up syndrome Am J Med Genet 1998 78 55 57 9637424 10.1002/(SICI)1096-8628(19980616)78:1<55::AID-AJMG11>3.0.CO;2-N
Niikawa N Kuroki Y Kajii T Matsuura N Ishikiriyama S Tonoki H Ishikawa N Yamada Y Fujita M Umemoto H Kabuki make-up (Niikawa-Kuroki) syndrome: a study of 62 patients Am J Med Genet 1988 31 565 589 3067577
Watanabe T Miyakawa M Satoh M Abe T Oda Y Kabuki make-up syndrome associated with chronic idiopathic thrombocytopenic purpura Acta Paediatr Jpn 1994 36 727 729 7871993
Hammarstrom L SCIE Hans D. Ochs CIESJMP Genetic approach to common variable immunodeficiency and IgA deficiency. Primary Immunodeficiency Diseases: A Molecular and Genetic Approach 1999 New York, Oxford University Press 250 262
Ramesh N GRSNLD Hans D. Ochs CIESJMP CD40 ligand and the hyper-IgM syndrome Primary Immunodeficiency Diseases: A Molecular and Genetic Approach 1999 New York, Oxford University Press 233 249
|
16042804
|
PMC1190177
|
CC BY
|
2021-01-04 16:03:33
|
no
|
BMC Med Genet. 2005 Jul 25; 6:28
|
utf-8
|
BMC Med Genet
| 2,005 |
10.1186/1471-2350-6-28
|
oa_comm
|
==== Front
BMC Med GenetBMC Medical Genetics1471-2350BioMed Central London 1471-2350-6-291607899610.1186/1471-2350-6-29Research ArticleAssociation study of functional genetic variants of innate immunity related genes in celiac disease Rueda B [email protected] A [email protected]ópez-Nevot MA [email protected]ín J [email protected] BPC [email protected] Instituto de Parasitología y Biomedicina "López-Neyra", CSIC, Granada, Spain2 Complex Genetics Group, Department of Biomedical Genetics, University Medical Center, Utrecht, The Netherlands3 Servicio de Inmunología, Hospital Virgen de las Nieves, Granada, Spain2005 3 8 2005 6 29 29 29 3 2005 3 8 2005 Copyright © 2005 Rueda et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Recent evidence suggest that the innate immune system is implicated in the early events of celiac disease (CD) pathogenesis. In this work for the first time we have assessed the relevance of different proinflammatory mediators typically related to innate immunity in CD predisposition.
Methods
We performed a familial study in which 105 celiac families characterized by the presence of an affected child with CD were genotyped for functional polymorphisms located at regulatory regions of IL-1α, IL-1β, IL-1RN, IL-18, RANTES and MCP-1 genes. Familial data was analysed with a transmission disequilibrium test (TDT) that revealed no statistically significant differences in the transmission pattern of the different genetic markers considered.
Results
The TDT analysis for IL-1α, IL-1β, IL-1RN, IL-18, and MCP-1 genes genetic variants did not reveal biased transmission to the affected offspring. Only a borderline association of RANTES promoter genetic variants with CD predisposition was observed.
Conclusion
Our results suggest that the analysed polymorphisms of IL-1α, IL-1β, IL-1RN, IL-18, RANTES and MCP-1 genes do not seem to play a major role in CD genetic predisposition in our population.
==== Body
Background
Celiac disease (CD) is an autoimmune disorder of the small intestine in which dietary gluten ingestion leads to a chronic inflammatory status of the mucosa [1]. There is strong evidence for a genetic component for CD, with the HLA genes being the strongest genetic locus associated with CD predisposition known to date. About 95% of CD patients are carriers of the DQ2 molecule, encoded by DQA1*05/DQB1*02 alleles, compared to ~10% of healthy control subjects. Furthermore, the DQ8 molecule (DQA1*0301/DQB1*0302 is also found more frequently in CD patients although to a lesser extent [2]. Finally, a role for genes located outside the HLA region has been suggested since the overall contribution of HLA genes to CD genetic predisposition is no more than 40% [1].
T CD4+ lymphocytes are key elements in the induction and progression of CD pathogenesis. Certain gluten peptides bound to DQ2 or DQ8 molecules cause proliferation and production of proinflammatory cytokines by lamina propria CD4 +T cells [3]. Besides this activation of adaptive immune response, recent evidences suggest that there is an implication of the innate immunity in the initial phases of CD [4]. In this regard, some gluten peptides have been demonstrated to drive a danger signal that leads to an activation of the innate immune system [5,6] and additionally it is thought that bacteria may play a role in CD [7]. In fact, CD patients show an up-regulation in the expression of pro-inflammatory cytokines typically related to the innate immune response, such us IL1, IL-18 and chemokines [6,8-10].
The IL1 gene cluster located in the chromosomal region 2q12-22 codifies for three proteins: IL-1α, IL-1β and IL-1 receptor agonist (IL-1RN), of which the two first are strong inducers of inflammation while IL-1RN is an effective antagonist binding to the IL-1 receptor without activating the target cell [11]. These genes are polymorphic bearing well-characterized single nucleotide polymorphisms (SNPs). Polymorphisms in IL-1α at position -889 C/T (rs1800587) and IL-1β at position -511 C/T (rs1143627) were described [12,13]. Furthermore, recent findings showed that the -511 C/T IL-1β genetic variant is related to differences in IL-1β protein secretion [14]. The IL-1RN gene contains within its second intron a variable number of an 86-bp tandem repeats (rs380092) [15], showing the allele 2 (IL-1RN*2; two repeats) an increased frequency in a variety of autoimmune and inflammatory disorders [16].
Another important member of the proinflammatory IL-1 family is IL-18, which is thought to be a key regulator of cytokine expression [17]. Furthermore, a role for IL-18 in the induction of an anti-gluten inflammatory response has been suggested [10,18,19]. It is thought that IL-18 gene variation in the promoter region regulates the expression of this cytokine [20]. Interestingly, in the IL-18 promoter region two SNPs -607 A/C (rs1946518) and -137 G/C (rs187238) were described, which are supposed to alter the IL-18 promoter activity [21].
Moreover, raised levels of chemokines such us RANTES (regulated upon activation, normal T-cells expressed and secreted) and monocyte chemoatractant protein-1 (MCP-1) have been observed in the primary immune response to gluten in CD patients [6,8]. Interestingly, genetic variants within regulatory regions that can affect trancription and protein production levels, RANTES -403 G/A (rs2107538) and -28 G/C (rs2280788) and MCP-1 -2518 G/A (rs1024611) SNPs, were described [22-24].
Taking into consideration these findings, in this work we aimed to investigate the possible implication of IL-1α, IL-1β, IL-1RN, IL-18, RANTES and MCP-1 functional polymorphisms in CD susceptibility.
Methods
Patients
In the present work we have analysed a panel of 105 celiac families characterised by the presence of an affected child with CD. The study participants were recruited at "Hospital Materno-Infantil" and "Hospital Clinico Universitario", Granda, (Spain) and were of Spanish Caucasian origin. All patients were diagnosed following the European Society of Paediatric Gastroenterology, Hepatology and Nutrition (ESPGHAN) criteria for CD [25]. Their age at study was 7.1 ± 3.9 years and the mean age for disease diagnosis was 2.7 ± 2.72. A 60% were women and 40 % men, showing an anthropometry at diagnosis (weight and height) of P 3–100 percentile. The mean age of gluten introduction was 6.4 ± 1.5 months. Typical symptoms were observed in 72.2 % of patients and 27.8 % showed atypical symptoms. All family members were genotyped in DRB1 and DQB1. DQA1 typing was deduced from DQB1 and DRB1 typing on the basis of the strong linkage disequilibrium among HLA class II alleles.
Genotyping
DNA from patients and controls was obtained from peripheral blood using standard methods. For all of the considered SNPs, except IL-1RN and RANTES -403, samples were genotyped using a Taqman 5' allelic discrimination assay. Table 1 shows the Taqman MGB probes sequences used for each polymorphism provided by the Custom-Taqman-SNP-Genotyping-Assay (Applied Biosystems, Foster City, CA, USA). PCR reaction was carried in a total reaction volume of 5 μl with the following amplification protocol: denaturation at 92°C for 10 min, followed by 50 cycles of denaturation at 92°C for 15 sec and annealing and extension at 58°C for 1 min. Post-PCR, the genotype of each sample was attributed automatically by measuring the allelic specific fluorescence on the ABI PRIM 7900 Sequence Detection Systems using the SDS 2.2.1 software for allelic discrimination (Applied Biosystems, Foster City, CA, USA). RANTES -403 genotyping was performed using a TaqMan SNP-Genotyping-Assay (part number: C__15874407_10, Applied Biosystems, Foster City, CA, USA).
The IL-1RN polymorphism was genotyped by PCR as previously described [26]. Briefly, we used a froward primer 5'- CTC AGC AAC ACCT CCT AT and reverse primer 5'- TCC TGG TCT GCA GGT AA, two amplify five possible alleles with different PCR fragment size: 410 bp (allele 1: 4 repeats), 240 bp (allele 2: two repeats), 325 bp (allele 3: 3 repeats), 500 bp (allele 4: 5 repeats), and 595 bp (allele 5: 6 repeats).
Statistical analysis
We used the UNPHASED software created for TDT and case-control analysis [27]. We performed a Transmission Disequilibrium Test (TDT), which assesses allele transmission rates in simplex families and tests for deviation from expected 50% transmission. For the haplotype analysis, pair-wise linkage disequilibrium measures were investigated and haplotypes constructed using the expectation-maximization (EM) algorithm implemented in UNPHASED software. The power of the study to detect an effect of a polymorphism in disease susceptibility was estimated using the Quanto v 0.5 software (Department of Preventive Medicine University of Southern California, California, USA) [28].
Results
IL1 gene cluster
The transmission pattern for IL-1α -889, IL-1β -511 and IL-1RN VNTR polymorphisms is shown in table 2. When transmission of these genetic variants was analysed, none of the alleles showed statistically significant skewing. IL-1α -889 T allele was slightly more transmitted to the affected children (58% transmission for allele T vs 47% for allele C), however the p value failed to reach statistically significant level (Table 2). With regard to IL-1RN we observed that alleles IL-1RN*1 and IL-1RN*2 were the most frequent in our population (71.6% and 26.8% respectively), accordingly with previously studies in Caucasian populations [26].
IL18 gene
The TDT analysis for -607 A/C and -137 G/C IL-18 promoter genetic variants did not reveal biased transmission of any of the alleles to the affected offspring (Table 2).
The haplotype estimation for the -607 A/C and -137 IL-18 promoter variants revealed complete linkage disequilibrium between the two variants (D' = 1). We observed three out of the four possible haplotypic combinations in CD families (Table 3). The transmission pattern of IL-18 promoter haplotypes did not show any statistically significant skewing (Table 3).
MCP-1 and RANTES
After analyzing the MCP-1 -2518 G/A alleles transmission we observed that none of the alleles was preferentially transmitted from heterozygous parents to the affected offspring (Table 2). Regarding to the RANTES promoter genetic variants the mutant alleles -403 A and -28 G showed an overall allele frequency similar to that expected for Caucasian populations (84.1% and 96.5% respectively in our population) [29,30]. The transmission of both -403G/A and -28 C/G SNPs showed a slightly deviation from the 50% expected transmission pattern (Table 2). Alleles -403 G and -28 C were more transmitted to the affected offspring with borderline significance (P = 0.04 and P = 0.06 respectively) (Table 2). In addition, we estimated haplotypes for both genetic variants. Three out of the four haplotypic combinations were observed, being the -403G/-28C and -403A/-28C haplotypes the most common in CD families. No significant distorted transmission pattern for RANTES promoter haplotypes was observed (Table 3).
Discussion
CD is considered a model for autoimmune disorders since many of the components that generate the altered immune response to gluten have been well characterized [1]. However, there are some relevant events of CD pathogenesis that remain unclear, for instance the stimuli that drives the high IFNγ levels in the small intestine of CD patients and why only one out of 20–30 DQ2-positive individuals develops CD [3]. An explanation for these questions might be provided from recent studies that point out a role for the innate immunity in CD [4]. This finding supports a novel focus of research in CD molecular and genetic basis, opening a new field for the functional search of CD candidate genes.
In this work, for the first time we have assessed the relevance of IL-1α, IL-1β, IL-1RN, IL-18, RANTES and MCP-1 genes in CD predisposition. All these genes have been previously associated with susceptibility to several autoimmune disorders [31-40]. However, we failed to detect an association of IL-1α, IL-1β, IL-1RN, IL-18, and MCP-1 genes with CD predisposition using a TDT analysis in our cohort of 105 simplex CD families. Only a borderline significant association of RANTES promoter genetic variants with CD predisposition was observed.
Several studies have focused on the role of RANTES -403G/A and -28 G/C promoter polymorphisms in susceptibility to different autoimmune disorders. The RANTES -403A allele has been associated with susceptibility to multiple sclerosis (MS) and polymialgia rheumatica [41,42]. On the other hand, the RANTES -28G allele was observed to be a genetic risk for clinical complications such us diabetic neprhopathy, early onset of MS, lower levels of C3 in SLE, and higher incidence of central nervous system lupus [37,38,41]. Both RANTES -403A and -28G alleles were associated with higher RANTES expression levels [22,23]. However, considering the multiple testing of the 6 different genes of our study, the association observed for RANTES promoter variants in our population can not be considered as being significant. Therefore, our results of RANTES suggest that further studies should be performed to clarify the role of RANTES in CD and autoimmune diseases in general.
Using a familial approach we eliminate the risk of population stratification derived from case-control association studies. In addition, we estimated that our study design would have considerable power to detect the effect of a polymorphism with moderate to high risk for CD. Assuming an additive model, a minor allele frequency of 0.30 (corresponding to a median value of the majority of markers considered) and RR of 1.8 we would reach 81% power to detect an association in our population. Nevertheless, under a dominant model the power drops to 49% and considering a lower disease allele frequency, for instance 0.16 as is the case of RANTES -28, our study power would decrease to a 64% for a RR of 1.8, and increases to 82% when we assume a RR of 2.0. For this reason, the low level of significance that our TDT analysis reached for RANTES promoter genetic variants might well reflect a true positive, and therefore needs further confirmation using a larger group of CD families.
Taking into account our findings, it is suggested that the analysed genetic polymorphisms of IL-1α, IL1-β, IL-1RN, IL-18, RANTES and MCP-1 genes seem not to play a major role in CD susceptibility in our population. It might be possible that the release of these cytokines and chemokines observed in CD patients could be derived from the activity of other innate immunity related pro-inflammatory mediators with higher influence in disease pathogenesis. In this regard, it is known that in CD the cytokine expression pattern in response to gluten is strongly dominated by IFNγ [43]. Of note, in a recent work we assessed the influence of a functional dinucleotide polymorphism of IFNγ gene in CD predisposition. An association of a higher IFNγ producer allele with CD was observed, supporting a possible explanation for the high levels of INFγ observed in intestinal mucosa of CD patients [44].
Other proinflammatory mediators related with innate immunity such as, TNF-α and IL-12, has been analysed with respect to CD susceptibility. In accordance with our findings no evidence of association was found between IL-12 and CD in two independent studies [45,46]. Regarding TNF-α it has been difficult to dissect the relevance of this genetic marker in CD since it maps within HLA clas III region and it shows linkage disequilibrium with CD disease predisposing DQ2 alleles. In fact controversial results have been obtained, and there is no consensus about an independent or due to linkage disequilibrium role of TNF-α in CD susceptibility [47,48].
Conclusion
Our results suggest that IL-1α, IL1-β, IL-1RN, IL-18, RANTES and MCP-1 genetic variants do not play a major role in CD genetic predisposition, although the suggestive evidence for RANTES deserves further investigation. Furthermore, we consider the innate response an intriguing focus of research and it should be of interest to investigate the role of other cytokines up-regulated in the early events of CD.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
B.R., carried out the genotyping and statistical analysis and drafted the manuscript.
A.Z., participated in the genotyping and helped in the use of the ABI PRIM 7900 Sequence Detection Systems and SDS 2.2.1 software.
M.A. L-N., collected the samples and revised the manuscript.
J.M., participated in the manuscript design and coordination and helped to draft the manuscript
B.K., reviewed the statistical analysis and helped to draft the manuscript.
All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was supported by grant SAF03-460 from Plan Nacional de I+D (CICYT), and in part by Consejería de Educación, Junta de Andalucía, grupo CTS-180.
B.K. is supported by the Dutch Diabetes Research Foundation, The Netherlands Organisation for Health Research and Development (ZonMW) and The Juvenile Diabetes Research Foundation International (JDRF) (2001.10.004).
Figures and Tables
Table 1 Taqman probes used for cytokine genotyping
Polymorphism Taqman probe sequence
IL-1A -889 C/T (rs1800587) VIC – CCTTCAATGGTGTTGCC
FAM – CCTTCAATGATGTTGCC
IL-1B -551 C/T (rs1143627) VIC – CTGTTTTTATGGCTTTCA
FAM – CTGTTTTTATAGCTTTCA
IL-18 -607 C/A (rs1946518) VIC – ATCATTAGAATTTTATGTAATAAT
FAM – ATCATTAGAATTTTATTTAATAAT
IL-18 -137 G/C (rs187238) VIC – ACTATTTTCATGAAATCTTTTCT
FAM – TTTTCATGAAATGTTTTCT
RANTES -28 G/C (rs2280788) VIC – CCCCTCAACTGGC
FAM – CCCCTGAACTGGC
MCP-1 -2518 G/A (rs1024611) VIC – CAGACAGCTGTCACTTT
FAM – CAGACAGCTATCACTTT
Table 2 Allelic frequencies and percentage of transmission of IL-1α, IL-1β, IL-1RN, IL-18, RANTES and MCP-1 genetic variants in CD families.
Alelle frequency in parents (%)
T:NT
% T
P
IL-1A -551
C 70.4 111:125 47 0.08
T 29.6 52:38 58 0.08
IL-1B -889
T 67.6 110:111 49.7 NS
C 32.4 47:46 50.5 NS
IL-1RN
VNTR
1 71.6 107:107 50 NS
2 26.8 43:42 50.6 NS
3 1.6 2:2 50 NS
IL-18 -607
C 61.7 102:103 49.7 NS
A 38.3 59:58 50.4 NS
IL-18 -137
G 75.9 124:124 50 NS
C 24.1 47:47 50 NS
RANTES -28
C 84.1 155:148 51 0.04
G 15.9 3:10 23 0.04
RANTES -403
G 85.7 141:128 52 0.06
A 14.3 23:36 39 0.06
MCP-1
A 72.8 117:129 47 NS
G 27.2 48:36 57 NS
T = transmitted, NT = not transmitted %T = percentage transmitted
Table 3 Transmission pattern of haplotypes inferred for IL-18 and RANTES promoter genetic variant in CD families
Gene Haplotype Transmitted (%) No transmitted (%)
P
Il-18 -607A/-137G 37 (24) 35 (22.5) NS
-607A/-137C 19 (12.2) 20 (13) NS
-607C/-137G 99 (63.8) 100 (64.5) NS
RANTES -403G/-28A 3 (1.4) 8 (5.2) NS
-403A/-28C 19 (12.3) 26 (16.8) NS
-403G/-28C 132 (85.7) 119 (81.5) 0.05
==== Refs
Sollid LM Coeliac disease: dissecting a complex inflammatory disorder Nature Review Immunology 2002 2 647 10.1038/nri885
Louka AS Sollid LM HLA in coeliac disease: Unravelling the complex genetics of a complex disorder Tissue Antigens 2003 61 105 117 12694579
Green PH Jabri B Coeliac disease Lancet 2003 362 383 391 12907013 10.1016/S0140-6736(03)14027-5
Schuppan D Esslinger B Dieterich W Innate immunity and coeliac disease Lancet 2003 362 3 4 12853189 10.1016/S0140-6736(03)13843-3
Maiuri L Ciacci C Ricciardelli I Vacca L Raia V Auricchio S Picard J Osman M Quaratino S Londei M Association between innate response to gliadin and activation of pathogenic T cells in coeliac disease Lancet 2003 362 30 37 12853196 10.1016/S0140-6736(03)13803-2
Tuckova L Novotna J Novak P Flegelova Z Kveton T Jelinkova L Zidek Z Man P Tlaskalova-Hogenova H Activation of macrophages by gliadin fragments: isolation and characterization of active peptide J Leukoc Biol 2002 71 625 631 11927649
Forsberg G Fahlgren A Horstedt P Hammarstrom S Hernell O Hammarstrom ML Presence of bacteria and innate immunity of intestinal epithelium in childhood celiac disease Am J Gastroenterol 2004 99 894 904 15128357 10.1111/j.1572-0241.2004.04157.x
Chowers Y Marsh MN De Grandpre L Nyberg A Theofilopoulos AN Kagnoff MF Increased proinflammatory cytokine gene expression in the colonic mucosa of coeliac disease patients in the early period after gluten challenge Clin Exp Immunol 1997 107 141 147 9010269 10.1046/j.1365-2249.1997.d01-892.x
Fornari MC Pedreira S Niveloni S Gonzalez D Diez RA Vazquez H Mazure R Sugai E Smecuol E Boerr L Pre- and post-treatment serum levels of cytokines IL-1beta, IL-6, and IL-1 receptor antagonist in celiac disease. Are they related to the associated osteopenia? Am J Gastroenterol 1998 93 413 418 9580142
Salvati VM MacDonald TT Bajaj-Elliott M Borrelli M Staiano A Auricchio S Troncone R Monteleone G Interleukin 18 and associated markers of T helper cell type 1 activity in coeliac disease Gut 2002 50 186 190 11788557 10.1136/gut.50.2.186
Arend WP The balance between IL-1 and IL-1Ra in disease Cytokine Growth Factor Rev 2002 13 323 340 12220547 10.1016/S1359-6101(02)00020-5
di Giovine FS Takhsh E Blakemore AI Duff GW Single base polymorphism at -511 in the human interleukin-1 beta gene (IL1 beta) Hum Mol Genet 1992 1 450 1301918
McDowell TL Symons JA Ploski R Forre O Duff GW A genetic association between juvenile rheumatoid arthritis and a novel interleukin-1 alpha polymorphism Arthritis Rheum 1995 38 221 228 7848312
Hall SK Perregaux DG Gabel CA Woodworth T Durham LK Huizinga TW Breedveld FC Seymour AB Correlation of polymorphic variation in the promoter region of the interleukin-1 beta gene with secretion of interleukin-1 beta protein Arthritis Rheum 2004 50 1976 1983 15188375 10.1002/art.20310
Tarlow JK Blakemore AI Lennard A Solari R Hughes HN Steinkasserer A Duff GW Polymorphism in human IL-1 receptor antagonist gene intron 2 is caused by variable numbers of an 86-bp tandem repeat Hum Genet 1993 91 403 404 8500797 10.1007/BF00217368
Witkin SS Gerber S Ledger WJ Influence of interleukin-1 receptor antagonist gene polymorphism on disease Clin Infect Dis 2002 34 204 209 11740709 10.1086/338261
Gracie JA Robertson SE McInnes IB Interleukin-18 J Leukoc Biol 2003 73 213 224 12554798 10.1189/jlb.0602313
Eaton AD Xu D Garside P Administration of exogenous interleukin-18 and interleukin-12 prevents the induction of oral tolerance Immunology 2003 108 196 203 12562328 10.1046/j.1365-2567.2003.01570.x
Lettesjo H Hansson T Bergqvist A Gronlund J Dannaeus A Enhanced interleukin-18 levels in the peripheral blood of children with coeliac disease Clin Exp Immunol 2005 139 138 143 15606624 10.1111/j.1365-2249.2005.02661.x
Tone M Thompson SA Tone Y Fairchild PJ Waldmann H Regulation of IL-18 (IFN-gamma-inducing factor) gene expression J Immunol 1997 159 6156 6163 9550417
Giedraitis V He B Huang WX Hillert J Cloning and mutation analysis of the human IL-18 promoter: a possible role of polymorphisms in expression regulation J Neuroimmunol 2001 112 146 152 11108943 10.1016/S0165-5728(00)00407-0
Liu H Chao D Nakayama EE Taguchi H Goto M Xin X Takamatsu JK Saito H Ishikawa Y Akaza T Polymorphism in RANTES chemokine promoter affects HIV-1 disease progression Proc Natl Acad Sci U S A 1999 96 4581 4585 10200305 10.1073/pnas.96.8.4581
McDermott DH Beecroft MJ Kleeberger CA Al-Sharif FM Ollier WE Zimmerman PA Boatin BA Leitman SF Detels R Hajeer AH Chemokine RANTES promoter polymorphism affects risk of both HIV infection and disease progression in the Multicenter AIDS Cohort Study Aids 2000 14 2671 2678 11125885 10.1097/00002030-200012010-00006
Rovin BH Lu L Saxena R A novel polymorphism in the MCP-1 gene regulatory region that influences MCP-1 expression Biochem Biophys Res Commun 1999 259 344 348 10362511 10.1006/bbrc.1999.0796
Revised criteria for diagnosis of coeliac disease. Report of Working Group of European Society of Paediatric Gastroenterology and Nutrition Arch Dis Child 1990 65 909 11 2205160
Craggs A West S Curtis A Welfare M Hudson M Donaldson P Mansfield J Absence of a genetic association between IL-1RN and IL-1B gene polymorphisms in ulcerative colitis and Crohn disease in multiple populations from northeast England Scand J Gastroenterol 2001 36 1173 1178 11686217 10.1080/00365520152584806
Dudbridge F Methods and software for association tests of uncertain haplotypes in case-parent trios Am J Hum Genet 2002 71 A2338
Gauderman WJ Candidate gene association analysis for a quantitative trait, using parent-offspring trios Genet Epidemiol 2003 25 327 338 14639702 10.1002/gepi.10262
al Sharif F Ollier WE Hajeer AH A rare polymorphism at position -28 in the human RANTES promoter Eur J Immunogenet 1999 26 373 374 10553504 10.1046/j.1365-2370.1999.00162.x
Hajeer AH al Sharif F Ollier WE A polymorphism at position -403 in the human RANTES promoter Eur J Immunogenet 1999 26 375 376 10553505 10.1046/j.1365-2370.1999.00163.x
Parks CG Cooper GS Dooley MA Treadwell EL St Clair EW Gilkeson GS Pandey JP Systemic lupus erythematosus and genetic variation in the interleukin 1 gene cluster: a population based study in the southeastern United States Ann Rheum Dis 2004 63 91 94 14672899 10.1136/ard.2003.007336
Timms AE Crane AM Sims AM Cordell HJ Bradbury LA Abbott A Coyne MR Beynon O Herzberg I Duff GW The interleukin 1 gene cluster contains a major susceptibility locus for ankylosing spondylitis Am J Hum Genet 2004 75 587 595 15309690 10.1086/424695
Mann CL Davies MB Stevenson VL Leary SM Boggild MD Ko Ko C Jones PW Fryer AA Strange RC Thompson AJ Interleukin 1 genotypes in multiple sclerosis and relationship to disease severity J Neuroimmunol 2002 129 197 204 12161036 10.1016/S0165-5728(02)00181-9
Kaijzel EL van Dongen H Bakker AM Breedveld FC Huizinga TW Verweij CL Relationship of polymorphisms of the Interleukin-1 gene cluster to occurrence and severity of rheumatoid arthritis Tissue Antigens 2002 59 122 126 12028539 10.1034/j.1399-0039.2002.590208.x
Ide A Kawasaki E Abiru N Sun F Fukushima T Ishii R Takahashi R Kuwahara H Fujita N Kita A Association of interleukin-18 gene promoter polymorphisms in type 1 diabetes and autoimmune thyroid disease Ann N Y Acad Sci 2003 1005 436 439 14679107 10.1196/annals.1288.074
Novota P Kolostova K Pinterova D Novak J Treslova L Andel M Cerna M Interleukin IL-18 gene promoter polymorphisms in adult patients with type 1 diabetes mellitus and latent autoimmune diabetes in adults Immunol Lett 2005 96 247 251 15585330 10.1016/j.imlet.2004.08.016
Nakajima K Tanaka Y Nomiyama T Ogihara T Ikeda F Kanno R Iwashita N Sakai K Watada H Onuma T RANTES promoter genotype is associated with diabetic nephropathy in type 2 diabetic subjects Diabetes Care 2003 26 892 898 12610055
Liao CH Yao TC Chung HT See LC Kuo ML Huang JL Polymorphisms in the promoter region of RANTES and the regulatory region of monocyte chemoattractant protein-1 among Chinese children with systemic lupus erythematosus J Rheumatol 2004 31 2062 2067 15468376
Tucci M Barnes EV Sobel ES Croker BP Segal MS Reeves WH Richards HB Strong association of a functional polymorphism in the monocyte chemoattractant protein 1 promoter gene with lupus nephritis Arthritis Rheum 2004 50 1842 1849 15188361 10.1002/art.20266
Kim HL Lee DS Yang SH Lim CS Chung JH Kim S Lee JS Kim YS The polymorphism of monocyte chemoattractant protein-1 is associated with the renal disease of SLE Am J Kidney Dis 2002 40 1146 1152 12460032 10.1053/ajkd.2002.36858
Gade-Andavolu R Comings DE MacMurray J Vuthoori RK Tourtellotte WW Nagra RM Cone LA RANTES: a genetic risk marker for multiple sclerosis Mult Scler 2004 10 536 539 15471370 10.1191/1352458504ms1080oa
Makki RF al Sharif F Gonzalez-Gay MA Garcia-Porrua C Ollier WE Hajeer AH RANTES gene polymorphism in polymyalgia rheumatica, giant cell arteritis and rheumatoid arthritis Clin Exp Rheumatol 2000 18 391 393 10895380
Nilsen EM Jahnsen FL Lundin KE Johansen FE Fausa O Sollid LM Jahnsen J Scott H Brandtzaeg P Gluten induces an intestinal cytokine response strongly dominated by interferon gamma in patients with celiac disease Gastroenterology 1998 115 551 563 9721152
Rueda B Martinez A Lopez-Nevot MA Mas-Fontao A Paco L Ortega E Fernandez-Arquero M Urcelay E Gomez de la Concha E Martin J A functional variant of IFNgamma gene is associated with coeliac disease Genes Immun 2004 5 517 519 15215891 10.1038/sj.gene.6364115
Louka AS Torinsson Naluai A D'Alfonso S Ascher H Coto I Ek J Giordano M Gudjonsdottir AH Mellai M Momigliano-Richiardi P The IL12B gene does not confer susceptibility to coeliac disease Tissue Antigens 2002 59 70 72 11972887 10.1034/j.1399-0039.2002.590117.x
Seegers D Borm ME van Belzen MJ Mulder CJ Bailing J Crusius JB Meijer JW Wijmenga C Pena AS Bouma G IL12B and IRF1 gene polymorphisms and susceptibility to celiac disease Eur J Immunogenet 2003 30 421 425 14675396 10.1111/j.1365-2370.2003.00428.x
de la Concha EG Fernandez-Arquero M Vigil P Rubio A Maluenda C Polanco I Fernandez C Figueredo MA Celiac disease and TNF promoter polymorphisms Hum Immunol 2000 61 513 517 10773355 10.1016/S0198-8859(99)00187-1
Woolley N Mustalahti K Maki M Partanen J Cytokine gene polymorphisms and genetic association with coeliac disease in the Finnish population Scand J Immunol 2005 61 51 56 15644122 10.1111/j.0300-9475.2005.01525.x
|
16078996
|
PMC1190178
|
CC BY
|
2021-01-04 16:03:33
|
no
|
BMC Med Genet. 2005 Aug 3; 6:29
|
utf-8
|
BMC Med Genet
| 2,005 |
10.1186/1471-2350-6-29
|
oa_comm
|
==== Front
BMC Med GenetBMC Medical Genetics1471-2350BioMed Central London 1471-2350-6-301608683610.1186/1471-2350-6-30Research ArticleHigh frequency of the IVS2-2A>G DNA sequence variation in SLC26A5, encoding the cochlear motor protein prestin, precludes its involvement in hereditary hearing loss Tang Hsiao-Yuan [email protected] Anping [email protected] John S [email protected] Fred A [email protected] Raye L [email protected] The Bobby R. Alford Department of Otorhinolaryngology and Communicative Sciences, Baylor College of Medicine, One Baylor Plaza, NA102, Houston, TX 77030, USA2 Huffington Center on Aging and Department of Molecular and Cellular Biology, Baylor College of Medicine, One Baylor Plaza, N710, Houston, TX 77030, USA2005 8 8 2005 6 30 30 8 3 2005 8 8 2005 Copyright © 2005 Tang et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Cochlear outer hair cells change their length in response to variations in membrane potential. This capability, called electromotility, is believed to enable the sensitivity and frequency selectivity of the mammalian cochlea. Prestin is a transmembrane protein required for electromotility. Homozygous prestin knockout mice are profoundly hearing impaired. In humans, a single nucleotide change in SLC26A5, encoding prestin, has been reported in association with hearing loss. This DNA sequence variation, IVS2-2A>G, occurs in the exon 3 splice acceptor site and is expected to abolish splicing of exon 3.
Methods
To further explore the relationship between hearing loss and the IVS2-2A>G transition, and assess allele frequency, genomic DNA from hearing impaired and control subjects was analyzed by DNA sequencing. SLC26A5 genomic DNA sequences from human, chimp, rat, mouse, zebrafish and fruit fly were aligned and compared for evolutionary conservation of the exon 3 splice acceptor site. Alternative splice acceptor sites within intron 2 of human SLC26A5 were sought using a splice site prediction program from the Berkeley Drosophila Genome Project.
Results
The IVS2-2A>G variant was found in a heterozygous state in 4 of 74 hearing impaired subjects of Hispanic, Caucasian or uncertain ethnicity and 4 of 150 Hispanic or Caucasian controls (p = 0.45). The IVS2-2A>G variant was not found in 106 subjects of Asian or African American descent. No homozygous subjects were identified (n = 330). Sequence alignment of SLC26A5 orthologs demonstrated that the A nucleotide at position IVS2-2 is invariant among several eukaryotic species. Sequence analysis also revealed five potential alternative splice acceptor sites in intron 2 of human SLC26A5.
Conclusion
These data suggest that the IVS2-2A>G variant may not occur more frequently in hearing impaired subjects than in controls. The identification of five potential alternative splice acceptor sites in intron 2 of human SLC26A5 suggests a potential mechanism by which expression of prestin might be maintained in cells carrying the SLC26A5 IVS2-2A>G DNA sequence variation. Additional studies are needed to evaluate the effect of the IVS2-2A>G transition on splicing of SLC26A5 transcripts and characterize the hearing status of individuals homozygous for the IVS2-2A>G variant.
==== Body
Background
Outer hair cells (OHCs) are sensory cells of the mammalian cochlea. These cells are cylindrical in shape, with stereocilia projecting from their apical surfaces and neuronal synapses associated with their basal surfaces. Mechanical deflection of OHC stereocilia, in response to sound pressure waves, results in variations in the OHC membrane potential that trigger somatic cell length changes in synchrony with the sound wave. Hyperpolarizing potentials elongate the cell and depolarizing potentials shorten the cell. Known as electromotility, this capability is believed to amplify cochlear vibrations and enable the acute hearing sensitivity and frequency selectivity of the mammalian cochlea [1-3].
Prestin is a multipass transmembrane protein of the OHC required for electromotility [2,3]. Developmental expression of prestin coincides with the appearance of electromotility [4]. When prestin is expressed in non-auditory mammalian cells in vitro, a non-linear, voltage-dependent membrane capacitance, a commonly used marker of electromotility, results [3,5]. Targeted deletion of prestin in mice eliminates OHC electromotility and results in a significant (40-60dB) loss of cochlear sensitivity and an increased high frequency hearing threshold [3,6,7]. The human gene SLC26A5, encoding prestin, contains 21 exons that are alternately spliced to create four isoforms of prestin with variable lengths of the C-terminus [8]. The expression profile of the four isoforms in OHCs and their significance with respect to electromotility are not known [8].
In humans, a single nucleotide change in the second intron of SLC26A5 has been reported in association with hearing loss [8]. The DNA sequence variation, IVS2-2A>G, is an A to G transition in the splice acceptor site for exon 3. The A nucleotide at position IVS2-2 of SLC26A5 was found to be conserved in human, mouse, and rat orthologs and mutation of this nucleotide is predicted to cause skipping of exon 3 [8-10]. The start codon for the prestin protein is encoded in exon 3 of SLC26A5 [8,10]. Skipping of exon 3 during RNA processing would be expected to result in a messenger RNA incapable of prestin protein production and cochlear hearing loss due to absence of OHC electromotility [8-10].
To further explore the relationship between the SLC26A5 IVS2-2A>G nucleotide substitution and hearing loss, and assess allele frequency, genomic DNA from hearing impaired and control subjects was genotyped at the SLC26A5 IVS2-2 position. In addition, SLC26A5 orthologs from various eukaryotic species were evaluated for conservation of the A nucleotide at position IVS2-2. An attempt was also made to identify possible alternative splice acceptor sites within intron 2 that might support alternate splicing of exon 3.
Methods
Subjects
Hearing impaired patients were identified and recruited from the outpatient clinical care centers of the Bobby R. Alford Department of Otorhinolaryngology and Communicative Sciences and the Department of Molecular and Human Genetics of Baylor College of Medicine. Ethnicity of cases was self-described. Control specimens were obtained from the Baylor Polymorphism Resource . In this resource, Caucasian subjects were European Americans. Ethnicity of controls was recorded by the Baylor Polymorphism Resource group at the time of specimen collection. The hearing status of controls is unknown however the process through which control subjects were recruited included verbal communication, reducing the likelihood that individuals with significant hearing impairment were unknowingly included in the control population.
IRB approval
This work was approved by the Baylor College of Medicine Institutional Review Board. Informed consent was obtained and documented from all subjects prior to specimen collection.
Specimen collection
Blood was collected from all subjects by peripheral venipuncture. Lymphoblastoid cell lines were established by standard Epstein Barr Virus mediated transformation protocols.
DNA isolation
DNA was isolated from cultured cells using the PUREGENE® DNA Purification Kit for cells (Gentra Systems, Inc., Minneapolis, Minnesota, USA), according to manufacturer's specifications.
Sequencing of the SLC26A5 exon 3 splice acceptor site
The region surrounding the exon 3 splice acceptor site of SLC26A5 was amplified from human genomic DNA by polymerase chain reaction (PCR) using the forward primer Prestin-5'F from [8] and, a reverse primer that was derived from primer Prestin-4R from [8] by the addition of 5 nucleotides to the 5' end as follows, 5'-GCAATTGTTTGAGGACAGCAAGGG-3'. PCR was conducted with 15 pmol of each primer, 1.25U Taq DNA polymerase (Amersham Pharmacia Biotech Inc., Piscataway, NJ, USA), and 1 × PCR buffer, as provided by the manufacturer, in a total volume of 25 μL. PCR was conducted as follows: 94°C for 2 minutes; 40 cycles of 94°C for 30 seconds, 58°C for 30 seconds, and 70°C for 1 minute; and, 70°C for 5 minutes. PCR fragments were sequenced using the forward primer and the ABI BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems Inc, Foster City, California, USA). Cycle sequencing was conducted as follows: 25 cycles of 96°C for 10 seconds; 50°C for 5 seconds; and, 60°C for 4 minutes. Sequencing reactions were analyzed on an ABI Prism 310 Genetic Analyzer according to the manufacturer's specifications (Applied Biosystems, Foster City, CA).
Sequencing of the coding regions of GJB2 (Connexin 26) and GJB6 (Connexin 30)
The coding region of GJB2 was amplified by PCR and sequenced using the following primers: forward PCR primer, 5'-AGGAAGAGATTTAAGCATGCT-3'; reverse PCR primer, 5'-TGGAGTTTCACCTGAGGC-3'; forward sequencing primer, 5'-CTGCAGCTGATCTTCGTG-3'; and, reverse sequencing primer, 5'-GTCGTACATGACATAGAAGACGT-3'. The coding region of GJB6 was amplified from human genomic DNA by PCR using the following primers: forward PCR primer derived from primer Cx30-1 from [11] by the addition of 3 nucleotides to the 5' end and the removal of 2 nucleotides from the 3' end as follows, 5'-GACTCAGGGATAAACCAGCGCA-3'; reverse PCR and sequencing primer Cx30-8 from [11], and reverse sequencing primer Cx30-4 from [11]. PCR and sequencing were conducted as described for the SLC26A5 exon 3 splice acceptor site.
GJB6 (Connexin 30) deletion analysis
The GJB6 gene was evaluated for the 342 kilobase pair deletion shown to be associated with hearing loss as described [11-13].
Mitochondrial 12S rRNA gene sequencing and restriction analysis for the 1555A>G mutation
DNA samples were amplified by PCR using primers derived from MITOMAP and PCR conditions as previously described [14,15]. The presence of the 1555A>G mutation was evaluated in some subjects by DNA sequencing as previously described [15] and in other subjects by restriction digestion using BsmAI (New England BioLabs, Beverly, MA, USA) according to manufacturer's specifications and as described [16].
Sequence analysis
Electropherograms were evaluated by visual inspection and pairwise alignment to reference sequences using the BCM Search Launcher BLAST2 Pairwise Sequence Alignment Tool from the Human Genome Sequencing Center of Baylor College of Medicine , and/or by interpretation using Mutation Surveyor software Version 2.41 (Softgenetics, Inc, State College, Pennsylvania, USA).
Statistical analysis
Fisher's exact test of 2 × 2 contingency tables was used to calculate the two-tailed p values associated with allele frequencies in case and control groups. Fisher's exact test was performed using the GraphPad QuickCalcs Online Calculator for Scientists .
Multiple sequence alignment of the exon 3 splice acceptor site of SLC26A5
SLC26A5 genomic DNA sequences from human, chimp, rat, mouse, zebrafish and fruit fly were obtained from the Ensembl Genome Browser . Prestin sequences at the intron 2 and exon 3 junction were compared using ClustalW multiple sequence alignment tool, Biology WorkBench 3.2 .
Evaluation of potential alternative splice acceptor sites within intron 2 of human SLC26A5
Alternative splice acceptor sites within intron 2 of human SLC26A5 were sought using a splice site prediction program from the Berkeley Drosophila Genome Project . Only sites having a confidence score above 0.99 out of a maximum possible score of 1.00 were considered to be potential alternative splice sites.
Results
Allele frequency of the IVS2-2A>G DNA sequence variation
Eighty-four hearing impaired cases and 246 controls were genotyped for the exon 3 splice acceptor site of SLC26A5, encoding prestin. Four hearing impaired cases were found to be heterozygous for the IVS2-2A>G nucleotide substitution: one Hispanic, two Caucasian, and one of uncertain ethnicity, possibly Caucasian or mixed Caucasian. Four controls were also found to be heterozygous for the IVS2-2A>G nucleotide substitution: one Hispanic and three Caucasian (Table 1). Four of the eight carriers were male: three cases and one control. Four of the eight carriers were female: three controls and one case. The IVS2-2A>G nucleotide substitution was not found in 106 subjects of Asian or African American descent nor was it found in a homozygous state in any of 330 total subjects (Table 1).
Data from the control population suggests an allele frequency for the IVS2-2A>G nucleotide substitution of approximately 0.007 in Hispanics (carrier frequency 1.3%) and 0.02 in European American Caucasians (carrier frequency 4%). The difference in allele frequency of the IVS2-2A>G nucleotide substitution between cases and controls was evaluated by derivation of two-tailed p values using Fisher's exact test of 2 × 2 contingency tables. The difference in allele frequency of the IVS2-2A>G variant between cases and controls among Hispanics and Caucasians in this population was not statistically significant (p = 0.45). If Asian and African American cases and controls are included in the Fisher's exact test analysis, the allele frequency distribution still does not reach statistical significance (p = 0.12, Table 1).
The phenotype and audiometric profile of each of the four hearing impaired subjects who were carriers of the SLC26A5 IVS2-2A>G nucleotide substitution was unique (Table 2). Of note, the hearing impaired sibling of case #4 did not carry the IVS2-2A>G DNA sequence variation. Among the four SLC26A5 IVS2-2A>G carriers, hearing loss associated mutations were not found in the coding regions of GJB2 or GJB6, encoding Connexin 26 and Connexin 30 respectively. The deletion mutation in GJB6 [11-13] and the 1555A>G mutation in the mitochondrial 12S rRNA gene [16-18] were also absent from these hearing impaired SLC26A5 IVS2-2A>G carriers. Neither genetic nor hearing tests were performed on the parents or other relatives of these subjects.
Multiple sequence alignment of the exon 3 splice acceptor site of SLC26A5
SLC26A5 genomic DNA sequences from human, chimp, rat, mouse, zebrafish and fruit fly were aligned and compared for evolutionary conservation of the exon 3 splice acceptor site and, in particular, the AG dinucleotide. The A and G nucleotides at positions -2 and -1, respectively, of the exon 3 splice acceptor site were invariably present and evolutionarily conserved among all eukaryotic prestin orthologs analyzed (Table 3).
Evaluation of alternative splice sites within intron 2 of SLC26A5
To identify possible alternative splice acceptor sites within intron 2 of SLC26A5 that might support alternate splicing of exon 3, the 21,515 base pair genomic DNA sequence of intron 2 was analyzed using a splice site prediction program from the Berkeley Drosophila Genome Project . Five alternative splice acceptor sites demonstrating a confidence score greater than 0.99 out of a maximum of 1.00 were identified within intron 2 of SLC26A5 (Figure 1).
Discussion
These data suggest that heterozygosity for the SLC26A5 IVS2-2A>G DNA sequence variation may not be, by itself, sufficient to cause hearing loss. Only one of the four hearing impaired carriers of the SLC26A5 IVS2-2A>G DNA sequence variation identified in this study reported a history of hearing loss in a parent. Further, various additional clinical findings in each of the hearing impaired carriers of the SLC26A5 IVS2-2A>G variant argue against SLC26A5 being the cause of hearing loss in each of these cases (Table 2).
The carrier frequency for the SLC26A5 IVS2-2A>G DNA sequence variation in the control populations used in this study was observed to be 1.3% in Hispanics and 4% in Caucasians. These data suggest that the SLC26A5 IVS2-2A>G DNA sequence variation may not be uncommon and that it occurs in multiple ethnic groups, in contrast to previous studies [8]. The SLC26A5 IVS2-2A>G DNA sequence variation was not found in Asians or African Americans. This observation may reflect absence of the variant in these ethnic groups or may be due to the small size of the study population.
Although nothing is known about the hearing status of the control group, the allele frequency of the SLC26A5 IVS2-2A>G variant was not significantly different between cases and controls. In addition, the SLC26A5 IVS2-2A>G variant was not found in homozygosity in any hearing impaired case. Given the high allele frequency of the IVS2-2A>G variant in the Caucasian and Hispanic control populations, if this variant were associated with hearing loss, it is reasonable to expect that homozygous hearing impaired cases would have been observed in the study group. This expectation is based on the following observations. First, congenital genetic hearing loss affects, conservatively, approximately 1 in 2,000 live births in the United States [19]. A carrier rate in Caucasians of 4% for the SLC26A5 IVS2-2A>G nucleotide substitution suggests that approximately 1 in 2,500 Caucasians should be homozygous for this mutation. If this DNA sequence variation were related to hearing loss, it should account for a large percentage of cases of early onset genetic hearing loss in Caucasians. Curiously, SLC26A5 has not been mapped by traditional linkage analysis in any of the families with hereditary hearing loss that have so far been studied worldwide (Van Camp G, Smith RJH. Hereditary Hearing Loss Homepage. . Further, the carrier rate among Caucasians for mutations in GJB2, encoding Connexin 26, widely believed to be the most common cause of early onset genetic hearing loss in the United States, is estimated to be approximately 3% [20]. This is 1% less than the carrier frequency of the SLC26A5 IVS2-2A>G DNA sequence variation observed among Caucasians in this study. The common 35delG mutation in GJB2 is, by itself, estimated to have a carrier rate in US Caucasians of approximately 2.5% [20]. In a group of 94 hearing impaired cases that includes the 84 subjects tested in this study for SLC26A5 IVS2-2A>G, five subjects homozygous for the 35delG mutation were found (unpublished observation). In contrast, no subjects homozygous for the IVS2-2A>G DNA sequence variation were observed. These observations challenge whether the SLC26A5 IVS2-2A>G DNA sequence variation is associated with hearing loss.
As shown in Table 3, the AG dinucleotide at the splice acceptor site of exon 3 of SLC26A5 is invariant and evolutionarily conserved amongst prestin orthologs. Analysis of gene structure in multiple species including vertebrates, invertebrates, plants and viruses suggested that the AG dinucleotide of 3' splice acceptor sites cannot tolerate mutation [9]. Thus, the IVS2-2A>G substitution creates a sequence variation in SLC26A5 that is expected to cause skipping of exon 3 [8,9]. However, since a number of potential alternative splice acceptor sites exist within intron 2 of human SLC26A5 (Figure 1), and the translation start codon for prestin is found within exon 3 [8,10], alternate splicing of exon 3 within intron 2 might be compatible with production of a correctly translated prestin protein. Alternatively, utilization of one or more alternate splice sites in a way that minimizes the number of transcripts missing exon 3 during functional maturation of the cochlea might minimally impact the level of prestin in mature OHCs. This hypothesis requires in vivo evaluation of splicing and steady state protein levels in cells carrying SLC26A5 genes with the IVS2-2A>G substitution, and in particular during functional maturation of the cochlea.
Further studies are needed to clarify and refine the IVS2-2A/G allele frequencies in various ethnic groups and examine the role of the IVS2-2A>G nucleotide substitution in hearing loss. These studies should include screening of additional case and control populations for the SLC26A5 IVS2-2A>G DNA sequence variation, and identification and audiometric testing of carriers and homozygous individuals.
Conclusion
The exon 3 splice acceptor sequence of SLC26A5 is evolutionarily conserved. Disruption of the canonical AG dinucleotide at this splice acceptor site by the IVS2-2A>G transition is expected to abolish splice site recognition and result in skipping of exon 3. The IVS2-2A>G transition is found in individuals of Hispanic and European American (Caucasian) ancestry, suggesting this DNA sequence variation is not limited to a single ethnic group. Among Hispanics and Caucasians, there is no statistically significant difference in allele frequency of the IVS2-2A>G nucleotide substitution between hearing impaired cases and controls, challenging whether this DNA sequence variation is associated with hearing loss. The detection of alternative splice acceptor sites within intron 2 suggests a potential mechanism by which expression of prestin might be maintained in cells carrying the SLC26A5 IVS2-2A>G DNA sequence variation. Further studies of the SLC26A5 IVS2-2A>G DNA sequence variation are needed.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
HYT participated in the design of this study, carried out the laboratory analyses, contributed to the analysis and interpretation of the data, and helped draft the manuscript. APX participated in the analysis and interpretation of the molecular genetic data, performed the sequence alignment and alternative splice site analyses, and helped to draft the manuscript. JSO participated in the design of this study, in the analysis and interpretation of the data, and in preparation of the manuscript. FAP participated in the design of the study, the sequence alignment and alternative splice site analyses, the analysis and interpretation of the data, and preparation of the manuscript. RLA participated in the design of the study, was responsible for subject enrollment and control group acquisition, analyzed and interpreted data from the study, and prepared the manuscript. All authors read and approved of the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors wish to thank Laura Molinari, Susan D. Fernbach, John W. Belmont, and the Baylor Polymorphism Resource for the availability of the control specimens used in this work, and Ignacio del Castillo for providing a control for the GJB6 deletion analyses. The authors also wish to thank Dr. Thomas Cooper for his consultation and insights on RNA splicing. This work was supported by the Allbritton-Alford Fund, the Alkek Foundation, the Brown Foundation (HYT, RLA), and NIDCD grants DC006671 (JSO) and DC004585 (FAP).
Figures and Tables
Figure 1 Potential alternate splicing of intron 2 of SLC26A5 containing the IVS2-2A>G variant. Alternative splice acceptor sites in intron 2 of SLC26A5 having a confidence score above 0.99 out of a maximum possible score of 1.00 are shown. Numbering of intron 2 begins at the +1 position. Dotted lines above the gene drawing show potential alternately spliced gene products. The solid line below the gene drawing shows the wild type splice donor and acceptor sites that join exon 2 to exon 3.
Table 1 Genotypes of cases and controls at position SLC26A5 IVS2-2.
Ethnicity Case Genotypes Control Genotypes Fisher's exact test two-tailed p value
A/A A/G A/A A/G
Hispanic 22 1 75 1 p = 0.41 p = 0.45
Caucasian 34 2 71 3 p = 0.66
Uncertain (Caucasian/Mixed Caucasian) 14 1 0 0 Not Done
Asian 3 0 50 0 Not Done
African American 7 0 46 0 Not Done
Total 80 4 242 4 p = 0.12
Table 2 Phenotypes of hearing impaired carriers of SLC26A5 IVS2-2A>G.
Case Type of Hearing Loss; Age at onset Associated Anomalies Family History Ethnicity GJB2, GJB6, 12S rRNA
1 Profound sensorineural on right, severe conductive on left; Congenital Internal auditory canal hypoplasia, hypoplastic vestibulocochlear nerve on right; External auditory canal atresia, malformed ossicles on left None reported Hispanic Heterozygous GJB2 V27I polymorphism, no mutations or other variants found
2 Severe to profound on right, moderate to severe on left, mixed, progressive; Early childhood, sudden, with head trauma and ear infection Bilateral enlarged vestibular aqueducts; Disequilibrium; History of ear infections and possible meningitis None reported Caucasian No mutations or variants found
3 Moderate to severe, bilateral, sloping audiogram, sensorineural, progressive; Childhood to young adulthood None Consistent with autosomal dominant inheritance Caucasian No mutations or variants found
4 Moderate on right, mild to moderate on left, sensorineural, progressive; Childhood None Hearing impaired sibling (progressive, mild on right, mild to moderate on left, sensorineural; childhood onset; renal malformation); hearing impaired grandparent (unknown severity, possibly noise-induced) Uncertain: Caucasian/Mixed Caucasian No mutations or variants found
Table 3 Multiple sequence alignment of the exon 3 splice acceptor site of SLC26A5.
Species Intron 2/EXON 3* Ensembl Gene ID
Human cccctag/TGACACT ENSG00000170615
Chimp cccctag/TGACACT ENSPTRG00000019554
Rat cctgcag/GCTTAGC ENSRNOG00000011616
Mouse cccttag/TGGCCAT ENSMUSG00000029015
Zebrafish ttttcag/CTGTTCG ENSDARG00000022424
Fruit fly ttttcag/CTCCTAA CG5485
*The invariant AG dinucleotide at the intron 2/exon 3 splice acceptor site is shown in bold and underlined.
==== Refs
Brownell WE Nature's Solutions for Listening Volta Review 1997 99 9 28
Liberman MC Gao J He DZ Wu X Jia S Zuo J Prestin is required for electromotility of the outer hair cell and for the cochlear amplifier Nature 2002 419 300 304 12239568 10.1038/nature01059
Zheng J Shen W He DZ Long KB Madison LD Dallos P Prestin is the motor protein of cochlear outer hair cells Nature 2000 405 149 155 10821263 10.1038/35012009
Belyantseva IA Adler HJ Curi R Frolenkov GI Kachar B Expression and localization of prestin and the sugar transporter GLUT-5 during development of electromotility in cochlear outer hair cells J Neurosci 2000 20 RC116 11125015
Santos-Sacchi J Navarrete E Voltage-dependent changes in specific membrane capacitance caused by prestin, the outer hair cell lateral membrane motor Pflugers Arch 2002 444 99 106 11976921 10.1007/s00424-002-0804-2
Cheatham MA Huynh KH Gao J Zuo J Dallos P Cochlear function in Prestin knockout mice J Physiol 2004 560 821 830 15319415 10.1113/jphysiol.2004.069559
Wu X Gao J Guo Y Zuo J Hearing threshold elevation precedes hair-cell loss in prestin knockout mice Brain Res Mol Brain Res 2004 126 30 37 15207913 10.1016/j.molbrainres.2004.03.020
Liu XZ Ouyang XM Xia XJ Zheng J Pandya A Li F Du LL Welch KO Petit C Smith RJ Webb BT Yan D Arnos KS Corey D Dallos P Nance WE Chen ZY Prestin, a cochlear motor protein, is defective in non-syndromic hearing loss Hum Mol Genet 2003 12 1155 1162 12719379 10.1093/hmg/ddg127
Shapiro MB Senapathy P RNA splice junctions of different classes of eukaryotes: sequence statistics and functional implications in gene expression Nucleic Acids Res 1987 15 7155 7174 3658675
Zheng J Long KB Matsuda KB Madison LD Ryan AD Dallos PD Genomic characterization and expression of mouse prestin, the motor protein of outer hair cells Mamm Genome 2003 14 87 96 12584604 10.1007/s00335-002-2227-y
del Castillo I Villamar M Moreno-Pelayo MA Del Castillo FJ Alvarez A Telleria D Menendez I Moreno F A deletion involving the connexin 30 gene in nonsyndromic hearing impairment N Engl J Med 2002 346 243 249 11807148 10.1056/NEJMoa012052
del Castillo I Moreno-Pelayo MA Del Castillo FJ Brownstein Z Marlin S Adina Q Cockburn DJ Pandya A Siemering KR Chamberlin GP Ballana E Wuyts W iel-Guerra AT Alvarez A Villamar M Shohat M Abeliovich D Dahl HH Estivill X Gasparini P Hutchin T Nance WE Sartorato EL Smith RJ Van CG Avraham KB Petit C Moreno F Prevalence and evolutionary origins of the del(GJB6-D13S1830) mutation in the DFNB1 locus in hearing-impaired subjects: a multicenter study Am J Hum Genet 2003 73 1452 1458 14571368 10.1086/380205
Lerer I Sagi M Ben-Neriah Z Wang T Levi H Abeliovich D A deletion mutation in GJB6 cooperating with a GJB2 mutation in trans in non-syndromic deafness: A novel founder mutation in Ashkenazi Jews Hum Mutat 2001 18 460 11668644 10.1002/humu.1222
Brandon MC Lott MT Nguyen KC Spolim S Navathe SB Baldi P Wallace DC MITOMAP: a human mitochondrial genome database--2004 update. Nucleic Acids Research 2005 33 D611 D613 15608272 10.1093/nar/gki079
Tang HY Hutcheson E Neill S Drummond-Borg M Speer M Alford RL Genetic susceptibility to aminoglycoside ototoxicity: how many are at risk? Genet Med 2002 4 336 345 12394346
Casano RA Bykhovskaya Y Johnson DF Hamon M Torricelli F Bigozzi M Fischel-Ghodsian N Hearing loss due to the mitochondrial A1555G mutation in Italian families Am J Med Genet 1998 79 388 391 9779807 10.1002/(SICI)1096-8628(19981012)79:5<388::AID-AJMG11>3.0.CO;2-N
Estivill X Govea N Barcelo E Badenas C Romero E Moral L Scozzri R D'Urbano L Zeviani M Torroni A Familial progressive sensorineural deafness is mainly due to the mtDNA A1555G mutation and is enhanced by treatment of aminoglycosides Am J Hum Genet 1998 62 27 35 9490575 10.1086/301676
Prezant TR Agapian JV Bohlman MC Bu X Oztas S Qiu WQ Arnos KS Cortopassi GA Jaber L Rotter JI . Mitochondrial ribosomal RNA mutation associated with both antibiotic-induced and non-syndromic deafness Nat Genet 1993 4 289 294 7689389 10.1038/ng0793-289
Keats BJ Berlin CI Genomics and hearing impairment Genome Res 1999 9 7 16 9927480
Green GE Scott DA McDonald JM Woodworth GG Sheffield VC Smith RJ Carrier rates in the midwestern United States for GJB2 mutations causing inherited deafness JAMA 1999 281 2211 2216 10376574 10.1001/jama.281.23.2211
|
16086836
|
PMC1190179
|
CC BY
|
2021-01-04 16:03:33
|
no
|
BMC Med Genet. 2005 Aug 8; 6:30
|
utf-8
|
BMC Med Genet
| 2,005 |
10.1186/1471-2350-6-30
|
oa_comm
|
==== Front
BMC Med EducBMC Medical Education1472-6920BioMed Central London 1472-6920-5-281602950910.1186/1472-6920-5-28Research ArticleFeedback on video recorded consultations in medical teaching: why students loathe and love it – a focus-group based qualitative study Nilsen Stein [email protected] Anders [email protected] Section for General Practice, Department of Public Health and Primary Health Care, University of Bergen, Kalfarveien 31, N-5009 Bergen, Norway2005 19 7 2005 5 28 28 7 4 2005 19 7 2005 Copyright © 2005 Nilsen and Baerheim; licensee BioMed Central Ltd.2005Nilsen and Baerheim; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Feedback on videotaped consultations is a useful way to enhance consultation skills among medical students. The method is becoming increasingly common, but is still not widely implemented in medical education. One obstacle might be that many students seem to consider this educational approach a stressful experience and are reluctant to participate. In order to improve the process and make it more acceptable to the participants, we wanted to identify possible problems experienced by students when making and receiving feedback on their video taped consultations.
Methods
Nineteen of 75 students at the University of Bergen, Norway, participating in a consultation course in their final term of medical school underwent focus group interviews immediately following a video-based feedback session. The material was audio-taped, transcribed, and analysed by phenomenological qualitative analysis.
Results
The study uncovered that some students experienced emotional distress before the start of the course. They were apprehensive and lacking in confidence, expressing fear about exposing lack of skills and competence in front of each other. The video evaluation session and feedback process were evaluated positively however, and they found that their worries had been exaggerated. The video evaluation process also seemed to help strengthen the students' self esteem and self-confidence, and they welcomed this.
Conclusion
Our study provides insight regarding the vulnerability of students receiving feedback from videotaped consultations and their need for reassurance and support in the process, and demonstrates the importance of carefully considering the design and execution of such educational programs.
==== Body
Background
During the last couple of decades, the importance of good patient-doctor communication has been increasingly emphasized, and teaching communication skills in medical school and post graduate courses is no longer a novelty. There is increasing evidence to suggest that this educational input results in an overall improvement in the communication skills of medical students and doctors, especially where the training includes some form of feedback on the trainees' performance [1,2].
One educational approach involves videotaping of consultations between medical student/ doctors and simulated or real patients, and providing personal feedback from others through later assessment of the taped consultations. Feedback on videotape or after direct observation has been shown to enhance development of both general communication skills and more specific consultation techniques [3,4]. This kind of feedback has also been demonstrated to have a more lasting impact on the students' communications skills than conventional education such as lectures or textbooks only, and it has been recommended that all medical students should be provided feedback training [5].
Despite this evidence only a minority of communication training programs provide this kind of individual feedback [6]. A common experience among medical teachers is that students or doctors participating in video-based training programs seem to be very apprehensive beforehand and consider this educational approach to be stressful [7]. This may represent an obstacle to implementation. The few studies assessing this mode of communication teaching have made use mainly of questionnaires or other quantitative research methods to collect information [6,7].
Roter et al did a pre/post comparison of residents' performance after undergoing an innovative video feedback programme. They used a coded, quantitative analysis to demonstrate this method as a powerful and effective teaching tool [6].
In another study, Paul at al made an assessment of the feasibility of video feedback in teaching clinical skills, and of the students' own perception of the effectiveness of this training. They used observer assessment and semi-structured interviews, making a quantitative rating of the students' performance. Among their findings was that feedback on video performances can be useful and effective for improving clinical skills. They also added some open questions, and in this part the students' apprehension before the course became evident [7].
We judged focus group interviews to be the most suitable assessment tool for our purpose, as we believed it would provide more detailed and rich insights into these questions. We wanted to explore this issue in more detail, in order to deal with it more effectively and make the process more acceptable to the participants.
This study explores students' experiences of receiving feedback on their videotaped consultations, and aims to provide insight about how this element of their education can be refined and further developed.
Methods
Communication and consultation training in Bergen University Medical School is partly given as courses in year 3 and 5, but the main emphasis takes places during the final term of the 6 year curriculum, where the main focus is on general practice. The students have a 4-week training period in a GP practice, and also take a course that focuses on different aspects of the consultation process. The course consists of two parts, the first containing practical exercises, group discussions, demonstrations and lectures. In the second part, which has been evaluated in this study, each student videotapes a consultation with a real patient. This is the first time the students' consultations have been videotaped. The consultation takes place at Bergen's community- based Emergency Department, where acutely ill patients from the inner city attend for advice and treatment. Informed consent to videotape the consultation and show it in a closed group of students is obtained from the patients prior to the consultation. The student can, if needed, seek advice and support from a teaching doctor during the consultation.
A few days later the students meet in groups of 6 or 7, led by a mentor experienced in general practice, and also trained in this type of group leadership. The videotapes are reviewed and analysed one at a time, inspired by the ALOBA (Agenda-led, outcome-based) guidelines [8], as follows: Before showing his/her video, the student is encouraged to define critical incidents and/or points of special interest. He/she is asked to appoint a "critical friend", who is assigned to give specific feedback on areas for possible improvement. The other students are also given specific roles in the feedback process, each paying attention to a certain aspect of the consultation. The tape is played back, generally in its entirety, after which the student shares his own assessment of his performance with the group. He/she is invited to define an agenda for discussion, in which the whole group then takes part, seeking to provide balanced feedback, focusing both on what works and areas where improvements could be made. There is opportunity for role-play, focusing on particular points of interest in the consultation, and critical incidents from the tape may be reviewed. The students themselves are encouraged to give one another feedback while the mentor supervises the process, providing comments or further advice when needed. The time allowed for the review session of each student is approx 45–60 minutes, giving a total of 6 hours. The tapes are erased immediately following the feedback session.
In our study, three video review groups out of a total of 12 (19 students out of 75), were invited to participate, and all consented. Their average age was 27.1 years, 58% were males, and all were of Norwegian descent. The members of the groups did not differ significantly from the general student cohort. The students were informed about the purpose of the interview, and given the opportunity to withdraw if they wished. They were also encouraged to observe confidentiality about the group discussion. Written consent to participate was not obtained. The interviews took place immediately after finishing the video feedback sessions.
The interviews were conducted as focus group interviews, as described by Kvale [9]. This approach was chosen for its suitability for collecting qualitative data from a group, especially when assessing attitudes, experience and values in an environment where individuals act or work together. There were 6 or 7 members in each, a size considered ideal for a focus group interview. The interviews were audio-recorded, and the interviewer was supported by an assistant, who took notes. Each interview lasted for 60–90 minutes. The interviews were afterwards transcribed fully by the interviewer. The text was then analysed in a qualitative mode, as described by Giorgi, modified by Malterud [10].
First, the transcripts were read, to get a general overview of the topics commented upon. They were scrutinized to identify all text elements concerning the different aspects of the consultation and the feedback. Each element was coded according to topic or type of factor. The codes were derived from the data, not decided beforehand. Similarly coded elements were interpreted for a common meaning, and were then summarized using expressions close to the students' own words.
The interviewer (SN), a general practitioner, analyzed the focus group interview text and drafted the manuscript. AB made a separate analysis of large parts of the text, and agreement was reached through discussion where differences in analysis appeared.
Results
From the text analysis 3 major themes appeared: Concerns, the feedback process and reassurance.
Concerns
A major theme in the students' evaluation dealt with their apprehension and anxiety prior to the video taped consultations and feedback group discussion. Their concerns evolved around both procedural elements and the feedback process itself.
Among their concerns were:
• carrying out a consultation in an unfamiliar setting or while being videotaped
• embarrassment at watching themselves on videotape together with fellow students
• fear of being shown up as lacking in medical knowledge
• fear of being thought to be inadequate in personality, or in basic communication skills
• fear that if they were judged incompetent at this late stage of their medical training, there would be inadequate time for improvement
One student, seemingly confident and experienced, expressed his worries this way: "The worst thing that could happen would be to demonstrate a complete lack of medical knowledge, to miss something obvious and crucial. "...That, I feel, would be a real blow to me " (male, 25 y, 2. interview). Another student commented that medical students in general worry more than other students about being unsuccessful: "... This is their nature, and to imagine making a video of your own failures is frightening"(female, 27 y, 3.interview). Other concerns were also expressed: One student voiced doubts as to whether specific consultation skills he felt obliged to demonstrate would not be appropriate for the consultation being videoed, and feared he would be criticized for not using these techniques. Another mentioned that being first to show her video made it more embarrassing, it would have been easier to have appeared later in the session.
The feedback process
Some of the respondents stated that despite feeling apprehensive before the start of the course, they experienced few problems once they got into it; and that eventually they realized that most of their fears had been groundless. One student who had technical problems playing back her tape put it this way: "I was not happy with my consultation and hoped that I might somehow lose it. When this actually happened I came to regret it, because I realized afterwards that getting feedback from the others would in fact have been a useful experience for me."(female, 25 y, 2. interview)
Some students gave positive comments on the feedback, as follows:
• the advice I got about what could be improved was being worded carefully and with respect, and did not make me think less of myself afterwards
• feedback always ended with a positive conclusion
• feedback was expressed in a constructive manner, so that possible improvements could be pointed out without anybody losing face.
One student mentioned that she found it easy to agree with the advice given regarding possible improvements, because she had already been able to observe her own areas for development while watching the tape. Another mentioned that although he didn't agree with all the feedback he was given, he still found the discussion valuable. Others described the group environment as feeling safe and protective. They gave several different reasons for this:
• few students in each group
• everybody sharing the same experience
• all showing a positive attitude and willingness to learn from each other
• this type of session provides an environment which is conducive for accepting criticism
Some students found the rules guiding the feedback helpful. One mentioned the way each of them was given a specific role for every new round of feedback, and emphasized the point that the student receiving feedback himself appointed his main "opponent". Another student approved of the expressed aim to make the feedback specific and detailed in order to be of best possible use to the recipient. Some students particularly liked the fact that most of the feedback was given by the students themselves, while the group leader had more of a supervisory role, supplying additional comments where appropriate.
Reassurance
Before watching the videotapes, some of the students said that they did not think their own consultations had gone very well. They admitted to being very self-critical and, due to their limited professional experience as doctors, insecure and vulnerable. Had the feedback from the group been insensitive or harshly-worded, the effect could have been damaging.
After having watched the tapes they were relieved to find that they in fact had done better than expected, and were reassured to have this confirmed by their fellow students.
A number of points were made by the students, following the session. For some, reassurance regarding their professional ability, and more specifically their consultation skills, was of paramount importance. Another put it this way: "Most of the students got positive feedback, as I did myself. I was happy about that: I don't find it easy to take a negative response to something I have put so much of myself into." (Male, 28, 1.interview).
The importance of treating each student individually, so that each is allowed to develop his own personal style was also mentioned. One noted that the session highlighted the fact that many different approaches can be valid, and he learnt a lot by observing others.
One student felt that carrying out one single consultation only added to the stress, since this left no opportunity to become familiar with the situation, and demonstrate improvement the next time. It was suggested that for the future, a number of consultations should be taped.
Discussion
This study showed that consultation training by way of video taping real consultations with subsequent feedback was in general considered acceptable, useful and inspiring by the participating students. However, they experienced a considerable amount of anxiety and apprehension before and during the course, resulting in a strong need for reassurance and a positive evaluation. The feedback process seemed to meet this need, and was described as respectful, easing unwarranted fear, increasing self -esteem and contributing to personal growth of these emerging doctors.
This study highlights aspects that have been noted, but not widely discussed in many earlier studies. An important aspect is the level of anxiety and feeling of vulnerability suffered by the students in this particular educational setting. This initial apprehension by trainees has been noted by others: Paul et al found that the majority of their students reported anxiety and resistance to videotaping, but that their inhibition diminished with increasing practice and experience [7]. Most students in their study believed that they would have gained confidence had they had the opportunity to view a few videotapes of standard consultations before making their own. Smith et al noted the same initial wariness in neurology trainees undergoing a similar educational process, and concluded that doctors may not yet be ready to accept these methods of training without first becoming more familiar with them [12]. However, Rees et al found that undergraduate students preferred experiential methods of learning communication skills to more conventional methods such as lectures, and their students did not emphasize the stressfulness of the experience [13]. One reason for this may be that their students are familiar with making video consultations form their first year of medical school (C. Rees, personal communication, 2005). An important practice implication of this study is that students should be introduced to such teaching methods from the start of their training rather than waiting until later stages, when the stakes are so much higher. Practice in safe environments such as clinical skills resource centres and with simulated patients would also possibly lessen the burden.
In our study we found that some of the observed anxiety and stress among the students was due to the fact that the recorded consultations were made in an environment completely unknown to them, and we suppose that recording the video in a more familiar setting would lessen this burden. This would be achievable if the consultations were taped during primary care training, or if the students were given the opportunity of a second performance after the initial feedback [14].
Many of the students admitted afterwards that their initial worries had been groundless, indicating that they had a misguided impression of the process beforehand. It should be possible to decrease their apprehension by paying more attention to the general information provided, and to specific preparation prior to the course. Another way of dealing with the fear of hostile criticism from others, is to allow the students to choose their own groups. This, however, could risk a too "friendly" environment where important criticisms are less likely to surface. Mavis et al suggest instead that students should be grouped randomly, instead of choosing their own group members, to counteract the observation that students were reluctant to provide less favourable feedback to peers [15].
Many students showed a very self-critical attitude towards their own performance and expressed genuine satisfaction and encouragement from receiving positive evaluation of their consultations. This desire for recognition and reassurance was also noted by Lings and Gray in a study of British GPs participating in an educational program to promote higher standards of care [16]. The doctors in this program were considered likely to be among the most confident to begin with, but still had a strong need for reassurance [17]. One would assume that this kind of support is even more important for students at the start of their careers. Well-earned personal encouragement may provide valuable inspiration to further development of communication skills, which may otherwise often be given low priority among all the other challenges a young doctor is about to face. The students' fears of appearing incompetent must be treated with respect, and mentoring of the groups and the adhesion to the guidelines ruling the feedback process should be given careful consideration. The mentor's role and responsibility in this educational process has not been given much attention so far, and should be investigated further.
The way in which the feedback sessions were conducted has been criticized in prior evaluations of this education model [6]. Among problems observed are superficial reviews of the tapes and difficulties in identifying critical incidences. The problems seem related to the time and cost of faculty training and preparations. To meet this challenge, different approaches have been developed. The method chosen for reviews of the tapes in our program, based on the ALOBA guidelines described earlier [8], seemed to work well, and was praised by many students. One of the keys to success might be the rather detailed guiding rules which the process adhered to, giving each student a specific role and a guide for delivering feedback. This provided a feeling of safety within the group and also gained attention for each part of the consultation, securing a balance between positive and corrective judgment. However, a possible drawback may be that students tend to overestimate the performance of their fellow students [15]. This puts a further responsibility on the mentor to ensure that the feedback is well-balanced.
The students groups interviewed were selected in such a way that each of the three groups had a different mentor, minimizing the possible effect of a specific mentor's personal impact. 19 of the 75 participating students (25%) were interviewed, securing a sufficiently broad sample of opinions. All students invited to participate did so: there were no refusals or withdrawals. The students knew each other well and appeared secure and relaxed in the interview setting, indicating that the interview group provided an environment where they could speak out freely. They were also granted anonymity as to the content of their given opinions. These factors may lessen an intra-group trend towards conformity which could otherwise present a problem, since participants who are unsure or alien to the group may tend to voice agreement with opinions already expressed by others [11]. The interviews were conducted immediately after the video feedback sessions, leaving no time for possible reflection and second thoughts about the experience. After a long and intense session, the students may have been tired and anxious to get away, and less likely to put much thought into their answers. On the other hand, carrying out the interviews straight away contributed to full participation from the students. The experience was fresh in their minds, perhaps resulting in a response more immediate than it would have been at a later stage, and therefore enhancing the validity of their comments.
The students were asked to participate while they were in a group doing the video feedback review, and the focus interview was carried out right afterwards. They did not get a written invitation prior to this, and therefore had limited opportunity to consider the invitation and refuse to participate. This might represent a coercion to participate, and could help explain a 100% participation.
The evaluation appeared to be overall very positive, and one might suspect that intrinsic factors would inadvertently facilitate this. One such factor might be that the interviewer (SN) also was a member of the faculty staff, and this could represent a potential coercion towards too positive views. SN had, however, only been a part time (20%) employee for 6 months when the interviews took place, and the majority of the students had never met him before the interview. Thus, the students had no reason to fear consequences of expressing a negative view. Also, an interview guide was used to secure varied approaches to the issues in question, and to encourage differing views. Another possible issue is the fact that the groups invited to be interviewed were not chosen in a truly randomly fashion, but for pragmatic reasons. Still, the selection was made in a way that should make them representative of the whole cohort.
The students' general enthusiasm found here is also in agreement with the results of the written free text assessment that all students attending this course carry out after each teaching term. These written assessments have earlier been analysed formally [18].
Conclusion
This study shows that students participating first time in peer feedback as part of their training in consultations, experience a high degree of anxiety and apprehension before and during the course; also demonstrating a strong need for reassurance and positive evaluation, both of their efforts and of their professional ability. The feedback process described here would appear to address this issue, and might contribute to the empowerment of the participants by increasing their confidence. This approach requires careful attention to design and procedure, in order to provide sufficient support for the students or trainees, both during the video recording sessions and the feedback process. The study also implicates that this educational approach should be introduced early in the medical school curriculum, in order to lessen the stress later.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
SN conducted the interviews, transcribed the text, performed the analysis and drafted the paper. AB was involved in the conception and design of the study, participated in analysis and edited the paper.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Thanks to Yngvild Skaatun Hannestad for assistance during the interviews, and for useful discussions during the process. Thanks to Angela Rowe for valuable advice concerning language. Thanks to the participating students for their willingness to participate and for enthusiastic cooperation
==== Refs
Yedidia MJ Gillespie CC Kachur E Effect of communications training on medical student performance JAMA 2003 290 1157 1165 12952997 10.1001/jama.290.9.1157
Aspegren K BEME Guide no 2: Teaching and learning communication skills in medicine: a review with quality grading articles 1999 Dundee: Association for Medical Education in Europe
Holmwood C Direct observation: A primer for supervisors of doctors in training Aus Fam Physician 1998 27 48 51 9503706
Beckman HB Frankel RM The use of videotape in internal medicine training J Gen Intern Med 1994 9 517 521 7996296
Maguire P Fairbairn S Fletcher C Consultation skills of young doctors: Benefits of feedback training in interviewing as students persists Br Med J (Clin Res Ed) 1986 292 1573 1576 3719282
Roter DL Larson S Shinitzky H Use of an innovate video feedback technique to enhance communication skills training Med Educ 2004 38 145 157 14871385 10.1111/j.1365-2923.2004.01754.x
Paul S Dawson KP Lanphear JH Cheema MY Video recording feedback: a feasible and effective approach to teaching history-taking and physical examination skills in undergraduate paediatric medicine Med Educ 1998 32 332 336 9743791 10.1046/j.1365-2923.1998.00197.x
Kurtz SM Silvermann JD Draper J Teaching and Learning Communication Skills in Medicine 1998 Abingdon: Radcliffe Medical Press
Kvale S Interviews: An Introduction to Qualitative research Interviewing 1996 Thousand Oaks: Sage Publications
Malterud K Kvalitative metoder i medisinsk forskning 1996 Oslo: Tano
Espeland A Baerheim A Factors affecting general practitioners' decisions about plain radiography for back pain- a qualitative study BMC Health Serv Res 2003 3 8 12659640 10.1186/1472-6963-3-8
Smith PEM Fuller GN Kinnersley P Brigley S Elwyn G Using simulated consultations to develop communications skills for neurology trainees Eur J Neurol 2002 9 83 87 11784381 10.1046/j.1468-1331.2002.00339.x
Rees C Sheard C McPherson A Medical students' views and experiences of methods of teaching and learning communication skills Patient Educ Couns 2004 54 119 121 15210269 10.1016/S0738-3991(03)00196-4
Eaooskoon W Sumawong V Silpakit C Evaluation of training medical students in patient-interviewing skills by three modes of learning J Med Assoc Thai 1996 79 526 530 8855636
Mavis BE Ogle KS Lovell KL Madden LM Medical students as standardized patients to assess interviewing skills for pain evaluation Med Educ 2002 36 135 140 11869440 10.1046/j.1365-2923.2002.01070.x
Lings P Gray DP Professional development for general practitioners through Fellowship by Assessment Med Educ 2002 36 360 365 11940177 10.1046/j.1365-2923.2002.01173.x
van Dalen J Assessment practices undermine self-confidence Med Educ 2002 36 310 311 11940168 10.1046/j.1365-2923.2002.01191.x
Barheim A Meland E Schei E Evaluating the patient-centered consultation course in Bergen Tidsskr Nor Legeforen 2000 120 2263 2265
|
16029509
|
PMC1190180
|
CC BY
|
2021-01-04 16:30:55
|
no
|
BMC Med Educ. 2005 Jul 19; 5:28
|
utf-8
|
BMC Med Educ
| 2,005 |
10.1186/1472-6920-5-28
|
oa_comm
|
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-5-231604865310.1186/1472-6947-5-23Research ArticleThe GuideLine Implementability Appraisal (GLIA): development of an instrument to identify obstacles to guideline implementation Shiffman Richard N [email protected] Jane [email protected] Cynthia [email protected] Abdelwaheb [email protected] Allen [email protected] George [email protected]'Connell Ryan [email protected] Yale Center for Medical Informatics, Yale University School of Medicine, New Haven, CT, USA2 Yale School of Nursing, Yale University, New Haven, CT, USA2005 27 7 2005 5 23 23 28 2 2005 27 7 2005 Copyright © 2005 Shiffman et al; licensee BioMed Central Ltd.2005Shiffman et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Clinical practice guidelines are not uniformly successful in influencing clinicians' behaviour toward best practices. Implementability refers to a set of characteristics that predict ease of (and obstacles to) guideline implementation. Our objective is to develop and validate a tool for appraisal of implementability of clinical guidelines.
Methods
Indicators of implementability were identified from the literature and used to create items and dimensions of the GuideLine Implementability Appraisal (GLIA). GLIA consists of 31 items, arranged into 10 dimensions. Questions from 9 of the 10 dimensions are applied individually to each recommendation of the guideline. Decidability and Executability are critical dimensions. Other dimensions are Global, Presentation and Formatting, Measurable Outcomes, Apparent Validity, Flexibility, Effect on Process of Care, Novelty/Innovation, and Computability. We conducted a series of validation activities, including validation of the construct of implementability, expert review of content for clarity, relevance, and comprehensiveness, and assessment of construct validity of the instrument. Finally, GLIA was applied to a draft guideline under development by national professional societies.
Results
Evidence of content validity and preliminary support for construct validity were obtained. The GLIA proved to be useful in identifying barriers to implementation in the draft guideline and the guideline was revised accordingly.
Conclusion
GLIA may be useful to guideline developers who can apply the results to remedy defects in their guidelines. Likewise, guideline implementers may use GLIA to select implementable recommendations and to devise implementation strategies that address identified barriers. By aiding the design and operationalization of highly implementable guidelines, our goal is that application of GLIA may help to improve health outcomes, but further evaluation will be required to support this potential benefit.
==== Body
Background
Tremendous resources have been invested in the development and implementation of clinical practice guidelines over the past 15 years [1-3]. In spite of these efforts, however, guidelines are not uniformly successful in improving care and several instances of implementation failure have been described, some resulting in substantial waste of time and resources [4-6]. In many cases, implementation failures have been related to factors extrinsic to the guideline itself – e.g., organizational and provider-specific obstacles inherent in a particular system of care that interfere with implementation success. In other cases, however, factors intrinsic to the guideline have contributed to implementation failure, e.g., ambiguity, inconsistency, and incompleteness [7,8]. We believe it is particularly important to identify these intrinsic factors, because in many cases they can be ameliorated or fully remedied by guideline authors while the guideline is being developed. If these problems are not captured during guideline development, they must be addressed during implementation.
Guideline implementation involves "the concrete activities and interventions undertaken to turn policies into desired results" [9]. We define implementability to refer to a set of characteristics that predict the relative ease of implementation of guideline recommendations. Measures of successful implementation include improved adherence to guideline-prescribed processes of care and, ultimately, improved patient outcomes. Indicators of implementability, on the other hand, focus on the ease and accuracy of translation of guideline advice into systems that influence care. In this paper, we describe a tool for appraisal of implementability that is intended to help anticipate barriers to implementation success. We first delineate the process by which indicators of implementability were chosen. Then we describe the steps we took in the preliminary validation of the evolving instrument and show an example of its application. Finally, we discuss how the instrument might be used in practice.
Methods
Instrument development
The first step in the measurement of implementability was to define its attributes. From a broad-based literature search, we identified several key papers and book chapters that describe the impact of a variety of factors on success of implementation. The report from the Institute of Medicine [9] included general definitions and several high level constructs relevant to implementation. Thorsen and Mäkelä [10] described critical factors in implementation strategy that facilitate use of guidelines and overcome barriers to adoption. Solberg et al. asked expert implementers about implementation success factors and identified 83 variables grouped into 5 clusters [11]. Applying diffusion of innovation theory, Grilli and Lomas identified guideline complexity, trialability, and observability as critical factors for successful implementation [12]. Grol and colleagues found that vagueness, controversy, demand for a change in routines, and absence of an evidence-base differentiated guidelines that were not followed from those that were [13]. Finally, we examined 3 instruments for appraisal of guideline quality – Cluzeau's 37-item instrument [14], the AGREE instrument [15], and Shaneyfelt's Guideline Quality Appraisal Questionnaire [16]– and extracted factors from each that addressed implementability.
The authors eliminated redundant factors, i.e., those that appeared in several sources or represented concepts that were subsumed by others, through open discussion and consensus. Factors that indicated guideline quality but not implementability were excluded. We decided early on to focus the GLIA on factors that were intrinsic to the guideline, because they could be addressed centrally by a guideline development committee. Thus, we eliminated many factors related to Solberg's medical group characteristics, organizational capability for change, infrastructure for implementation, and external environment. Extrinsic items relating to a recommendation's effect on the process of care and items relating to the novelty or innovation of a guideline statement were retained in the instrument because developers can anticipate these barriers and offer potential strategies for implementation success.
All remaining factors were grouped into categories of related constructs, hereafter referred to as dimensions. We then devised specific questions to characterize each dimension and phrased them so that negative responses identified barriers. These questions ultimately became items of the instrument. We iteratively refined the items, further clarified definitions, and re-categorized items into the most appropriate dimensions.
Validation
To explore the construct of implementability and its measurement, we carried out a series of validation activities. The steps are summarized in Table 2 and are described below in the sequence in which they were carried out. Concomitant with these validation activities, the GLIA instrument underwent iterative refinement and revision.
Table 2 Summary of validation activities.
Step Process Purpose or Validity Type Results
1 Ranking of implementability of 3 recommendations by experts at COGS Validity of the construct of implementability Consistent ranking
2 Guideline review with early GLIA versions To refine GLIA Demonstrate feasibility of measurement of implementability
3 Expert review of GLIA items and dimensions by HL7 experts Content validity of GLIA GLIA items rated generally as relevant and clear. Two new items were added, definitions were clarified, and explanatory material was added.
4 Review of the recommendations ranked in Step 1 Construct validity of GLIA GLIA assessment of these three guidelines was consistent with experts' rankings of implementability (Step 1). Barriers to implementation were explicitly identified.
Results
Final instrument
The table shown in [Additional file 1] summarizes the 10 dimensions of the instrument and their definitions. (The GLIA instrument is available for download at .) Of the 31 items in GLIA, the first dimension (Global) contains 7 items that relate to the guideline as a whole. The remainder of the dimensions focus on the individual recommendation as the unit of implementability, since a single guideline may contain recommendations that vary widely in their implementability. For each GLIA item, each recommendation is rated using one of four response options (see Table 1). Additional comments that explain each response are encouraged. GLIA users should discuss all divergent ratings and try to achieve consensus. Any items scored with "?" should be resolved: often, this requires the help of an expert in the guideline's topic area.
Table 1 GLIA response options.
Y The recommendation meets this criterion fully.
N The recommendation does not meet this criterion.
? Rater is unable to address this question because of insufficient knowledge or experience in this area.
NA Criterion is not applicable to this recommendation.
When any GLIA item is assigned an "N" response, its corresponding barrier to implementation is recorded on the summary sheet with a brief description of why the recommendation failed the criterion (see Figure 1). Examination of the barriers recorded on the summary sheet should provide an understanding of predicted impediments to implementation of the recommendation. The summary sheet also contains a column where suggested remedies can be described.
Figure 1 Example of a GLIA Summary Report on draft recommendations from a guideline for diagnosis and management of otitis media with effusion.
Validation
Step 1: Evaluation of implementability as a concept that is understood by experts
In April 2002, the Conference on Guideline Standardization (COGS) brought together 23 national and international leaders in guideline development, dissemination and implementation [17]. This gathering provided the opportunity to explore the concept of implementability. We selected 3 recommendations for review by the experts based on the following criteria:
• Each recommendation concerned a common clinical problem generally understood by the attendees.
• The three recommendations as a group represented a broad range of implementation challenges.
Using our own global subjective judgment of implementability, one of the recommendations was considered to be straightforward to implement, another was considered to be quite difficult, and the remaining recommendation was considered to occupy an intermediate position. The recommendations we selected concerned: aspirin therapy in management of acute myocardial infarction [18], diagnosis of urinary tract infection in young children with unexplained fever [19], and evaluation of unexplained syncope [20].
Divided into 2 groups (10 individuals with considerable experience in guideline implementation in one group, 11 individuals experienced in guideline development and dissemination in the other), participants were asked to rank the relative implementability of the guideline recommendations. After discussion, each group came to a consensus ranking of the three recommendations' implementability. These rankings were consistent between the two groups and also with the authors' expectation. Both groups agreed that the aspirin recommendation would be easiest to implement and the syncope algorithm would be most challenging.
This step provided support for our basic assumption that the construct of implementability is real. Moreover, implementability varies among recommendations in a way that can be systematically recognized by experts.
Step 2- Initial guideline review with early GLIA versions
The authors next tested an early version of GLIA with a convenience sample of 20 guideline recommendations. These reviews provided our first experience using the GLIA in practice and led us to identify areas of rating controversy. Early attempts to apply a simple numeric scoring system failed because all the items were not of equal importance. In addition, we recognized the advantage of qualitative rating in understanding specific barriers to implementation. We appreciated the importance of including a variety of expertise among members of the GLIA team. This experience led to refinement of the instrument, as well as to a deeper understanding of the optimal process for conducting a review using GLIA.
The 20 replications of our process seemed quite adequate to provide us with assurance that creating an instrument to assess implementability was a feasible goal. They also gave us a view of the problems still to be addressed.
Step 3 - content validation: Expert review of GLIA items and dimensions
To investigate the content validity of the draft GLIA, we systematically examined each item's clarity, its relevance to its superordinate dimension, and the comprehensiveness of the GLIA as a whole [21]. We prepared a Questionnaire for Expert Review on which relevance and clarity ratings were scored on a four point scale, with higher numbers representing greater relevance and greater clarity. We distributed the Questionnaire to volunteers attending a Workgroup Meeting of the HL7 Special Interest Group (SIG) on Clinical Guidelines. The SIG includes representatives from academia, vendors of electronic health record software, and healthcare providers, who meet three times each year to discuss standardization of guideline components.
Judgments about the GLIA were obtained from 7 guideline implementers. Mean relevancy ratings by item ranged from 2.7 to 4.0. On average, 26 of 30 items were rated as "moderately" or "highly relevant." No reviewer used a relevancy rating of 1 ("not at all relevant") for any item.
Mean clarity ratings by item ranged from 2.3 to 4.0. On average, 29 of 30 items were rated as "clear" or "very clear." The correlation of relevancy ratings and clarity ratings by item was .23 (non-significant), indicating that the two ratings of each item provided different information.
Raters were also asked for suggestions about the items and numerous comments were provided. Based on this feedback from implementers, GLIA was again revised. Items that received low clarity or relevance scores were carefully reviewed, definitions were clarified, and explanatory material was added to better communicate the meaning of the item. Based on suggestions from the reviewers, two new items that explored recommendation sequencing and the internal consistency of the guideline were added to the Global dimension.
This process of obtaining judgment data from implementers proved to be a rich source of ideas for refinement of items. It also provided supportive evidence for the general content validity of the developing instrument.
Step 4 -Review of ranked recommendations
Next, we used GLIA to formally assess the three recommendations whose implementability had been ranked previously by experts at the COGS Meeting (described above in Step 1). Results using GLIA were consistent with the earlier expert rankings, i.e., more critical implementation barriers were identified in the recommendation ranked least implementable and no barriers were identified in the recommendation ranked most implementable. Moreover, appraisal with GLIA allowed us to itemize specific obstacles, thereby clarifying particular impediments to implementation in contrast to the global subjective evaluation performed in Step 1.
This finding of agreement with expert rankings indicated that GLIA results reflect the construct of implementability as conceptualized by experts. This consistency provided preliminary evidence that supports the construct validity of GLIA assessments.
Step 5 – Application to a draft guideline from the American Academy of Pediatrics
To assess the value of application of GLIA in a real world guideline development effort, we applied a late version of the appraisal to a draft clinical guideline for the management of otitis media with effusion (OME) that was in preparation by a joint committee of national professional societies – the American Academy of Pediatrics, the American Academy of Family Physicians, and the American Academy of Otolaryngology. We received an intermediate draft of the guideline for quality appraisal and evaluation of implementability.
After independently rating the draft guideline, we met as a group to discuss the barriers to implementation that we identified. While not every barrier was identified by every rater, no single barrier was uniquely identified. Remarkably little discussion was needed to reach a consensus on anticipated barriers to implementation.
Our report to the guideline authors identified several instances of problems with decidability and executability of individual recommendations. For example, the draft recommended, "During the initial assessment of the child with OME, the clinician should document (a detailed set of physical findings)" (italics added). The GLIA report drew attention to the fact that the guideline's users might not consistently determine when in the course of continuous care an assessment is initial. Adding to the confusion, the draft guideline text later states that these findings should be ascertained at every medical encounter. The recommendation that was ultimately published stated: "Clinicians should document (the physical findings) at each assessment of the child with OME." Vagueness was also inherent in use of the terms "academic risk", "when necessary", and "individualized management" without clear definitions. Following the GLIA report, each of these was clarified in the final guideline publication [22].
GLIA appraisal also identified extrinsic barriers to implementation that were reported to the Joint OME Implementation Committee of the professional societies. These barriers included:
• Recommendations to perform pneumatic otoscopy, tympanometry, hearing screening, and language assessment would require acquisition of new equipment and skills on the part of many physicians.
• Recommendations against prescription of antihistamines, decongestants, antimicrobials and corticosteroids for effusions may not be compatible with patient expectations.
The committee considered these potential barriers in their design of a guideline implementation strategy.
Discussion
Many clinical guidelines – developed at substantial cost and effort – have proven to be difficult or impossible to operationalize. We developed the GuideLine Implementability Appraisal to facilitate guideline implementation. The instrument is designed to systematically highlight barriers to implementation.
The GLIA is intended to provide feedback about a guideline's implementability to two distinct audiences: the authors of the guideline and those individuals who choose guidelines for application within a health care delivery system. As a guideline is being developed, GLIA can provide feedback to guideline authors about potentially remediable defects. Developers may choose to make modifications to the guideline document before it is finalized and disseminated. Implementers can use GLIA to help select a guideline, to identify potential obstacles, and to target efforts toward addressing identified barriers. Thus, GLIA can be used to help select guidelines that are more easily implementable and also to devise implementation strategies that address identified barriers.
Two GLIA dimensions are of particular importance because failure to address them adequately will result in inconsistent implementation [23,24]. Any recommendation that does not clearly communicate what to do (i.e., it fails executability criteria) or when to do it (i.e., fails decidability criteria) is not fully ready for implementation. If possible, guideline authors should revise such a recommendation before it is disseminated for implementation. If problems in decidability and executability are not corrected prior to dissemination, different implementers may well interpret the guideline authors' intent in a discordant manner.
GLIA incorporates an optional dimension – computability – to indicate the ease with which a recommendation might be operationalized in an electronic information system. Guideline implementation strategies – e.g., education, academic detailing, audit and feedback, administrative sanctions – need not necessarily involve computers [25]. Because of the success of electronic implementations in influencing clinician behavior [26-28], the wide variation in electronic information systems, and the current lack of guidance regarding computability, these extrinsic considerations were retained in the instrument. Items in the computability dimension address the availability of data to trigger the recommendation, the level of specificity of the triggers and recommended actions, and whether there is a clear path from recommendation to electronic implementation.
Implementability must be differentiated from guideline quality. Quality assessments relate primarily to determining the scientific validity of guidelines and, generally, quality is assessed for the guideline as a whole. Implementability, on the other hand, is one component of guideline quality, but its assessment is applied largely to individual recommendations within a guideline.
In a comprehensive review of 13 tools for guideline quality appraisal, Graham [29] identified instruments developed by Cluzeau [14] and Shaneyfelt et al [16] as the best developed. Since that time, the AGREE Instrument [15] has emerged as the leading exemplar of guideline quality appraisal and it has been endorsed by the Guidelines International Network [30]. Application of GLIA can complement quality appraisal in identifying guideline deficiencies. Several GLIA items overlap with items in the AGREE scale. However, to the best of our knowledge, GLIA is the only tool that emphasizes implementation concerns at the level of the individual recommendation.
Application of GLIA to measurement of decidability and executability requires that users translate guideline recommendations into statements comprising conditions and actions [31]. Training and practice may be required to assure consistency of logical analysis.
Application of GLIA requires team effort and a consequent resource investment. The team that applies GLIA should include members with skills in guideline implementation as well as members with specific understanding of the clinical domain. Our experience has demonstrated that rating a guideline that contains 15 recommendations might require several hours of an individual's time. Additional time must be spent in resolving divergent ratings, although this effort usually yields an improved understanding of implementation issues. We are currently developing an electronic version of GLIA that will provide a more efficient means of rating, scoring, and reporting results.
During development, GLIA is best applied once evidence has been synthesized and draft recommendations have been formulated. When applied too late in the authoring process, GLIA may have limited impact because authors may have already become committed to recommendations as written and thus not open to making modifications. Assuring that guideline authors understand the importance of implementability early on may help to overcome premature commitment.
The classic challenge of instrument development is to arrive at the correct number of items to minimize burden and avoid redundancy, while including a sufficient number to be comprehensive. GLIA contains 7 global items that are applied once to each guideline, 20 items that are applied to each rated recommendation, and 4 optional items rating computability that are applicable when an electronic implementation is planned. Further use of GLIA is likely to result in clarification and perhaps modification of the number of items.
Limitations
We have performed a series of activities to provide preliminary evidence of GLIA's validity. However, neither the inter-rater reliability of GLIA, the test-retest reliability, the factor structure of the dimensions, nor its predictive validity has yet been established. Plans for this testing are underway.
It should be noted that the authors performed the activities described as Step 4. Until verified by an independent panel, these results should be considered preliminary and subject to potential bias.
In addition, GLIA addresses primarily factors intrinsic to the guideline. It is clear that extrinsic factors are critical in a successful implementation. Future extensions to GLIA will help to identify extrinsic barriers to implementation.
Conclusion
The GuideLine Implementability Appraisal provides a tool designed to help developers and implementers better understand and anticipate barriers to successful implementation. By aiding the design and operationalization of highly implementable guidelines, our goal is that application of GLIA may help improve health outcomes. Demonstration of GLIA's effectiveness will require prospective testing.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RS conceived the instrument, performed the literature review, participated in all phases of instrument development and validation, and edited the manuscript.
JD participated in all phases of instrument development and validation and edited the manuscript.
CB, AE, AH, GM, and RO participated in all phases of instrument development and validation and reviewed the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
GLIA dimensions and example items
Click here for file
Acknowledgements
The work was supported through grants R01-LM-07199, R29- LM05552, and T15-LM07065 from the U.S. National Library of Medicine and by R13 HS10962 from the Agency for Healthcare Research and Quality.
==== Refs
Grimshaw JM Thomas RE MacLennan G Fraser C Ramsay CR Vale L Whitty P Eccles MP Matowe L Shirran L Wensing M Dijkstra R Donaldson C Effectiveness and efficiency of guideline dissemination and implementation strategies. Health Technol Assess 2004 8 1 72 Review
Larson E Status of practice guidelines in the United States: CDC guidelines as an example. [Review] [64 refs] Preventive Medicine 2003 36 519 524 12689796 10.1016/S0091-7435(03)00014-8
Burgers JS Grol R Klazinga NS Makela M Zaat J AGREE Collaboration Towards evidence-based clinical practice: an international survey of 18 clinical guideline programs Int J Qual Health Care 2003 15 31 45 12630799 10.1093/intqhc/15.1.31
Eccles M McColl E Steen N Rousseau N Grimshaw J Parkin D Effect of computerised evidence-based guidelines on management of asthma and angina in adults in primary care: cluster randomised controlled trial BMJ 2002 325 941 948 12399345 10.1136/bmj.325.7370.941
Tierney WM Overhage JM Murray MD Harris LE Zhou XH Eckert GJ Smith FE Nienaber N McDonald CJ Wolinsky FD Effects of computerized guidelines for managing heart disease in primary care J Gen Intern Med 2003 18 967 976 14687254 10.1111/j.1525-1497.2003.30635.x
Switzer GE Halm EA Chang CC Mittman BS Walsh MB Fine MJ Physician awareness and self-reported use of local and national guidelines for community-acquired pneumonia J Gen Intern Med 2003 18 816 823 14521644 10.1046/j.1525-1497.2003.20535.x
Tierney WM Overhage JM Takesue BY Harris LE Murray MD Vargo DL McDonald CJ Computerizing guidelines to improve care and patient outcomes: the example of heart failure J Am Med Inform Assoc 1995 2 316 322 7496881
Katz DA Barriers between guidelines and improved patient care: an analysis of AHCPR's unstable angina clinical practice guideline Health Serv Res 1999 34 377 389 10199682
Institute of Medicine Field MJ, Lohr KN Guidelines for clinical practice: from development to use 1992 Washington, DC , National Academy Press
Thorsen T Mäkelä M Changing professional practice: theory and practice of clinical guidelines implementation 1999 Copenhagen , Danish Institute for Health Services Research and Development 216
Solberg LI Brekke ML Fowles J Jacobsen DN Kottke TE Mosser G O'Connor PJ Ohnsorg KA Rolnick SJ Lessons from experienced guideline implementers:attend to many factors and use multiple strategies Jt Comm J Qual Improv 2000 26 171 188 10749003
Grilli R Lomas J Evaluating the message: the relationship betweenm the compliance rate and the subject of a practice guideline Medical Care 1994 32 202 213 8145598
Grol R Dalhuijsen J Thomas S Veld C Rutten G Mokkink H Attributes of clinical guidelines that influence use of guidelines in general practice: observational study BMJ 1998 317 858 861 9748183
Cluzeau FA Littlejohns P Grimshaw JM Feder G Moran SE Development and application of a generic methodology to assess the quality of clinical guidelines Int J Qual Health Care 1999 11 21 28 10411286 10.1093/intqhc/11.1.21
AGREE Collaboration Development and validation of an international appraisal instrument for assessing the quality of clinical practice guidelines: the AGREE project Qual Saf Health Care 2003 12 18 23 12571340 10.1136/qhc.12.1.18
Shaneyfelt TM Mayo-Smith MF Rothwangl J Are guidelines following guidelines? the methodological quality of clinical practice guidelines in the peer-reviewed medical literature JAMA 1999 281 1900 1905 10349893 10.1001/jama.281.20.1900
Shiffman RN Shekelle P Overhage JM Slutsky J Grimshaw J Deshpande AM Standardized reporting of clinical practice guidelines: a proposal from the Conference on Guideline Standardization Ann Intern Med 2003 139 493 498 13679327
Ryan TJ Antman EM Brooks NH Califf RM Hillis LD Hiratzka LF rapaport E Riegel B Russell RO Smith EE weaveer WD 1999 Update: ACC/AHA guidelines for the management of patients with acute myocardial infarction
Anonymous Practice parameter: the diagnosis, treatment, and evaluation of the initial urinary tract infection in febrile infants and young children. American Academy of Pediatrics. Committee on Quality Improvement. Subcommittee on Urinary Tract Infection [published errata appear in Pediatrics 1999 May;103(5 Pt 1):1052, 1999 Jul;104(1 Pt 1):118 and 2000 Jan;105(1 Pt 1):141] [see comments] Pediatrics 1999 103 843 852 10103321
Linzer M Yang EH Estes M Wang P Vorperian VR Kapoor WN Diagnosing syncope: Part 2. unexplained syncope Ann Intern Med 1997 127 76 86 9214258
Lynn MR Determination and quantification of content validity Nursing Research 1986 35 382 385 3640358
Anonymous Otitis media with effusion Pediatrics 2004 113 1412 1429 15121966 10.1542/peds.113.5.1412
Michie S Johnston M Changing clinical behaviour by making guidelines specific BMJ 2004 328 343 345 14764503 10.1136/bmj.328.7435.343
Shekelle PG Kravitz RL Beart J Marger M Wang M Lee M Are nonspecific practice guidelines potentially harmful? a randomized trial of the effect of nonspecific versus specific guidelines on physician decision making Health Services research 2000 34 1429 1448 10737446
Greco PJ Eisenberg JM Changing physicians' practices N Engl J Med 1993 329 1271 1274 8413397 10.1056/NEJM199310213291714
Hunt DL Haynes RB Hanna SE Smith K Effects of computer-based clinical decision support systems on physician performance and patient outcomes: a systematic review JAMA 1998 280 1339 1346 9794315 10.1001/jama.280.15.1339
Shiffman RN Liaw Y Brandt CA Corb GJ Computer-based guideline implementation systems: a systematic review of functionality and effectiveness J Am Med Inform Assoc 1999 6 104 114 10094063
Kaplan B Evaluating informatics applications--clinical decision support systems literature review Int J Med Inform 2001 64 15 37 Review 11673100 10.1016/S1386-5056(01)00183-6
Graham ID Calder LA Hebert PC Carter AO Tetroe JM A comparison of clinical practice guideline appraisal instruments Int J Tech Assess Health Care 2000 16 1024 1038 10.1017/S0266462300103095
Ollenschlager G Marshall C Qureshi S Rosenbrand K Slutsky J Burgers J Makela M Board of Trustees 2002, Guidelines International Network (G-I-N) Improving the quality of health care: using international collaboration to inform guideline programmes - by founding the Guidelines International Network (G-I-N) Qual Safe Health Care 2004 13 455 460 10.1136/qshc.2003.009761
Shiffman RN Representation of clinical practice guidelines in conventional and augmented decision tables J Am Med Inform Assoc 1997 4 382 393 9292844
|
16048653
|
PMC1190181
|
CC BY
|
2021-01-04 23:52:11
|
no
|
BMC Med Inform Decis Mak. 2005 Jul 27; 5:23
|
utf-8
|
BMC Med Inform Decis Mak
| 2,005 |
10.1186/1472-6947-5-23
|
oa_comm
|
==== Front
BMC Med Inform Decis MakBMC Medical Informatics and Decision Making1472-6947BioMed Central London 1472-6947-5-241607638510.1186/1472-6947-5-24Research ArticleAudio computer-assisted self-interviewing (ACASI) may avert socially desirable responses about infant feeding in the context of HIV Waruru Anthony K [email protected] Ruth [email protected]är Thorkild [email protected] Centres for Disease Control and Prevention (CDC), Kenya Medical Research Institute (KEMRI), P.O Box 1578, 040100 Kisumu, Kenya2 Network for AIDS Researchers in Eastern and Southern Africa (NARESA), P.O Box 10654, 00100 Nairobi, Kenya3 Centre for International Health, University of Bergen, Armauer Hansen Bld, N-5021 Bergen, Norway2005 2 8 2005 5 24 24 19 11 2004 2 8 2005 Copyright © 2005 Waruru et al; licensee BioMed Central Ltd.2005Waruru et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Understanding infant feeding practices in the context of HIV and factors that put mothers at risk of HIV infection is an important step towards prevention of mother to child transmission of HIV (PMTCT). Face-to-face (FTF) interviewing may not be a suitable way of ascertaining this information because respondents may report what is socially desirable. Audio computer-assisted self-interviewing (ACASI) is thought to increase privacy, reporting of sensitive issues and to eliminate socially desirable responses. We compared ACASI with FTF interviewing and explored its feasibility, usability, and acceptability in a PMTCT program in Kenya.
Methods
A graphic user interface (GUI) was developed using Macromedia Authorware® and questions and instructions recorded in local languages Kikuyu and Kiswahili. Eighty mothers enrolled in the PMTCT program were interviewed with each of the interviewing mode (ACASI and FTF) and responses obtained in FTF interviews and ACASI compared using McNemar's χ2 for paired proportions. A paired Student's t-test was used to compare means of age, marital-time and parity when measuring interview mode effect and two-sample Student's t-test to compare means for samples stratified by education level – determined during the exit interview. A Chi-Square (χ2test) was used to compare ability to use ACASI by education level.
Results
Mean ages for intended time for breastfeeding as reported by ACASI were 11 months by ACASI and 19 months by FTF interviewing (p < 0.001). Introduction of complementary foods at ≤3 months was reported more frequently by respondents in ACASI compared to FTF interviews for 7 of 13 complementary food items commonly utilized in the study area (p < 0.05). More respondents reported use of unsuitable utensils for infant feeding in ACASI than in FTF interviewing (p = 0.001). In other sensitive questions, 7% more respondents reported unstable relationships with ACASI than when interviewed FTF (p = 0.039). Regardless of education level, respondents used ACASI similarly and majority (65%) preferred it to FTF interviewing mainly due to enhanced usability and privacy. Most respondents (79%) preferred ACASI to FTF for future interviewing.
Conclusion
ACASI seems to improve quality of information by increasing response to sensitive questions, decreasing socially desirable responses, and by preventing null responses and was suitable for collecting data in a setting where formal education is low.
==== Body
Background
Research conducted in low-income countries often yield contradictory results where sensitive issues are concerned even in similar studies and populations [1]. One source of bias is socially desirable responses in face-to-face (FTF) interviewing. Consequently, data from studies on stigma-associated diseases such as HIV/AIDS may be flawed with potentially far-reaching implications on decision-making and public health policies.
The foregoing argument implies that traditional FTF interviewing is unsuitable for collecting potentially sensitive data and calls for alternative methods of interviewing that increase privacy and are less confrontational. ACASI is a method of interviewing conducted in a private setting and the interaction happens only between the respondent and the computer. This person-machine interaction eliminates the need for respondents to reply in a socially desirable manner, as would be the case in FTF interviewing. Studies have shown that privacy provided by ACASI increases reporting of sensitive behaviours [2-7]. Apparently, this increased reporting is not attributable to random errors [4,8]. A plausible explanation given by Cooley and Turner suggests that increased reporting is because respondents are more willing to respond in an ACASI than in FTF interviewing [9]. ACASI is more so apt where education is low since it eases the response task [2,7,8,10]. ACASI program has other advantages such as preventing non-response by guiding respondents through the interview by providing prompts. Skips and branching are automated preventing punching errors occurring in manual data entry. With ACASI, the interview is standardised hence avoiding interviewer variations that would occur in many research settings where FTF interviewing is carried out with the assistance of several interviewers.
ACASI has been shown to be feasible among women in Africa [8]. However, information on its use for interviewing mothers with low education and on sensitive issues in a clinical setting was lacking.
Methods
Respondents
This study was carried within the PMTCT pilot project at the Karatina District Hospital, Nyeri, Kenya. Mothers enrolled in the project regularly visiting the antenatal clinic during a three-months period (Sept – Nov 2002) were invited to participate in the study. By the end of the study, we had approached 104 mothers and 80 of them had consented to be interviewed. This number formed the sample. Kikuyu and Kiswahili were the languages of instructions and questioning and all respondents were able to hear and understand either.
Study design
Using a crossover design, all the respondents were interviewed with each method of interviewing (ACASI and FTF). The first block of 70 respondents were interviewed using ACASI first, and the second block of 10 respondents were interviewed using FTF first. In the subsequent interviews, respondents undertook the opposite interview from the one that they had the first time. To investigate preferences, and attitudes for the two interview modes, we asked respondents questions about their experiences with both methods of interviewing in an exit FTF interview.
Study expectations and questions
In the context of HIV, mothers face stigma when choosing feeding options and they would want to conform to what is socially desirable. It was hypothesised that in FTF interview, respondents could give socially desirable responses. It was also hypothesised that ACASI could increase response rates for sensitive questions on relationships. The study sought to assess response differences when ACASI was used compared FTF interviewing. We also wanted to explore feasibility, usability, and acceptability of ACASI in a clinical setting.
Study instruments and data analysis
To handle the questioning and provide instructions and cues to respondents, a graphic user interface (GUI) of the ACASI program was developed using Macromedia Authorware® [11]. We adopted 52 questions (a part of the English version of the questionnaire used in the PMTCT project), which were translated into Kikuyu and Kiswahili versions. Audio recording for both the Kikuyu and Kiswahili questions was done using Audiotools® shareware version [12]. Audio editing was done in Adobe Premiere® [13]. Finally, individual audio clips were inserted to correspond to questions in the ACASI program and a test-run of the program.
For implementation, a laptop, attached keypad (Targus™ Numeric keypad), and headphones were used (figure 1). All respondents went through two practice questions after which they were left on their own. It was necessary to guide respondents through the two practice questions in order to familiarise them with the use of the input device (external keypad) especially on how much pressure to exert on the keys since too much pressure would have resulted in repeated character entry. At the end of the ACASI, each interviewee's inputs were automatically added to a tab-delimited text record file and the amount of time that each respondent spent answering questions with ACASI was similarly recorded automatically. In the FTF interviewing inputs were directly entered by use of the keypad. At the end of the FTF interview, a differently named text file similar to the one produced in the ACASI was automatically generated. To determine preferences and attitudes for the two interviewing methods, we asked respondents questions about their experiences with both methods in an exit interview.
Figure 1 The audio computer-assisted self-interviewing (ACASI) set-up as it was used in Kenya. The mother being interviewed is sitting in front of a portable computer listening to the questions in her own language in the headset. She provides the answers by the small keypad with colour-coded buttons (enlarged bottom left). Note: Figure has been obscured for anonymity.
Data were analysed data using SPSS version 10.1 [14] and Epi 6.04 [15]. To compare means for age, marital-time and parity, we used a paired Student's t-test when measuring interview mode effect. We measured time taken by each mother to complete ACASI and used a Student's t-test to compare mean time taken by two groups of mothers – those educated up primary school level compared to mothers educated up to secondary school level. Level of education was determined for each mother from the exit interview. A Wilcoxon Signed Rank test was used for continuous variables for which distributions were skewed. For categorical variables, a Pearson Chi-Square (χ2test) was used to compare proportions for independent variables and when the expected value was less than 5, the Fisher's Exact Test was used. McNemar's χ2test was used to compare differences between paired proportions for categorical variables. For qualitative data, responses were reported literally and numbers of respondents giving similar responses indicated. Since the sample was larger than 30, we assumed normality in all analysis and a p-value of < 0.05 was considered statistically significant.
The Kenya PMTCT pilot project and the questionnaires used in this study had ethical approval from the Kenyatta national hospital ethical review committee and the national council of science and technology. The regional committee for medical research ethics, Norway-West also approved the study. Additional approval was not sought but all respondents orally consented to participate in the study.
Results
Demographic characteristics
Respondents comprised mostly of young mothers and their mean age was 25 years, (SD ± 5, range 17–38 years). Most respondents 49 (61%) had attained primary education and the rest 31 (39%) had attained secondary education. Most respondents (84%) were living in monogamous relationships. Mean duration for marital relationship was 4 years, (SD ± 4) by both modes of interviewing and parity ranged from 1 to 9. For all the demographic variables, data were not significantly different by mode of interviewing. Children to whom questions on infant feeding refer were more than 9 months old at the time of the interview.
Infant feeding practices
On the intended time for breastfeeding, data varied significantly by mode of interviewing. Fourteen respondents (18%) indicated that they wanted to breastfeed their infants for an extended period of 24 months by ACASI while in the FTF interview, 44 (55%) indicated that they wanted to breastfeed their infants for 24 months (p < 0.001).
Respondents were asked whether they had introduced a selection of 13 food items (table 1). In all cases apart from 'baby formula' and 'fresh packet milk', the number of respondents who indicated that they had introduced the foods ≤3 months of age was fewer in the FTF interview than in the ACASI. These differences were significant (p < 0.05) for 7 food items compared to differences for only 3 food items in proportions of mothers who had introduced food items at 1 year of an infants age (table 1). A graphical presentation of 3 of the 13 food items: milk, protein-rich foods, and herbal tea is presented in (figure 2).
Table 1 Comparison of proportions of mothers who claimed to have introduced different food items at 3 and at 12 months by mode of interviewing: Audio computer-assisted self-interviewing (ACASI) or face-to-face (FTF) interview.
Foods and fluids At ≤ 3 months of age At 1 year of age
ACASI n (%) FTF n (%) p a) ACASI n (%) FTF n (%) p a)
Plain water 54 (68) 50 (63) ns b) 72 (90) 79 (98) 0.016
Glucose water 30 (38) 25 (31) ns 39 (49) 41 (51) ns
Juice 12 (15) 3 (4) 0.012 18 (23) 12 (15) ns
Tea 20 (25) 6 (8) 0.001 40 (50) 41 (51) ns
Baby formula 4 (5) 4 (5) ns 5 (6) 6 (8) ns
Powder milk 0 (0) 0 (0) ns 0 (0) 0 (0) ns
Fresh packet milk 3 (4) 3 (4) ns 6 (8) 5 (6) ns
Cow's or goat milk 35 (44) 18 (23) 0.002 70 (88) 76 (95) ns
Other fluids 27 (34) 23 (29) ns 56 (70) 72 (90) 0.001
Fruits and vegetables 41 (51) 17 (21) 0.000 80 (100) 80 (100) ns
Cassava and plantains 61 (76) 76 (95) 0.000 61 (76) 76 (95) 0.000
Starchy foods 31 (39) 17 (21) 0.016 69 (86) 72 (90) ns
Eggs/poultry/fish/meat 26 (33) 4 (5) 0.000 56 (70) 56 (70) ns
a) McNemars χ2 test
b) ns not significant
Figure 2 Three examples of difference in reporting infant feeding by method of interviewing: Audio computer-assisted self-interviewing (ACASI) or face-to-face (FTF). Number of mothers, who report giving milk or protein-rich foods (egg, meat or fish), whole milk and herbal tea to their babies according to the age of the child (x-axis). A higher proportion of mothers reported having given milk or protein-rich foods to their children below 6 months of age with the ACASI compared to FTF, interpreted as a reflection of the fact that it is socially desirable not to give milk or protein-rich foods before 6 months of age.
Irrespective of interview mode, equal proportions of respondents 42 (53%) said they used a cup and spoon to feed their children in both ACASI and FTF interviewing respectively (table 2). However, more respondents reported use of a cup with holes on the snout in the ACASI than in the FTF interview (p = 0.001).
Table 2 Reported utensils used to feed infants by mode of interviewing: Audio computer-assisted self-interviewing (ACASI) or face-to-face (FTF) interview.
Utensils ACASI FTF
Cup 15 (19%) 6 (8%)
Cup and spoon 42 (53%) 42 (53%)
Closed cup with holes on the snout 20 (25%) 5 (6%)
Other 3 (4%) 27 (34%)
p a) 0.001
a) McNemars χ2 test
Respondents' experiences with the ACASI program
We recorded time taken to complete the ACASI interview for 74 respondents. The mean time taken by mothers who had attained primary education was 25 minutes and 46 seconds and for mothers who had attained secondary education 25 minutes and 56 seconds a non-significant difference of only 10 seconds. In the exit interview, 56, (70%) of the respondents said they were able to answer questions without any problems and 24, (30%) indicated they had some hitches. However, of the 24, only 11 asked for help and the rest were able to continue on their own. Sixteen, (20%) of the respondents did not press the repeat key. Of the 64 who did, 61 pressed only a few times -usually once- while the other three pressed the button often. All the respondents listened to the recorded audio questions and instructions at all times. Ability to use ACASI was not different across education levels (table 3).
Table 3 Comparison of the ability to use audio computer-assisted self-interviewing (ACASI) between mothers with primary versus secondary education.
Indicators for ability to use ACASI Primary education (n = 49) Secondary education (n = 31) p a)
Yes No Yes No
Difficult questions 4 45 5 26 0.298 b)
Unable to answer a question 16 33 8 23 0.515
Asked for help 9 7 2 6 0.303 c)
Pressed repeat key 42 7 22 9 0.108
a) Pearson χ2
b) Fisher's exact test used
c) The question was relevant for only those who were unable to answer a question
Preferences for and attitudes to mode of interviewing
Most respondents 52 (65%) preferred ACASI to FTF interviewing and the rest preferred either FTF (18 respondents, 23%), or both methods of interviewing (10 respondents, 13%). Preference though, did not vary by education level. Major reasons given for ACASI preference were: privacy, faster interviewing, and aspects of program usability such as (audio repetitions and prompts, the simultaneous on-screen text, clear instructions, the easy-to-use colour-coded number keypad). Some preferred FTF interviewing because the interviewer could clarify questions and respondents could ask other questions related to their own health in a FTF encounter. Three respondents felt that the computer was rigid in the way it asked questions.
Perceived privacy and confidentiality
Majority of the respondents (84%) thought ACASI offered more privacy than FTF interviewing. Five respondents thought FTF interviewing offered more privacy, 3 considered both methods equally private and 5 considered neither method private. Preference and perceived privacy were similar between both levels of education.
To assess the importance that respondents assigned to confidentiality, we asked them to tell us how important it was for their responses not to be seen by others. Twenty-six, (33%) indicated that 'it did not matter' while the majority 54 (68%) indicated that 'it was important that others do not see their responses'.
Discussion
This is the first study comparing ACASI and FTF conducted in an African rural clinical setting among mothers with low formal education. ACASI and FTF interviewing yielded similar results to neutral questions but differed consistently whenever questions of a sensitive nature were asked. Since our aim was not to validate any of the methods of interviewing, any of the two methods could have yielded the more correct answers. However, consistency of data divergence between the methods suggests that ACASI yielded responses that were less influenced by perceptions of what was socially desirable compared to FTF interviewing.
At the antenatal clinic, the nurses stress in their morning health talk sessions the importance of exclusive breastfeeding for the first 6 months of an infant's life and advocate for extended breastfeeding for at least two years. Our study demonstrates early (≤ 3 months) and extensive introduction of food items, irrespective of interview mode. This information concurs with a study done in Malawi showing that exclusive breastfeeding is uncommon and complementary foods are introduced early in rural families [16]. In the ACASI mode of interviewing even higher proportions of mothers reported early introduction of feeds, which suggests that the FTF responses were biased towards socially desirable responses.
Our data also indicates a bias of the FTF data in the direction of socially desirable responses for the intended length of breastfeeding. A larger proportion of mothers responded that they intended to breastfeed their infants for 24 months during FTF interviewing (44 respondents, 55%) as opposed to during ACASI (11 respondents, 14%). A plausible reason would be that respondents would want to confine their responses in FTF interviews to what they had been taught during the health talk sessions. Reported mean age of introducing protein-rich foods such as meat, eggs, or fish by ACASI mode was 117 days (approximately 4 months) and 173 days (approximately 6 months) by FTF interviewing. This discrepancy could mean that respondents knew that they should introduce protein-rich foods when their babies were older (as was taught during health talks) but in practice, they introduced protein-rich foods to babies at a young age.
More respondents (25%) indicated that they used a closed cup with holes on the snout for infant feeding in the ACASI interview compared to 6% in the FTF interview. Since this utensil is hard to clean, its use is discouraged in the morning health talks. It appears that the respondents gave socially desirable responses in the FTF interview hence the lower proportion reporting.
This study had some limitations. The ratio of respondents who undertook each of the interviewing methods in the first and subsequent interview were dissimilar (70:10 – ACASI first to FTF interviewing first). This was not due to design but because of lack of sufficient number of respondents in the second sequencing arm and we could not therefore effectively measure the effect of sequencing on responses. However, changing the order of interviewing only resulted to a decline in time respondents took to complete the ACASI interview among those who started on FTF interview first and ACASI second (t (72) = 2.626 p = 0.011, CI: 2 – 13 minutes) and had no effect on responses. Secondly, assignment of respondents to each of the interviewing sequencing arms was not random since it was done as a group of 70 (ACASI interviewing first) and the second group of 10 (FTF interviewing first) rather than an alternate assignment of respondents during recruitment to each of the interviewing mode.
Conclusion
Studies have shown that ACASI increases reporting of sensitive issues in high-income countries but until recently, few studies had tried this methodology in resource-poor settings. This is the first study on ACASI conducted in an African rural clinical setting among mothers with low formal education. Our study results are in agreement with results of several studies that have shown that ACASI increases reporting of sensitive issues [2-7]. Differences in responses with the two methods of interviewing show that data on infant feeding in the sensitive context of HIV when collected during face-to-face interviews may be influenced by the respondents tendency to give socially desirable responses. Relying on such data may lead to imprecise decisions on public health issues. Use of ACASI in research or program settings involving sensitive issues can improve data, information and conciseness of decision-making.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
A Waruru participated in all aspects of the study including planning, design of the ACASI program, data collection and data analysis, and in writing the manuscript, R Nduati designed the initial questionnaire, prepared the study site, supervised data collection, participated in analysis and interpretation of data and in the writing of the manuscript. T Tylleskär had the initial idea for the study, supervised the study planning, the design of the ACASI interviews and participated in the analysis and interpretation of data and in the writing of the manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Research assistants of the PMTCT project at Karatina District Hospital-Kenya helped with the translation of the English questionnaire to Kikuyu and Kiswahili, did voices for the ACASI and were instrumental in informing the respondents about the study. We also thank the staff at the Karatina District Hospital, antenatal clinic for their cooperation. The University of Bergen, Norway funded the study and had no influence on any decisions taken in the study.
==== Refs
Barbara S. Mensch Paul C. Hewett Annablel Erulkar The Reporting of Sensitive Behaviour Among Adolescents: A Methodological Experiment in Kenya 2001 New York, Population Council
Debra A. Murphy Stephen Durako Larry R. Muenz Craig M. Wilson Marijuana Use among HIV-positive and High-risk Adolescents: A Comparison of Self-report through Audio Computer-assisted Self-administered Interviewing and Urinalysis American Journal of Epidemiology 2000 152 805 813 11085391 10.1093/aje/152.9.805
David S. Metzger Beryl Koblin Charles Turner Helen Navaline Francesca Valenti Sarah Holte Michael Gross Amy Sheon Heather Miller Philip Cooley George R. Seage III Randomized Controlled Trial of Audio Computer-Assisted Self-Interviewing: Utility and Acceptability in Longitudinal Studies American Journal of Epidemiology 2000 152 99 106 10909945 10.1093/aje/152.2.99
Charles F. Turner Leighton Ku Frenya L. Warnecke R Impact of ACASI on Reporting of Male-Male Sexual Contacts: Preliminary Results From the 1995 National Survey of Adolescent Males: ; Breckenridge, Colorado. 1995 DHHS 171 176
Jessica Clark Newman Don C. Des Jarlais Charles F Turner Jay Gribble Phillip Cooley Denise Paone The Differential Effects of Face-to-Face and Computer Interview Modes American Journal of Public Health 2002 92 294 297 11818309
Roger Tourangeau Tom W. Smith Asking Sensitive Questions: The Impact of Data Collection Mode, Question Format and Question Content Public Opinion Quarterly 1996 60 275 304 10.1086/297751
Don C. Des Jarlais Denise Paone Judith Milliken Charles F Turner Heather Miller James Gribble Qiuhu Shi Holly Hagan Samuel R Friedman Audio-computer interviewing to measure risk behaviour for HIV among injecting drug users: a quasi randomized trial Lancet 1999 353 1657 1662 10335785 10.1016/S0140-6736(98)07026-3
Janneke van de Wijgert Nancy Padian Stephen Shiboski Charles Turner Is audio computer-assisted self-interviewing a feasible method of surveying in Zimbabwe? International Journal of Epidemiology 2000 29 885 890 11034973 10.1093/ije/29.5.885
Philip C. Cooley Charles F. Turner Implementing Audio-CASI on Windows Platforms Computers in Human Behavior 1998 14 195 207 10.1016/S0747-5632(98)00001-6
M. L. Williams R. C. Freeman A. M. Bowen L. Saunders The acceptability of a computer HIV/AIDS risk assessment to not-in-treatment drug users AIDS Care 1998 10 701 711 9924525 10.1080/09540129848325
Macromedia Authorware 1987 5.1.0.0 , Macromedia Inc
Unrelated-Inventions Audiotools 1998 3.70.3.0 , Unrelated Inventions
Adobe Adobe Premiere 1991 6.0.0.0 , Adobe Systems Incorporated
SPSS SPSS 10.1 for Windows 1989 10.1 Chicago, Chicago Inc.
Centers for Disease Control and Prevention World Health Organization Epi Info 6.04 2001 6.04 Atlanta, Centers For Disease Control and Prevention (CDC), U.S.A and World Health Organization, Geneva, Switzerland
Vaahtera M Kulmala T Hietanen A Ndekha M Cullinan T Salin ML Ashorn P Breastfeeding and complementary feeding practices in rural Malawi Acta Paediatr 2001 90 328 332 11332176 10.1080/080352501300067730
|
16076385
|
PMC1190182
|
CC BY
|
2021-01-04 23:52:11
|
no
|
BMC Med Inform Decis Mak. 2005 Aug 2; 5:24
|
utf-8
|
BMC Med Inform Decis Mak
| 2,005 |
10.1186/1472-6947-5-24
|
oa_comm
|
==== Front
BMC Med Res MethodolBMC Medical Research Methodology1471-2288BioMed Central London 1471-2288-5-251610916210.1186/1471-2288-5-25Research ArticleA priori postulated and real power in cluster randomized trials: mind the gap Guittet Lydia [email protected] Bruno [email protected] Philippe [email protected] Département d'Epidémiologie, Biostatistique et Recherche Clinique, Groupe Hospitalier Bichat-Claude Bernard (AP-HP) – Université Paris 7, Paris, France2 INSERM U 738, Université Paris 7, Paris, France3 INSERM CIC 202, Faculté de Médecine, Université François Rabelais, Tours, France4 INSERM U 717, Université Paris 7, Paris, France2005 18 8 2005 5 25 25 8 3 2005 18 8 2005 Copyright © 2005 Guittet et al; licensee BioMed Central Ltd.2005Guittet et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Cluster randomization design is increasingly used for the evaluation of health-care, screening or educational interventions. The intraclass correlation coefficient (ICC) defines the clustering effect and be specified during planning. The aim of this work is to study the influence of the ICC on power in cluster randomized trials.
Methods
Power contour graphs were drawn to illustrate the loss in power induced by an underestimation of the ICC when planning trials. We also derived the maximum achievable power given a specified ICC.
Results
The magnitude of the ICC can have a major impact on power, and with low numbers of clusters, 80% power may not be achievable.
Conclusion
Underestimating the ICC during planning cluster randomized trials can lead to a seriously underpowered trial. Publication of a priori postulated and a posteriori estimated ICCs is necessary for a more objective reading: negative trial results may be the consequence of a loss of power due to a mis-specification of the ICC.
==== Body
Background
A cluster randomized trial involves randomizing social units or clusters of individuals, rather than the individuals themselves. This design, which is increasingly used for evaluating health-care, screening and educational interventions [1-3], presents specific constraints that must be considered during planning and analysis [4,5].
The responses of individuals within a cluster tend to be more similar than those of individuals of different clusters. This correlation leads to an increased required sample size in randomized trials of clusters compared with that of individuals, although this clustering effect is rarely taken into account. Thus, in a recent review of cluster randomized trials in primary care, Eldridge et al [6] reported that only 20% of studies accounted for clustering in the sample size calculation. Similar results were found in other reviews, as listed by Bland [7]. The increase in sample size is measured through an inflation factor, which is a function of both the cluster size and the intraclass correlation coefficient (ICC), which appraises the correlation between individuals within the same cluster [1-3,8]. Therefore an a priori value for this correlation must be postulated during planning. However, estimates of this correlation are rarely available, and, if available, are often uncertain. Indeed the correlation would differ according to outcome, setting, intervention, covariate adjustment and also sampling [5,9,10]. Therefore, a discrepancy between a priori postulated and a posteriori estimated ICCs may occur.
The discrepancy between a priori postulated and a posteriori estimated ICCs may be reduced by intermediate estimation of the ICC, thus allowing a re-estimation of the required sample size [11]. However, the room to manœuvre to increase the sample size may be restricted. Indeed, including new clusters may be difficult, either because the number of clusters is limited [10,12-15] (which may occur when the randomization unit is defined by a geographic area or hospital, for example) or because clusters are frequently randomized all at once and not one at a time. The cluster size itself may also be limited (e.g., by the size of a family or because the number of patients followed up in a clinical practice cannot be increased [16]), which then disallows the increase in sample size by increasing cluster size.
The purpose of our study was to assess the consequence on power of the ICC and to what extent the discrepancy between a priori postulated and a posteriori estimated ICCs may induce a loss in power in cluster randomized trials.
Methods
We considered a completely cluster randomized design with a continuous outcome (normally distributed) measured at a single time point. We assumed an equal number of clusters randomized to each arm and a fixed common cluster size. The sample size is calculated as follows [1]:
where m is the cluster size, g is the number of clusters per arm, ρ is the ICC, ES is the effect size (defined as ratio between the absolute difference between the two intervention-specific means (|Δ|) and the standard deviation (σ)) and z1-α/2 and z1-β are the critical values of the standard normal distribution corresponding to error rates α (two-sided) and β, respectively. One recognizes the sample size calculation for an individually randomized trial inflated by a factor equal to [1 + (m - 1)ρ] defined as the variance inflation factor. When the cluster size varies, m refers to the average cluster size.
Power contour graphs
To quantify the influence of the ICC on the power, we drew two kinds of power contour graphs. First, considering an effect size and an a priori postulated ICC, we considered several combinations of numbers of clusters and cluster sizes that allow for achieving 80% power. Then considering these combinations, we plotted the real power as a function of the ICC, which may differ from the a priori postulated value. Two values of a priori postulated ICC (0.005, 0.02) and five numbers of clusters per intervention arm (3, 5, 10, 20 and 40) were considered for these graphs. The effect size was fixed at 0.25.
We also drew power contour graphs, showing combinations of cluster sizes and number of clusters leading to a pre-specified power, with type I error fixed at 5%. Four power levels were considered (90, 80, 60 and 40%), 3 effect sizes (0.25, 0.50 and 0.75) and 4 levels of ρ (0.005, 0.020, 0.050 and 0.100). These ICC values were chosen according to previously published estimates [3,6,12,16-23].
Maximal theoretical achievable power
We determined the maximal achievable power given a limited number of randomized clusters (i.e., considering an infinite cluster size) or a limited cluster size (i.e., considering an infinite number of clusters). For a limited number of clusters, results were graphically illustrated by considering 5 numbers of clusters per intervention arm (3, 5, 10, 20 and 40) for 2 effect sizes (0.25, 0.5).
Results
Influence of the discrepancy between a priori postulated and a posteriori estimated ICCs on power
Figure 1 displays the real power associated with a study whose a posteriori estimated ICC would differ from the a priori postulated one. With an a priori ICC of 0.02, as few as 5 clusters per intervention arm is not enough to achieve a power of 80% to detect an effect size of 0.25. Power decreases as the ICC increases, and the loss is all the more important when the number of clusters is small. For example, if the a priori ICC was fixed at 0.005 and the a posteriori ICC is as high as 0.01, the power falls to 70.8% with 5 clusters per intervention arm, instead of the targeted 80% power, whereas the power is almost safeguarded with 20 clusters per intervention arm (real power 77.7%).
Figure 1 Real power of cluster randomized trials according to the discrepancy between the a priori postulated and a posteriori estimated intraclass correlation coefficients. The effect size to be detected is fixed at 0.25 and power at 80%. g is the number of clusters per arm, m is the average cluster size and N is the total number per intervention arm considering an a priori postulated ICC of 0.005 or 0.02.
Figure 2 displays power contour graphs for combinations of numbers of clusters and cluster sizes. First, let us consider the situation of a fixed number of clusters. In many situations, even a slight increase in the ICC has a great influence on power and leads to a major increase in the required cluster size to keep the desired power. In some situations, even reaching the required power may no longer be possible: the power contour curves tend to be infinite. For instance, assuming that 15 clusters are randomized to each arm and we want to detect an effect size of 0.25 with 80% power, we would need an average cluster size of 25 patients with an a priori ICC fixed at 0.02. To keep 80% power, the required mean cluster size should be increased to 98 for an ICC of 0.05, which represents 1095 more subjects per arm. If the ICC actually equals 0.10, 80% power is no longer achievable without recruiting additional clusters. The phenomenon is all the more acute when the number of fixed clusters is low.
Figure 2 Power contour graphs for several intraclass correlation coefficients (ICCs) and effect sizes*. Effect size is presented in columns and ICC in rows. In situations above or to the right of the red curve, the statistical power is greater than 90%. In situations between the red and blue curves, the statistical power is between 80% and 90%. In situations between the blue and red curves, the statistical power is between 60% and 80%. For vertical curves, increasing the cluster size is pointless. The number of subjects required, assuming individual randomization, is 24 to achieve a power of 40% to detect an effect size of 0.50, and 38, 28, 18 and 11 to achieve powers of 90%, 80%, 60% and 40%, respectively, to detect an effect size of 0.75, thus, the reason why curves are truncated. *Effect size = absolute difference between the two intervention-specific means divided by the S.D. of the response variable.
Second, when the mean cluster size is limited but the number of clusters is not, an increase in the ICC may also be of great consequence. As an example, considering a mean cluster size of 100, we would need to randomize 8 clusters per arm to detect a 0.25 effect size with 80% power when the ICC is fixed at 0.02. This number of clusters is raised to 15 and 28 when the ICC is fixed at 0.05 and 0.10, respectively, or 700 and 2000 additional subjects, respectively, per arm.
Maximal theoretical power with infinite cluster size
In a cluster randomized trial aimed at detecting an effect size ES at a pre-specified α level with an a priori postulated ICC equal to ρ, changing the cluster size m and/or the number of clusters g per group changes β and therefore power. Power is thus related to the f(m, g) = (z1-α/2 + z1-β)2 function defined as
When m, the mean cluster size, tends to be infinite, f(m, g) tends to be an asymptotic value, but there is no limit when g, the number of clusters, is infinite:
Therefore, although power is not theoretically limited when the number of clusters can be increased, a maximal reachable power is possible when this number is fixed and only the cluster size can be increased. This maximum theoretical power is defined as:
where Φ-1( ) refers to the inverse cumulative function associated with the standard normal distribution. This maximal theoretical power decreases when the ICC increases and/or the number of clusters decreases (Figure 3). In some cases, an 80% or 90% power is not achievable even with a theoretical situation of infinite cluster sizes. Thus, when 5 clusters are randomized to each arm, a power of 80% to detect an effect size of 0.50 cannot be achieved if the ICC is greater than 0.079 (and this limit equals 0.058 when 90% power is considered). For an effect size of 0.25, this upper ICC limit is 0.019 for 80% power and 0.014 for 90% power.
Figure 3 Theoretical maximal power assuming an infinite cluster size for several fixed numbers of clusters according to two different effect sizes *. *Effect size = absolute difference between the two intervention-specific means divided by the S.D. of the response variable.
Discussion
The ICC is a nuisance parameter that has to be a priori specified when planning a cluster randomized trial. The magnitude of this coefficient has a major impact on power, particularly with a small number of randomized clusters. Our results were derived considering a continuous outcome, but in their simulation study, Donner and Klar [24] showed that power never differs from more than one percentage point in continuous or binary outcomes. Moreover, we did not take into account any potential variability in cluster size, which is already known to reduce power [25]. When planning cluster randomized trials, variability in cluster size is rarely taken into account, and the cluster size m is generally replaced by the mean cluster size. An underestimation of the ICC may therefore be expected to have similar consequences when cluster size is constant. In the end, an underestimation of the ICC during planning could therefore lead to a severely underpowered study and thus questionable results.
In cluster randomized trials, it is known that for a fixed total number of subjects, the higher the number of clusters (and thus the smaller the average cluster size), the higher the power [2,4,5,14,24,26,27]. In the extreme case, in clusters of size one, individuals are randomized, with no loss of power because of correlation between subjects. Moreover, it has also been shown that increasing cluster size improves the power up to a certain threshold, which depends on the value of the ICC [24,27]. Therefore, when planning a cluster randomized trial, the optimal strategy is indeed to randomize a large number of clusters [1,2,12,29]. Such a strategy first allows for decreasing the total sample size for a pre-specified power and second, as our results show, protects against a loss of power induced by an underestimation of the ICC when planning. However, because of logistic constraints, the number of randomized clusters may be limited, and indeed, the review by Eldridge et al [6] noted that half of the cluster randomized trials analyzed had fewer than 29 clusters in each arm. Therefore, for most cluster randomized trials, the a priori postulated value of the ICC has a great impact on power.
When planning trials, the a priori postulated ICC will rarely be very reliable. During the study, an intermediate estimation of the ICC can be assessed, thus allowing a sample size adjustment [11]. But the determination of this intermediate estimation is not without error, as was shown in the study by Moore et al [28], in which the intermediate ICC was 0.012 and the final one 0.031. A sensitivity analysis must therefore be undertaken when planning, to account for uncertainty of the ICC. In the extreme situations, when very few clusters can be randomized, such a sensitivity analysis may illustrate the high risk of performing an underpowered study and thus highlight arguments for not performing the study.
When reporting the study results, investigators should publish both the ICC used during the planning and the a posteriori estimated one, as recommended initially by some authors and recently by the extension of the CONSORT statement for cluster randomized trials [27,29-31]. However, such information is rarely available. We studied cluster randomized trials published between January 2003 and December 2004 in the British Medical Journal, "which contains more such reports than any other journal" [7], and the published extension of the CONSORT statement [30]). Of 16 published studies, 5 (31.2%) did not report an a priori postulated ICC and 2 reported no sample size calculation. Only 5 (31.2%) reports provided a posteriori estimated ICCs (without any confidence intervals). Such under-reporting disallows assessing the discrepancy between the a priori postulated ICC and the a posteriori estimated one. However, reporting both ICCs would help readers "assess the appropriateness of the original sample size calculations as well as the magnitude of the clustering for each outcome" [30] and help investigators design future trials [1,27,31]. It would also help readers understand trial results, particularly negative ones: a study may prove to be negative just by a loss of power induced by an a priori underestimation of the ICC. On a formal point, the publication format of the a posteriori estimated ICC should follow the recommendation by Campbell et al., who advocate specifying a description of the data set and information on the method used to assess it and the precision of the estimate [32].
In conclusion, our study supports modifications in investigators' practices when planning trials and reporting results, taking into account the uncertainty of the ICC by favoring a high number of clusters and publishing this parameter. For readers, an objective reading of trial results, particularly negative results, requires knowledge of a priori and a posteriori estimated ICCs.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
This study was designed by LG, BG and PhR. LG performed the statistical analysis and drafted the article, which was then revised by BG and PhR.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors are indebted to Dr. Sandra Eldridge for constructive comments.
This work was funded by a grant from the Foundation for Medical Research (FRM).
==== Refs
Donner A Klar N Design and Analysis of Cluster Randomization Trials in Health Research 2000 London, England: Arnold
Murray DM Design and Analysis of Group-Randomized Trials 1998 New York, NY: Oxford University Press Inc
Reading R Harvey I McLean M Cluster randomised trials in maternal and child health: implications for power and sample size Arch Dis Child 2000 82 79 83 10630921 10.1136/adc.82.1.79
Murray DM Varnell SP Blitstein JL Design and analysis of group-randomized trials: a review of recent methodological developments Am J Public Health 2004 94 423 432 14998806
Donner A Klar N Pitfalls of and controversies in cluster randomization trials Am J Public Health 2004 94 416 422 14998805
Eldridge SM Ashby D Feder GS Rudnicka AR Ukoumunne OC Lessons for cluster randomized trials in the twenty-first century: a systematic review of trials in primary care Clinical trials 2004 1 80 90 16281464 10.1191/1740774504cn006rr
Bland JM Cluster randomised trials in the medical literature: two bibliometric surveys BMC Med Res Methodol 2004 4 21 15310402 10.1186/1471-2288-4-21
Kerry SM Bland JM Statistics notes: sample size in cluster randomisation BMJ 1998 316 549 9501723
Adams G Gulliford MC Ukoumunne OC Eldridge S Chinn S Campbell MJ Patterns of intra-cluster correlation from primary care research to inform study design and analysis J Clin Epidemiol 2004 57 785 794 15485730 10.1016/j.jclinepi.2003.12.013
Turner RM Prevost AT Thompson SG Allowing for imprecision of the intracluster correlation coefficient in the design of cluster randomized trials Stat Med 2004 23 1195 1214 15083478 10.1002/sim.1721
Lake S Kammann E Klar N Betensky R Sample size re-estimation in cluster randomization trials Stat Med 2002 21 1337 1350 12185888 10.1002/sim.1121
Ukoumunne OC Gulliford MC Chinn C Sterne JAC Burney PGJ Methods for evaluating area-wide and organisation-based interventions in health and health care: a systematic review Health Technology Assessment (Winchester, England) 1999 3 iii 92 10982317
Ukoumunne OC Gulliford MC Chinn S Sterne JAC Burney PGJ Donner A Evaluation of health interventions at area and organisation level BMJ 1999 319 376 379 10435968
Flynn TN Whitley E Peters TJ Recruitment strategies in a cluster randomized trial-cost implications Stat Med 2002 21 397 405 11813226 10.1002/sim.1025
Campbell MK Thomson S Ramsay CR MacLennan GS Grimshaw JM Sample size calculator for cluster randomized trials Comput Biol Med 2004 34 113 25 14972631 10.1016/S0010-4825(03)00039-8
Gulliford MC Adams G Ukoumunne OC Latinovic R Chinn S Campbell MJ Intraclass correlation coefficient and outcome prevalence are associated in clustered binary data J Clin Epidemiol 2005 58 246 251 15718113 10.1016/j.jclinepi.2004.08.012
Hannan PJ Murray DM David RJ Mc Govern PG Parameters to aid in the design and analysis of community trials: intraclass correlations from the Minnesota Heart Health Program Epidemiology 1994 5 88 95 8117787
Siddiqui O Hedeker D Flay BR Hu FB Intraclass correlation estimates in a school-based smoking prevention study Am J Epidemiol 1996 144 425 433 8712201
Smeeth L Siu-Woon Ng E Intraclass correlation coefficients for cluster randomized trials in primary care: data from the MRC Trial of the Assessment and Management of Older People in the Community Control Clin Trials 2002 23 409 421 12161083 10.1016/S0197-2456(02)00208-8
Martinson BC Murray DM Jeffery RW Hennrikus DJ Intraclass correlation for measures from a worksite health promotion study: estimates, correlates, and applications Am J Health Promot 1999 13 347 357 10557507
Murray DM Phillips GA Birnbaum AS Lytle LA Intraclass correlation for measures from a middle school nutrition intervention study: estimates, correlates, and applications Health Educ Behav 2001 28 666 679 11720271
Murray DM Clark MH Alexander CW Intraclass correlation from a community-based alcohol prevention study: the effect of repeat observations on the same communities J Stud Alcohol 2000 61 881 890 11188494
Murray DM Short BJ Intraclass correlation among measures related to tobacco use by adolescents: estimates, correlates, and applications in intervention studies Addictive behaviours 1997 22 1 12 10.1016/0306-4603(95)00099-2
Donner A Klar N Statistical considerations in the design and analysis of community intervention trials J Clin Epidemiol 1996 49 435 439 8621994 10.1016/0895-4356(95)00511-0
Manatunga AK Hudgens MG Chen S Sample size estimation in cluster randomized studies with varying cluster size Biom J 2001 1 75 86 10.1002/1521-4036(200102)43:1<75::AID-BIMJ75>3.0.CO;2-N
Kerry SM Bland JM Unequal cluster sizes for trials in English and Welsh general practice: implications for sample size calculations Stat Med 2001 20 377 390 11180308 10.1002/1097-0258(20010215)20:3<377::AID-SIM799>3.0.CO;2-N
Donner A Some aspects of the design and analysis of cluster randomization trials Appl Statistics 1998 47 95 113
Moore L Campbell R Whelan A Mills N Lupton P Misselbrook E Frohlich J Self help smoking cessation in pregnancy: cluster randomised controlled trial BMJ 2002 325 1383 1388 12480850 10.1136/bmj.325.7377.1383
Murray DM McKinlay SM Martin D Donner AP Dwyer JH Raudenbush SW Graubard BI Design and analysis issues in community trials Evaluation Review 1994 18 493 514
Campbell MK Elbourne DR Altman DG CONSORT statement: extension to cluster randomised trials BMJ 2004 328 702 708 15031246 10.1136/bmj.328.7441.702
Elbourne DR Campbell MK Extending the CONSORT statement to cluster randomized trials: for discussion Stat Med 2001 20 489 496 11180315 10.1002/1097-0258(20010215)20:3<489::AID-SIM806>3.0.CO;2-S
Campbell MK Grimshaw JM Elbourne DR Intracluster correlation coefficients in cluster randomized trials: empirical insights into how they should be reported BMC Med Res Methodol 2004 4 9 15115554 10.1186/1471-2288-4-9
|
16109162
|
PMC1190183
|
CC BY
|
2021-01-04 16:32:51
|
no
|
BMC Med Res Methodol. 2005 Aug 18; 5:25
|
utf-8
|
BMC Med Res Methodol
| 2,005 |
10.1186/1471-2288-5-25
|
oa_comm
|
==== Front
BMC Musculoskelet DisordBMC Musculoskeletal Disorders1471-2474BioMed Central London 1471-2474-6-401602274110.1186/1471-2474-6-40Technical AdvanceThe clubfoot assessment protocol (CAP); description and reliability of a structured multi-level instrument for follow-up Andriesse Hanneke [email protected]ägglund Gunnar [email protected] Gun-Britt [email protected] Departments of Orthopedics, Lund University Hospital, SE-221 85 Lund, Sweden2 Departments of Health Science, Division of Physical Therapy, Lund University, Lasarettsgatan 7, SE-221 85, Sweden2005 18 7 2005 6 40 40 11 3 2005 18 7 2005 Copyright © 2005 Andriesse et al; licensee BioMed Central Ltd.2005Andriesse et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
In most clubfoot studies, the outcome instruments used are designed to evaluate classification or long-term cross-sectional results. Variables deal mainly with factors on body function/structure level. Wide scorings intervals and total sum scores increase the risk that important changes and information are not detected. Studies of the reliability, validity and responsiveness of these instruments are sparse. The lack of an instrument for longitudinal follow-up led the investigators to develop the Clubfoot Assessment Protocol (CAP).
The aim of this article is to introduce and describe the CAP and evaluate the items inter- and intra reliability in relation to patient age.
Methods
The CAP was created from 22 items divided between body function/structure (three subgroups) and activity (one subgroup) levels according to the International Classification of Function, Disability and Health (ICF). The focus is on item and subgroup development.
Two experienced examiners assessed 69 clubfeet in 48 children who had a median age of 2.1 years (range, 0 to 6.7 years). Both treated and untreated feet with different grades of severity were included. Three age groups were constructed for studying the influence of age on reliability. The intra- rater study included 32 feet in 20 children who had a median age of 2.5 years (range, 4 months to 6.8 years).
The Unweighted Kappa statistics, percentage observer agreement, and amount of categories defined how reliability was to be interpreted.
Results
The inter-rater reliability was assessed as moderate to good for all but one item. Eighteen items had kappa values > 0.40. Three items varied from 0.35 to 0.38. The mean percentage observed agreement was 82% (range, 62 to 95%). Different age groups showed sufficient agreement. Intra- rater; all items had kappa values > 0.40 [range, 0.54 to 1.00] and a mean percentage agreement of 89.5%. Categories varied from 3 to 5.
Conclusion
The CAP contains more detailed information than previous protocols. It is a multi-dimensional observer administered standardized measurement instrument with the focus on item and subgroup level. It can be used with sufficient reliability, independent of age, during the first seven years of childhood by examiners with good clinical experience.
A few items showed low reliability, partly dependent on the child's age and /or varying professional backgrounds between the examiners. These items should be interpreted with caution, until further studies have confirmed the validity and sensitivity of the instrument.
==== Body
Background
Clubfoot is a collection of different abnormalities [1-5] with different etiologies [3,6]. Consequently, severity varies with difficulties in evaluating treatment strategies and outcome results.
Most assessment instruments available for clubfoot aim towards classification or cross-sectional outcome and concentrate on variables belonging to the domains of body functions and structures [6-13]. Variables on activity and participation are sparsely used and addressed only generally [8,12-16]. Teasing, for example, is addressed in one patient based questionnaire [17]. A literature review on Medline, Libris and Elin shows that reliability and validation studies are rare [13,17-19] and regard only six out of a numerous amount of instruments described in clubfoot articles [6,15].
The International Classification of Function, Disability and Health (ICF), developed by the World Health Organization (WHO), is a classification of health and health related domains that describe body function and body structure, activity and participation [20,21]. For studies on outcome, the ICF can be used as a tool to systematically describe measures according to these domains.
The lack of an instrument that is useful during the child's growth, and follows the guidelines of the ICF, led to the development of the Clubfoot Assessment Protocol (CAP). The aims of this study were to i) describe this new instrument, ii) to investigate item inter-rater reliability between two experienced clinicians with different professional backgrounds, iii) to investigate item intra-rater reliability and iv) to investigate the influence of age on reliability.
Methods
The Clubfoot Assessment Protocol (CAP)
The purpose of the CAP is to provide an overall profile of the clubfoot child's functional status within the domains of body function/structure and activity on single assessment occasions and over time. Furthermore, the CAP aims to provide structure and standardization for follow-up procedures from 0 to11 years of age in daily clinical decision making. It is an observer administered test. The selection of important items to be included in the protocol and scoring system was an act of balance between considerations of clinical utility and scientific interest. Literature studies, expert opinions and clinical experience on what patients /parents present as important factors formed the platform for the CAP prototype.
The CAP (shown in its entirety, as used in daily practice on side 19), (Table 3) contains 22 items in four sub-groups: mobility (8 items), muscle function (3 items), morphology (4 items), and motion quality I and II (7 items). The first three sub-groups relate to body function/structures and the last to activity according to ICF-2001 [8]. Questions about pain, stiffness and daily activity /sport participation are standard. These subjective items are not included in this reliability study.
Each item is described in a manual along with the criteria for scoring. The scoring is divided systematically in proportion to what is regarded as normal variation and its supposed impact on perceived physical function ranging from 0 (severe reduction/ no capacity) to 4 (normal). Score grading can vary between 3 to 5 levels. For sub-groups the sum of the items scores are calculated and can be visualized as profiles (transformed to a 0–100 scale score, with 0 = extremely deviant and 100 within normal variance; sub-group transformation score = actual score/maximal possible score × 100). Missing item assessment is treated by submitting the average scoring for that item. The CAP is not intended for total scores.
Administration time varies between 10–15 minutes dependent on the child's cooperation. Seven items assess motion quality and are age dependent. At the age of three years all children are presumed to be able to perform Motion Quality part I. At the age of 4 all children are also expected to be able to perform Motion Quality part II. Knowledge and experience on normal child neuro-motor development is a prerequisite for enabling proper assessment of the sub-groups muscle function and movement quality.
Procedure
The reliability study took place over a four month period at routine follow-ups at the clubfoot unit and in a normal clinical setting. The project was regarded as quality control in clinical work. The children were familiar with the examiners. Parents and older children were informed about the testing procedure of the instrument and its importance in increasing the quality of our follow-up program. They were also informed that they could withdraw whenever they wanted. They all gave their consent to participate.
Two examiners, one physical therapist (HA) and one pediatric orthopedic surgeon (GH), both well acquainted with clubfoot problems, assessed consecutively and independently of each other the children in random order. Both had been participants in developing the protocol. HA had clinical experience working with the CAP. GH carefully studied the manual and the protocol before entry. After the first eight patients, the two observers consulted with each other before continuing. To enhance the stability of the phenomenon tested and to prevent the children of getting bored and tired, the examiners took turns in instructing the children while testing the items of domain "motion quality".
The intra-rater reliability test was done by HA.
Patients
In the inter- rater study, 13 girls and 35 boys born with idiopathic clubfoot, median age of 2.1 years (range, 0 to 6.7 years) were assessed. Twenty-seven children had unilateral and twenty-one had bilateral clubfoot, which gave a total of 69 assessed feet. The feet's severity spectrum in new-born ranged from very mild to very severe [10]. The feet were assessed in different phases of our treatment program. This includes intensive stretching and manipulations on a daily basis during the first 2 month after birth supplemented with an adjustable splint worn 22 hours a day. At the age of 2 month in most cases an Achilles tenotomy and posterior-medial release was needed followed by a 5 week period of casting. At the age of 4.5 month old these children's clubfeet were fully corrected and treatment continued with a special designed dynamic orthosis. In the beginning these orthosis were used 18 hours a day and later on only at night (minimum of 8 hours) until four years of age.
The children were divided into three age groups:
I. Newborn – walking debut (n = 22 feet, median age 3.2 months, range 0 to 1.1 years).
II. Walking debut – four years (n = 25 feet, median age 2.1 years, range 1.2 to 3.9 years).
III. Four years – seven years (n = 22 feet, median age 4.9 years, range 4.0 to 6.7 years).
The intra – rater portion of this study consisted of 20 children, considered to be in a clinical stable phase and a median age of 2.5 years (range, 4 months to 6.8 years). A total of 32 feet, were assessed dispersed in the three age groups as following; 8:14:10. The mean re-examination time was 2.1 months (range, 0.5 to 3.0 months).
Most missing values were seen in age group II in the sub-group motion quality, especially for heel and toe walking (12 out of 25 assessments). This was caused by immaturity in the motor development. In three cases, the child refused to co-operate with one or the other of the observers (Table 2).
Table 2 Inter-rater reliability. Unweighted Kappa values, confidence interval (CI) and overall agreement (Po)in percentage for the three age groups separately.
Age group I Age group II Age group III
Item S n Kappa (95% CI) Po % N Kappa (95%CI) Po% N Kappa(95%CI) Po %
1. Dorsiflexion 5 22 0.94 (0.81–1.00) 95 25 0.44 (0.16–0.71) 68 22 0.73 (0.5–0.96) 82
2. Plantarflexion 5 22 0.65 (0.34–0.96) 82 25 0.43 (0.14–0.71) 64 20 0.77 (0.35–1.00) 95
3. Subtalar 5 22 0.74 (0.54–0.95) 82 23 0.26 (-0.17–0.69) 68 22 0.71 (0.46–0.96) 82
4. Derotation 5 22 0.53 (0.30–0.76) 73 25 0.65 (0.02–1.00) 96 22 0.62 (0.32–0.93) 82
5. Adduktion 5 22 1.00 (0.00–0.00) 100 25 0.58 (0.23–0.92) 84 22 0.53 (0.21–0.84) 73
6. Tightness 4 22 0.66 (0.43–0.90) 77 23 0.41 (0.10–0.73) 84 21 0.85 (0.65–1.00) 90
7. Flx.dig.long 3 13 0.00 (0.00–0.00) 77 25 -0.06(-0.14–0.0)! 88 18 0.00 (0.00–0.00) ! 94
8. Flx.dig.hall. 3 13 0.00 (0.00–0.00) 85 25 0.83 (0.52–1.00) 96 18 0.00 (0.00–0.00) ! 94
9. M. Peroneus 3 22 0.58 (0.30–0.86) 73 25 0.54 (0.24–0.85) 76 22 0.57 (0.26–0.88) 72
10. M. ext.dig.ln 3 22 0.21 (0.02–0.40) 64 24 0.60 (0.21–0.99) 84 22 0.26 (-0.15–0.66) 72
11. M. sol/gastr. 3 20 0.52 (0.12–0.93) 80 21 0.00 (0.00–0.00) ! 81 22 0.00 (0.00–0.00)! 91
12. Tib.rotation 3 22 0.74 (0.40–1.00) 91 25 0.71 (0.41–1.00) 88 22 0.62 (0.16–1.00)! 91
13. Calc.pos. 3 22 0.78 (0.56–1.00) 86 21 -0.07(-0.2–0.04)! 84 22 0.43 (0.00–0.98) 86
14. Forefootpos. 3 22 0.82 (0.60–1.00) 82 21 0.63 (0.26–1.00) 86 19 0.41 (0.06–0.75) 68
15. Foot arch 3 22 1.00 100 21 0.00 (0.00–0.00) 95 21 0.43 (0.10–0.75) 81
16. Walking 4 n.a 24 0.71 (0.45–0.97) 83 22 0.64 (0.37–0.92) 77
17. Toe walking 4 n.a 12 0.60 (0.15–1.00) 83 22 0.38 (0.02–0.74) 68
18. Heelwalking 4 n.a 12 0.37 (0.06–0.68) 58 22 0.52 (0.23–0.81) 68
19. Squatting 4 n.a 18 0.50 (0.08–0.92) 78 n.a
20. Running 4 n.a 18 0.67 (0.35–0.98) 83 22 0.13 (-0.21–0.47) 45
21. One legstand 4 n.a n.a 22 0.66 (0.41–0.91) 77
22. Hop 1 leg 4 n.a n.a 22 0.94 (0.82–1.00) 95
S = amount of scales, n = number of feet assessed
! = limited variation of cell frequency
n.a. = not applicable, due to age dependent activities.
The distribution of the assessments scores were more equally spread in the age group I and for all ages together. Age group II and III had assessment shifting more to the right of the scale for the first 15 items.
Statistics
Unweighted Kappa (k) statistics for agreement were used [22-24] with 95% confidence interval. It calculates agreement beyond chance. As kappa values can become unstable under certain conditions [24,25], the observed percentage agreement (Po) was also calculated. A Po > 75% was regarded as good. In cases with limited distribution of cell frequency, the Po was preferred instead of k. The amount of categories is also regarded as kappa values decrease when categories increase [25]. The kappa has a maximum of 1 when agreement is perfect, but a value of 0 indicates no agreement better than chance, and negative values show worse than chance agreement. According to Altman [22] the kappa values are to be interpreted as follows: <0.20 as poor agreement, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as good and > 0.80 as very good agreement.
A good reliability was considered when the kappa value was high, or a low kappa value combined with a high in Po. A sufficient reliability was considered in cases with fair- moderate kappa values and good percentage agreement.
The SPSS 12.00 and StatXact (version 3) was used for the statistical analyses.
Results
Inter- rater reliability and age influence
Altogether 1196 assessments were made by each examiner.
For all children (n = 48, 69 clubfeet), 18 out of 22 items had kappa values > 0.40 (range, 0.52 to 1.00). Two items ranged from 0.35 to 0.36 but had good Po (Table 1). Item 7 had a negative kappa score caused by skewed frequency distribution but a good Po = 87%. Item 20 had a kappa value of 0.38 and Po = 62% and is assessed as fair agreement.
Table 1 Inter- and intra reliability. Unweighted Kappa values, confidence interval (CI) and overall agreement (Po) for the inter- and intra rater reliability tests and age interval 0 – 7 years.
Inter-rater Intra-rater
Item S n Kappa (95% CI)) Po (%) n Kappa (95% CI) Po (%)
1. Dorsiflexion 5 69 0.73 (0.60–0.86) 85 32 0.72 (0.50–0.94) 84
2. Plantarflexion 5 67 0.62 (0.45–0.79) 79 32 0.66 (0.41–0.91) 66
3. Subtalar 5 67 0.65 (0.49–0.79) 79 32 0.83 (0.66–1.00) 91
4. Derotation 5 69 0.62 (0.45–0.79) 84 32 0.93 (0.76–1.10) 97
5. Adduktion 5 69 0.71 (0.54–0.88) 85 32 0.94 (0.83–1.05) 97
6. Tightness 4 66 0.68 (0.53–0.83) 80 32 0.61 (0.35–0.87) 78
7. Flx.dig.long 3 56 -0.03 (-0.09–0.03) ! 87 32 0.72 (0.36–1.08) 94
8. Flx.dig.hall. 3 56 0.57 (0.20–0.94) ! 93 32 0.54 (0.27–0.81) 81
9. M. Peroneus 3 69 0.60 (0.44–0.76) 75 26 0.63 (0.34–9.92) 94
10. M. ext.dig.lgn 3 68 0.36 (0.16–0.56) 75 28 1.00 100
11. M. sol/gastr. 3 63 0.35 (0.05–0.60) 84 24 1.00 100
12. Tib.rotation 3 69 0.70 (0.50–0.90) 90 31 0.84 (0.63–1.00) 94
13. Calc.pos. 3 65 0.63 (0.43–0.83) 85 32 0.80 (0.53–1.00) 94
14. Forefootpos. 3 62 0.63 (0.44–0.82) 82 32 0.82 (0.63–1.00) 91
15. Foot arch 3 64 0.5 (0.30–0.86) ! 92 31 1.00 100
16. Walking 4 46 0.68 (0.50–0.86) 80 27 0.81 (0.63–0.99) 89
17. Toe walking 4 34 0.49 (0.21–0.77) 73 15 0.61 (0.27–0.95) 80
18. Heelwalking 4 34 0.47 (0.23–0.71) 65 15 0.79 (0.52–1.00) 87
19. Squatting 4 19 0.56 (0.18–0.94) 79 19 0.80 (0.54–1.00) 89
20. Running 4 40 0.38 (0.14–0.62) 62 27 0.78 (0.59–0.97) 85
21. One legstand 4 22 0.70 (0.45–0.95) 77 9 0.68 (0.28–1.00) 77
22. Hop 1 leg 4 22 0.94 (0.82–1.06) 95 9 0.70 (0.33–1.07) 77
S = amount of scales, n = number of feet assessed
! = limited variation of cell frequency
The two examiners agreed totally in 82% of the assessments (range, 62 to 95%). (Table 2). A one – category disagreement was seen in 17% of the cases, whereas a two-category disagreement was seen in 1 %. We conclude that all but one item had moderate to good agreement.
For age group I, 12 /15 items had kappa values > 0.40 (range, 0.52 to 1.00) (Table 2).
Items 7 and 8 had poor kappa values (kappa = 0.00) due to skewed frequencies but acceptable observer agreement (77% respectively 85%). Item10 had fair reliability with a kappa of 0.21 and Po of 64%.
For age group II, 14/ 20 items had kappa values > 0.40 (range, 0.41 to 0.83). Items 7, 11, 13 and 15 had poor values caused by limited distribution of cell frequency but good observer agreement (respectively 88%, 81%, 84% and 95%). Items 2 and 18 had kappa values of 0.26 and 0.37 respectively, and a Po of 68% and 58% respectively and are regarded as having fair reliability.
For age group III, 16/21 items had kappa values > 0.40 (range, 0.41 to 0.94). Items 7, 8, and 11 had poor kappa values also due to skewed distribution but very good observer agreement (94%, 94% and 91% respectively). Item 17 had a fair kappa value and a Po= 68 %. Item 20 had both poor kappa values and poor observer agreement (45%). Item 19 (squatting) was not assessed in this age group.
Taking into account the kappa values, the Po and amount of scales, no age group showed clearly poor reliability values for its items except for item 20, running, in age group III.
Intra – rater reliability
A total of 587 assessments were done twice. All items had kappa values > 0.40 (range, 0.54 to 1.00) (Table 1). Total agreement was reached in 89 %. A one-category disagreement was seen in 10 % and a two-category disagreement in 0.3 %.
Discussion
The CAP protocol items had moderate to very good inter-rater reliability for all the items in the age group 0–7 years and for most of the items when regarding the specific age groups.
The intra-rater test showed good to excellent reliability and indicates a good standardization of the protocol.
Most items in our protocol had moderate to excellent inter-observer reliability especially concerning sub-groups "passive mobility" and "morphology". This is a positive finding in the light of the fine-grained protocol with up to five different categories and the two observers' different professions and different experience with the protocol.
Methodological issues
Reliability studies in children are difficult to perform. The risk for errors is high as the children's co-operation and task understanding may vary from day to day and between different examiners. A child-friendly environment and familiarity with the examiners are important factors in enhancing reliability. We also wanted a situation that was comparable with a normal clinical setting where the instrument is intended to be normally used. These are the reasons why the investigation was unblinded and no more than two examiners were involved.
The fact that one of the examiners had extensive practical experience with the instrument while the other had only co-operated with the development of the protocol might have influenced the result.
In clinical practice teamwork often is the norm and therefore we chose two different professions. However Flynn et al. [18] observed in his study that including a physical therapist decreases reliability; agreement should be expected to increase if assessment is kept within the same profession.
The children available for our study represented the clubfoot spectrum [6] and illustrated the clinical development. Gender distribution corresponded well with the 3:1 (male/female) ratio normally described [3].
Statistics
When working with ordered categorical data as, in the case of our protocol, the right way of analyzing agreement is said to be Kappa [22,24]. We chose the unweighted Kappa as we wanted to know how the exact agreement would be for our finely graded instrument. It is more common though to use the weighted Kappa statistics that take into account the degree of disagreement [22]. These values are usually higher. We recalculated our kappa's to weighted and found that the values increased between 0.01 and 0.20. For example, our kappa value for the item "running" in age group III, changed from 0.13 to 0.46 when using weighted kappa statistics. This indicates that we can increase our reliability by combining categories. Within research, the finely graded protocol should be prioritized. Care should be taken when interpreting kappa statistics as the value of kappa depends upon the proportion of subjects in each category [24,26]. Haas [24] emphasizes that kappa becomes unstable under certain conditions. The problem-limited variation occurs when there is a large proportion of agreement and most of the agreement is limited to only one possible rating choice. We saw this problem for example in item 7. When all children between 0–7 years were included, untreated, treated and relapsing feet were assessed which meant that the whole scoring spectrum was used. Problems with limited distribution therefore became less. The older children generally had scores that lay more to the right on the protocol which caused a certain ceiling effect. Thus the CAP detects differences in severity which confirms part of its construct validity.
Another possibility for assessing reliability would have been to calculate statistical differences between the total sub scores for each observer, as Flynn et al. [18] did in their reliability study comparing the Pirani [13] and Dimeglio scores [11]. Another way might be to use the mean difference and calculate the 95% limits of agreement as Altman describes [22]. This could give us information on how much we can expect every new assessment to differ between new examiners and individuals and its clinical relevance.
Results
We have described the CAP; an alternative assessment tool for both short-and long-term follow-ups of children treated for clubfoot. Our protocol differs from most others through scoring grades with smaller intervals and incorporates a broader assessment on movement quality. It is also intended to be used longitudinally during the child's growth. The focus is primarily on item level and secondary on subgroup level. With sum scores and categorization/classification, important information can be lost and it should therefore be avoided [26,27]. Research profiles can be made for each item-score or subgroup(s) scores from the CAP at a certain time or over a time interval on group or individual level. In daily clinical work, the CAP is a promising tool in increasing the quality of follow -up procedures and clinical decision making through standardization and gives the possibility of a visual feedback. It also will give us the possibility to analyze factors influencing the clubfoot development.
With outcome studies, a holistic approach is of importance. The CAP should be supplemented with a patient- and parent-based questionnaire with items specifically focusing on symptoms and limitations in daily life, such as the patient- based questionnaire developed by Roye et al [17]. The Laaveg- Ponseti [12] rating system also has a score distribution emphasizing the importance of patient satisfaction and participation. Recently, several outcome measures focusing on the child's physical functioning in her or his environment, such as the Pediatric Outcomes Data Collection Instrument (PODCI) [28] and the Activity Scales for Kids (ASK) [29,30], have been developed. The use of these kinds of outcome instruments in the future will increase our knowledge of factors that are probably of more importance for patient satisfaction than range of motion, strength and radiographic changes. In the future these factors will become more and more important when discussing outcome results [10,31,32].
Face validity (whether a test appears to measure what it is supposed to measure) and content validity (the extent to which the measures represents functions or items of relevance given the purpose and matter of issue) [34] are enhanced through the developmental procedure. This is based on literature studies, discussions, clinical experience and patient information. Through clinical trial the tool was adjusted several times during the years used at the clubfoot-clinic and might be further adjusted.
Reliability for the different age groups is, with respect to the difficulties met in assessing children, within acceptable limits. Items, which demanded maturity, co-operation and task comprehension such as muscle function, are more vulnerable for different assessment results as research conditions can change between the observers. This is clearly seen in the total group for item 10, kappa value of 0.36, and item 11 kappa value 0.35.
Distinguishing differences in running quality is not easy to assess which is expressed in a low kappa value of 0.38 (fair). It is a fast movement and to observe slight variations is difficult. In our study nearly all differences lay between slightly deviant and normal.
Wainwright et al. [21] assessed the reliability of four classification systems from Catterall [9], Dimeglio et al. [11], Harrold and Walker [35] and Ponseti and Smoley [3]. These instruments are only comparable with the CAP mobility domain. Nine children (13 clubfeet) were assessed by four examiners at different stages in the first 6-months of life (= 180 examinations). The results showed kappa values varying between 0.14 and 0.77. It is not reported if the kappa is weighted or unweighted. The kappa values for our CAP-mobility items vary between 0.57 – 0.73 for ages 0–7 years and ages 0-walking debut between 0.32 – 1.00. We consider this to be positive in the light of the fine graded scales in our protocol.
Future research
Further studies on psychometric aspects are ongoing and are needed before the CAP can be used in a scientifically sound way. Changes in items used and item groupings are therefore expected.
Conclusion
The CAP contains more detailed information than previous protocols. It is a multidimensional observer-administered measurement instrument with the focus on item and subgroup level. It can be used with sufficient reliability independent of age during the first seven years of childhood by examiners with good clinical experience.
A few items showed low reliability, partly dependent on the child's age and /or varying professional backgrounds between the examiners. These items should be interpreted with caution, until further studies have confirmed the validity and sensitivity of the instrument.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
HA and GH designed the study and collected the data. HA analyzed the data and drafted the manuscript. HA and G-BJ interpreted the data. HA, GH and G-BJ revised the manuscript. All three authors read and approved the final manuscript.
Table 3 The Clubfoot Assessment Protocol (version 1.0)
Name: Date of birth:
Date of assessment: Assessment number:
Side: O Left O Right
Rating 0 1 2 3 4
Passive mobility
1. Dorsiflexion < -10° -10°-< 0° 0°- < +10° +10°- +20° >+20°
2. Plantar flexion 0°- < 10° 10°- < 20° 20°-< 30° 30°- 40° >40°
3. Varus/valgus >20°var 20°-< 10°var 10° -< 0°var 0°- neutral >0°vlg
4. Derotation >20°inv 20°-< 10°inv 10°- < 0°inv 0°- 10°evr >10°evr
5. Add/abd >20°add 20°-< 10°add 10°- < 0°add 0°- neutral >0°abd
6. Tightness + tight tight soft-tight soft
7. Flx.dig.long. + reduced reduced normal
8. Flx.dig.hall. + reduced reduced normal
Muscle function (strength)
9. M. peroneus absent/poor reduced normal
10. M. ext.dig.long absent/poor reduced normal
11. M. sol./gastr. absent/poor reduced normal
Morphology
12. Tibial rotation + inw. inw. normal
13. Calcaneus position >10 varus >0 varus <10 neutral/vlg
14. Forefoot position >20° add. >10 add. <20° add<10°
15. Foot arch + cavus cavus normal
Motion quality
I
16. Walking + deviant deviant slightly deviant normal
17. Toe walking cannot deviant slightly deviant normal
18. Heel walking cannot deviant slightly deviant normal
19. Squatting cannot deviant slightly deviant normal
20. Running + deviant deviant slightly deviant normal
II
21. One legstand cannot deviant slightly deviant normal
22. Hop1leg cannot deviant slightly deviant normal
Extra notes: Structured questions about pain, stiffness, shoe problems, physical condition, activity level, sports and social participation and patient/parent satisfaction.
+ = pronounced / very, var= varus, vlg= valgus, inv = inversion, evr = eversion, add = adduction, abd = abduction. inw = inward rotation, flx.dig.long. = length of M. flexor digiti longus, flx.dig.hall. = length of M. flexor digiti hallucis
©The CAP is copyright to Hanneke Andriesse, 2003.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This study was supported by Lund University, Vårdrådet and Lund University Hospital, Skåne Region in the county of Skåne.
Special thanks to Per-Erik Isberg at the Department of Statistics, Lund University, for statistical advice.
==== Refs
Bensahel H Huguenin O Themar-Noel C The functional anatomy of clubfoot J Pediatr Ortop 1983 3 191 5
Fukuhara K Schollmeier G Uhthoff HK The pathogenesis of clubfoot J Bone Joint Surg Br 1994 76 450 57 8175852
Ponseti IV Congenital clubfoot 1996 Oxford: University Press
Sirca A Erzen J Pecal F Histochemistry of abductor hallucis muscle in children with idiopathic clubfoot and controls J Pediatr Orthop 1990 10 477 482 2358485
Zimny ML Willig SJ Roberts JM DïAmbrosia RD An electron microscopic study of the fascia from the medial and lateral sides of clubfoot J Pediatr Ortop 1985 5 577 81 4044817
Simon GW The clubfoot 1993 New York: Springer Verlag
Atar D Lehman WB Grant AD Strongwater A Functional rating system for evaluating the results of clubfoot surgery Orthop Rev 1990 19 730 35 2216546
Bensahel H Dimeglio A Souchet P Final evaluation of clubfoot J Pediatr Orthop B 1995 4 137 41 7670980
Catterall A A method of assessment of the clubfoot deformity Clin Orthop Rel Res 1991 264 48 53
Cohen-Sobel E Caselli M Giorgini R Giorgini T Stummer S Longterm follow- up of clubfoot surgery: analysis of 44 patients J Foot Ankle Surg 1993 32 411 23 8251997
Dimeglio A Bensahel H Souchet P Bonnet T Classification of clubfoot J Pediatr Orthop 1995 4 129 36
Laaveg SJ Ponseti IV Longterm results of treatment of congenital clubfoot J Bone Joint Surg Am 1980 62 23 31 7351412
Pirani S Outerbridge H Moran M Sawatsky BJ A method of evaluating the virgin clubfoot with substantial inter-observer reliability POSNA 1995 71 Miami, Florida 99
Mc Kay DW New concepts of and approach to clubfoot treatment. Section III. Evaluation and results J Pediatr Orthop 1983 3 141 148 6863519
Uglow MG Clarke NMP The functional outcome of staged surgery for the correction of talipes equinovarus J Pediatr Orthop 2000 20 517 523 10912611 10.1097/00004694-200007000-00018
Harvey AR Michael GU Clarke NMP Clinical and functional outcome of relapse surgery in severe congenital talipes equinovarus J Pediatr Orthop B 2003 12 49 55 12488772 10.1097/00009957-200301000-00009
Roye BD Vitale MG Gelijns AC Roye DP Jr Patient-based outcomes after clubfoot surgery J Pediatr Orthop 2001 21 42 49 11176352 10.1097/00004694-200101000-00010
Flynn JM Donohoe M Mackenzie WG An independent assessment of two clubfoot–classification systems J Pediatr Orthop 1998 18 323 27 9600557 10.1097/00004694-199805000-00010
Wainwright AM Auld T Benson MK Theologis TN The classification of congenital talipes equinovarus, J Bone Joint Surg Br 2002 84 1020 4 12358365 10.1302/0301-620X.84B7.12909
International classification of functioning, disability and health. Geneva: World Health Organisation accessed; 2004/9/23
World Health Organization Classification, Assessment, Surveys and Terminology Team. ICIDH 2000 Geneva Switserland. Prefinal draft Fullversion
Altman DG Practical statistics for medical research 1997 London: Chapman & Hall 401 9
Cohen J Weighted kappa Nominal scale agreement with provision for scaled disagreement or partial credit Psychological Bulletin 1968 70 213 20
Haas M Statistical methodology for reliability studies J Manipulative Physiol Ther 1991 14 119 32 2019821
Kundel HL Polansky M Measurement of observer agreement Radiology 2003 228 303 308 12819342
Roos E Roos H Lohmander LS Ekdahl C Beynnon BD Knee Injury and Osteoarthritis Outcome Score(KOOS)- Development of a Self- Administered Outcome Measure J Orthop Sports Phys Ther 1998 28 88 96 9699158
Sgaglione NA Del Pizzo W Fox JM Friedman MJ Critical analysis of knee ligamnet rating systems Am J Sports Med 1995 23 660 667 8600730
Daltroy LH Liang MH Fossel AH Goldberg MJ The POSNA pediatric musculoskeletal functional health questionnaire: report on reliability, validity and sensitivity to change J Pediatr Orthop 1998 18 561 571 9746401 10.1097/00004694-199809000-00001
Plint AC Caboury I Owen J Young NL Activities Scale for Kids, an analysis of normals J Pediatr Orthop 2003 23 788 790 14581784
Young NL Williams JI Yoshida KK Wright JG Measurement properties of the Activities Scale for Kids J Clin Epidemiol 2000 53 125 137 10729684 10.1016/S0895-4356(99)00113-4
Campos da Paz A Ramalho A Momura A Braga L Almeida M Simon GE Gait analysis in clubfoot: An experimental study The clubfoot 1993 New York: Springer Verlag 81 3
Rejeski WJ Martin KA Miller ME Ettinger WH JrRapp S Perceived importance and satisfaction with physical function in patients with knee osteoarthritis Ann Behav Med 1998 20 141 8 Spring 9989320
Katula JA Rejeski WJ Wickley KL Berry MJ Perceived difficulty, importance, and satisfaction with physical function in COPD patients Health Qual Life Outcomes 2004 2 18 15056389 10.1186/1477-7525-2-18
Johnston MV Keith RA Hinderer SR Measurement standards for interdisciplinary medical rehabilitation Arch Phys Med Rehabil 1992 73 S3 S23 1463386
Harrold AJ Walker CJ Treatment and prognosis in congenital clubfoot J Bone Joint Surg Br 1983 65 8 11 6822607
Ponseti IV Smoley EN Congenital clubfoot: the results of treatment J Bone Joint Surg (Am) 1963 45-A 261 344
|
16022741
|
PMC1190184
|
CC BY
|
2021-01-04 16:32:04
|
no
|
BMC Musculoskelet Disord. 2005 Jul 18; 6:40
|
utf-8
|
BMC Musculoskelet Disord
| 2,005 |
10.1186/1471-2474-6-40
|
oa_comm
|
==== Front
BMC NeurolBMC Neurology1471-2377BioMed Central London 1471-2377-5-141602661510.1186/1471-2377-5-14Research ArticleObservations on comatose survivors of cardiopulmonary resuscitation with generalized myoclonus Thömke Frank [email protected] Jürgen J [email protected] Oliver [email protected] Thomas [email protected]ägele Stefan [email protected] Jascha [email protected] Sacha L [email protected] Department of Neurology, Johannes Gutenberg-Universität, Langenbeckstrasse 1, D- 55101 Mainz, Germany2 Internal Medicine II, Johannes Gutenberg-Universität, Langenbeckstrasse 1, D- 55101 Mainz, Germany2005 18 7 2005 5 14 14 30 3 2005 18 7 2005 Copyright © 2005 Thömke et al; licensee BioMed Central Ltd.2005Thömke et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
There is only limited data on improvements of critical medical care is resulting in a better outcome of comatose survivors of cardiopulmonary resuscitation (CPR) with generalized myoclonus. There is also a paucity of data on the temporal dynamics of electroenephalographic (EEG) abnormalities in these patients.
Methods
Serial EEG examinations were done in 50 comatose survivors of CPR with generalized myoclonus seen over an 8 years period.
Results
Generalized myoclonus occurred within 24 hours after CPR. It was associated with burst-suppression EEG (n = 42), continuous generalized epileptiform discharges (n = 5), alpha-coma-EEG (n = 52), and low amplitude (10 μV <) recording (n = 1). Except in 3 patients, these EEG-patterns were followed by another of these always nonreactive patterns within one day, mainly alpha-coma-EEG (n = 10) and continuous generalized epileptiform discharges (n = 9). Serial recordings disclosed a variety of EEG-sequences composed of these EEG-patterns, finally leading to isoelectric or flat recordings. Forty-five patients died within 2 weeks, 5 patients survived and remained in a permanent vegetative state.
Conclusion
Generalized myoclonus in comatose survivors of CPR still implies a poor outcome despite advances in critical care medicine. Anticonvulsive drugs are usually ineffective. All postanoxic EEG-patterns are transient and followed by a variety of EEG sequences composed of different EEG patterns, each of which is recognized as an unfavourable sign. Different EEG-patterns in anoxic encephalopathy may reflect different forms of neocortical dysfunction, which occur at different stages of a dynamic process finally leading to severe neuronal loss.
==== Body
Background
Comatose survivors of cardiopulmonary resuscitation (CPR) developing generalized myoclonus have a poor prognosis. Most of these patient die and those surviving the acute stage almost always remain in a persistent vegetative state [1-8]. There are recent reports of single patients with myoclonus who have a good outcome [9,10], suggesting poor prognosis is not invariably the case. There is only limited data on whether advances in critical care medicine are associated with a better outcome. We address this issue on the basis of 50 comatose survivors of CPR with generalized myoclonus, whom we examined and treated during the past 8 years. We also report the results of serial electroencephalographic (EEG) recordings, there being only limited data on the temporal dynamics of EEG abnormalities in the acute stage after CPR. Finally, we discuss the relation of generalized myoclonus and status epilepticus, since terms, such as "myoclonic status epilepticus" [4,5], "generalized status myoclonicus" [2], and "myoclonus status" [6], imply that comatose survivors of CPR with generalized myoclonus suffer from a severe form of convulsive status epilepticus.
Methods
Over an 8-year-period (between winter 1996/1997 and winter 2004/2005), we observed 50 consecutive patients, who developed generalized myoclonus within 24 hours after CPR. There were 24 women, aged 26 to 83 years, mean age: 55 years; and 26 men, aged 20 to 79 years, mean age 53 years. Forty-five patients were resuscitated outside and the remaining 5 inside the hospital. Forty-four patients had cardiac arrest or ventricular fibrillation, 4 acute respiratory failure, and 2 circulatory collapse due to gastrointestinal bleeding.
All patients had bipolar 8-channel EEG recordings with needle electrodes positioned according to the standard 10–20 system of electrode placement (Fp2-T4, T4-O2, Fp2-C4, C4-O2, Fp1-T3, T3-O1, Fp1-C3, C3-O2.). Filter setting was 0.53 Hz and 70 Hz. The first EEG was done 6 to 24 hours after CPR in all 50 patients. At that time, all had generalized myoclonus and none were on sufficiently high doses of drugs that would produce a burst-suppression pattern: Approximately 0.2 mg midazolam/kg body weight and 0.5 mg fentanyl were usually given when the patients were admitted to the intensive care unit. Subsequent EEGs were recorded in survivors at day 2 (42 patients), at day 3 or 4 (29 patients), at day 5 or 6 (15 patients), at 7 or 8 (12 patients), and during the 2nd week (4 patients). Several medications (phenytoin, valproic acid, diazepam, clonanzepam, lorazepam, midazolam, propofol) were used in an attempt to suppress generalized myoclonus after the first EEG. EEGs at day 2 were usually done after the administration of one these drugs given in doses usually used for convulsive status epilepticus, i.e. phenytoin: 1500 mg within 30 to 60 minutes; valproate: 1600 to 3200 mg within 30 minutes; diazepam up to 40 mg, clonazepam or lorazepam up to 8 mg, or midazolam up to 15 mg.
Diagnosis of a burst-suppression EEG (BS-EEG) was based on the Guidelines of the International Federation of Clinical Neurophysiology: "bursts of theta and/or delta waves, at times intermixed with faster waves, and intervening periods of low amplitude (below 20 μV)" [11], and on Niedermeyer and Lopes da Silva's Electroencephalography: "high-voltage bursts of slow waves with intermingled sharp transients or spikes occur against a depressed background or complete flatness" [12]. In accordance with others [13-15], we fixed the duration of isoelectric or low amplitude interburst intervals to at least 1 second to exclude patients with generalized continuous epileptiform discharges. Flat recordings were those with amplitudes below 20 μV, and isoelectric recordings those without any detectable activity (sensitivity of the recording system: 2 μV/mm). Serial determinations of serum neuron-specific enolase (NSE) were done one, two and three days after CPR in 27 of the 50 patients.
Results
All patients with generalized myoclonus were comatose, needed mechanical ventilation, and had loss of some brainstem reflexes. Myoclonus was highly variable ranging from single myoclonic jerks to nearly continuous myoclonus. It was generalized in approximately two thirds of patients and multifocal in one third. In all patients, myoclonus involved the facial muscles, more often associated with bilateral eye closure than with bilateral eye opening. Occasionally, myoclonus was restricted to facial muscles and manifested as eye or jaw opening. Most patients also had myoclonus of the shoulder and proximal arm muscles and the diaphragm. Involvement of the legs occurred less often but was seen in approximately half of the patients. Myoclonus almost always increased or was triggered by acoustic stimuli, touch, and tracheal suctioning.
In general, intravenous phenytoin, valproate, or various benzodiazepines were ineffective when given in doses usually used in patients with convulsive status epilepticus, i.e. phenytoin: 1500 mg over 30 to 60 minutes; valproate: 1600 to 3200 mg over 30 minutes; diazepam up to 40 mg, clonanzepam or lorazepam up to 8 mg, or midazolam up to 15 mg. Intravenous propofol (100 to 300 mg) was given to the last 7 patients after the recording of the 2nd EEG. This was always followed by a flat (below 10 μV) EEG recordings and cessation of generalized myoclonus, the cessation persisting during continuous propofol infusion (150 to 250 mg/h). Except in one patient whose generalized myoclonus persisted with decreasing intensity until her death 9 days after CPR, generalized myoclonus usually ceased within 1 to 2, but occasionally 3 days.
In 42 patients, generalized myoclonus was associated with a BS-EEG (Figure 1), which, in 12 patients was interrupted by trains of continuous epileptiform discharges for 10 to 55 seconds (Figure 2). Burst and trains of continuous epileptiform discharges were usually associated with generalized myoclonus. Sometimes bursts of activity without any visible myoclonus were recorded and at other times myoclonic jerks were seen without associated bursts. In the remaining 8 patients, generalized myoclonus was associated with continuous generalized epileptiform discharges (n = 5), alpha-coma-EEG (n = 2), and a low amplitude (below 10 μV) recording (n = 1), i.e. without associated EEG bursts. Except in one patient, who had a BS-EEG until her death 9 days after CPR, BS-EEG and any other EEG-pattern were only a transient phenomenon always followed within one day by another nonreactive EEG pattern (Figure 3). The most frequent nonreactive EEG-patterns on the 2nd day were alpha-coma-EEG (with or without some theta activity) (n = 10) (Figure 4) generalized continuous epileptiform discharges (n = 9), (Figure 5), theta EEG (n = 5), or isoelectric recordings (accompanied by clinical signs of brain death, n = 5). Subsequent recordings disclosed highly variable EEG sequences characterized by transitions between these patterns including reappearance of a BS-EEG and, finally, a low amplitude or isoelectric recording (Figures 3 and 6). Transitions between different EEG patterns were often subtle, especially transitions between between BS-EEG and continuous epileptiform discharges or between alpha-and theta-coma-EEG (Figure 7). Coexistence of different EEG-patterns in the same recording was seen in 18 patients, mainly BS-EEG with trains of generalized continuous epileptiform discharges (between 10 and 55 seconds) in 12 patients (Figure 2). Other coexisting patterns included alpha-coma-EEG with stretches of epileptiform discharges (up to 10 s) (n = 1) or with intervening periods of low amplitude (for 2 s) (n = 1), generalized continuous epileptiform discharges with trains of alpha-theta-acticity (for 2–5 s) (n = 2), theta-EEG with bursts of epileptiform discharges (for 1–3 s) (n = 1), BS-EEG with trains of alpha-theta-activity (for 2–4 s) (n = 1), and BS-EEG with episodes of alpha (theta) activity and periods of generalized continuous epileptiform discharges (n = 1).
Figure 1 Three examples of a burst-suppression-EEG of 3 comatose survivors with generalized myoclonus within 24 hours after cardiopulmonary resuscitation.
Figure 2 Burst-suppression-EEG (upper recording) with periods of continuous epileptiform discharges (lower recording) in a comatose survivors of cardiopulmonary resuscitation with generalized myoclonus.
Figure 3 EEG-sequences and outcome in 50 comatose survivors of cardiopulmonary resuscitation with generalized myoclonus.
Figure 4 Examples of alpha-coma-EEGs on the 2nd day after cardiopulmonary resuscitation in 2 comatose survivors with burst-suppression EEG on the preceding day.
Figure 5 Examples of continuous epileptiform discharges on the 2nd day after cardiopulmonary resuscitation in 2 comatose survivors with burst-suppression EEG on the preceding day.
Figure 6 EEG-sequence of a comatose survivors of cardiopulmonary resuscitation with a burst-suppression-EEG on the 1st day, which was followed by an alpha-coma-EEG on the 2nd and an isolelectric recording on the 4th day.
Figure 7 "Transitional"-EEG-pattern between burst-suppression-EEG and continuous epileptiform discharges (upper recording) or between alpha-and theta-coma EEG (lower recording).
Serum NSE was elevated in all 27 patients with maximum values between 36 und 540 ng/ml (upper normal limit in our laboratory: 17 ng/ml). Fifteen patients died within 24 hours and another 9 within 4 days after the resuscitation. Determinations regarding the level of care to be provided were made 3 to 4 days after CPR in 26 patients surviving up to this point. If clinical examination, EEG, and serum-NSE levels indicated a poor prognosis, treatment was restricted to mechanical ventilation and intravenous fluids. At the time of this decision, each patient had (a) no motor response on painful stimuli, (b) loss of the vestibular-ocular reflex, (c) NSE-levels between 67 and 540 ng/ml, and (d) three recordings of an unfavourable EEG pattern, i.e. burst-suppression EEG, continuous epileptiform discharges, or alpha-coma-EEG. Moreover, about one third also had bilateral absence of the pupil light reaction and/or bilateral absence of the corneal reflex. Twenty-one of these patients died between the 5th and 12th days after the resuscitation. None needed mechanical ventilation for more than 7 day. The remaining five patients survived in a persistent vegetative state.
Discussion
Two different forms of myoclonus may occur in patients with hypoxic injury of the brain. An acute form, the focus of the present paper, is different from the chronic form. The chronic form is seen in patients who regain consciousness and have the myoclonus begin days to weeks after CPR. Chronic hypoxic myoclonus is an action myoclonus, which only occurs when the conscious patient moves his arm or leg and is restricted to the extremity being moved. This myoclonus is thought to be generated by the cerebral cortex and probably reflects impairment of serotonergic transmission [16]. These patients often have only a mild or moderate intellectual impairment and most also suffer from cerebellar ataxia [17-19]. For reasons of clarity, only these patients should be classified as "Lance-Adams syndrome".
The acute form seen in our patients occurs in comatose survivors within one day after CPR. It is characterized by generalized myoclonic jerks which occur spontaneously and increase with acoustic or tactile stimuli. This type of myoclonus is thought to be generated by brainstem structures, the hypoxic damage of the neocortex likely being too severe to generate myoclonus [16]. Generalized myoclonus in comatose survivors of CPR is a transient phenomenon and implies a poor prognosis. The majority of our patients died during the acute stage in less than 2 weeks, and the few survivors remained in a permanent vegetative state. This is in agreement with previous studies [1-8].
Improvement in the critical care of patients has not resulted in a better outcome (Table 1). All pathological studies in such patients disclosed severe and extensive neuronal loss in the cerebral cortex, basal ganglia, thalamus, cerebellar cortex and – when examined – also in the spinal cord [2,4-6]. Extensive neuronal loss is also reflected by the regular elevation of serum NSE in our patients, which, as an isolated phenomenon, is known to indicate a poor prognosis [20-25].
Table 1 Outcome of comatose survivors of cardiopulmonary resuscitation with generalized myoclonus in previous and the present series
Study No. of patients died persistent vegetative state complete recovery
Butenuth and Kubicki 1971 [1] 12 12
Celesia et al. 1988 [2] 13 8 4 1
Krumholz et al. 1988 [3] 19 19
Jumao-as and Brenner 1990 [4] 11 11
Young et al. 1990 [5] 15 15
Wijdicks et al. 1994 [6] 40 40
Reeves et al. 1997 [8] 9 9
this series 50 45 5
There are, however, reports on individual patients with a good outcome despite generalized myoclonus after CPR, but these reports leave some questions. Celesia et al. [2] mentioned complete recovery in one of 13 patients with generalized myoclonus, but their paper lacks any additional information, especially EEG-data. We were able to find another four more recently reported patients, all with status asthmaticus as the cause of cerebral hypoxia [9,10]. In these patients, myoclonus often occurred with reduction of benzodiazepines, and responded in all patients to anticonvulsant drugs. This would be very unusual for generalized myoclonus due to severe hypoxic injury of the brain and raises the question whether these patients may have suffered from epileptic myoclonus, i.e. myoclonic jerks as a symptom of epileptic seizures. As none of these patients had EEG recordings, this possibility cannot be excluded. At least in one of these patients, additional generalized tonic-clonic seizures were reported (case 1 in the paper of Morris et al. [10].
Epileptic myoclonus or convulsive status epilepticus need to be considered in another patient with anoxic encephalopathy due to ventricular fibrillation reported by Mori et al. [26]. This patient suffered from "uncontrollable generalized tonic-clonic convulsion" for about 24 hours followed by transient eye opening, which was synchronized with a burst-suppression EEG. The latter was controlled by 50 mg intravenous diazepam, an unusual finding in patients with generalized postanoxic myoclonus. This patients survived with "neurologic resisuduals, ie, parkinsinism, dementia, and seizures".
Both, Arnoldus and Lammers [9] and Morris et al. [10], cited another two previously reported patients from Harper and Wilkes [27] as examples of a good outcome despite generalized myoclonus. These patients, however, had action myoclonus. The fact, that both patients were no longer comatose after finishing sedation with benzodiazepines had escaped observation, and the correct diagnosis of multifocal action myoclonus was determined during the course of the illness [27]. Obviously, Harper and Wilkes were aware of the correct diagnosis as they entitled their paper "Posthypoxic myoclonus (the Lance-Adams syndrome) in the intensive care unit" [27], whereas Arnoldus and Lammers [9] and Morris et al. [10] likely misinterpreted clinical findings in these patients. Thus, we are not aware of any comatose survivor of CPR with generalized myoclonus and a good outcome when accompanied by a BS-EEG or generalized continuous epileptiform discharges.
As in to previous series, generalized myoclonus in comatose survivors of CPR was mainly associated with a BS-EEG, and occasionally with generalized continuous epileptiform discharges, alpha-coma-EEG or flat recordings [1,3,4,6-8]. Each of these EEG-patterns is recognized as an unfavourable sign [1-7,13,15,28-35]. Our data indicate that each of these EEG-pattern is a transient phenomenon in comatose survivors of CPR with generalized myoclonus. It is followed by another transient EEG-pattern from this group. The temporal EEG-dynamics after CPR is characterized by a variety of variable EEG-sequences composed of these unfavourable patterns, and finally resulting in isoelectric or flat recordings (figures 3 and 6). This confirms our own findings in a previously reported smaller group of patients with postanoxic BS-EEG [8]. Our more recent findings indicate the existence of similar temporal dynamics in comatose survivors of CPR with other, less frequently seen EEG-patterns (Figure 3). These data correspond to previous observations of Wijdicks et al. [6], who performed one repeat-EEG in nine patients with BS-EEG. In six patients, BS-EEG persisted, and three had transition to alpha-coma-EEG. They also performed repeated EEGs in one patient disclosing frequent alternations between BS-EEG and an alpha coma pattern [6]. This paper, however, lacked information on the time interval between BS-EEG and repeat-EEG, and we are not aware of any other study dealing with the temporal dynamics of postanoxic EEGs.
Nineteen of our 50 patients also showed coexistence of different unfavourable EEG-patterns in the same recording, mainly BS-EEG with trains of continuous epileptiform discharges (Figure 2), and occasionally alpha-coma-EEG with trains of epileptiform discharges or intervening periods of low amplitude, theta-coma-EEG with bursts of epileptiform discharges, and BS-EEG with episodes of alpha-theta-activity. Such a coexistence seems to be a frequent phenomenon occurring more often than suggested by previous observations in single patients with transitions between alpha-and theta-coma-EEG, and between BS-EEG and alpha-coma-EEG [6,36-40].
Different EEG-patterns and EEG-sequences in comatose survivors of CPR with generalized myoclonus probably reflect different forms of dysfunction of severely damaged neocortical neurons, and occurring at different stages of a dynamic process, finally leading to severe neuronal loss. BS-EEG and alpha-coma-EEG are the most frequent patterns and have been the subject of discussion in the literature. Alpha-coma-EEG was attributed to a deafferentiation of cerebral neurons with the amygdaloid nuclei (and other subcortical structures) functioning as the final pacemakers of electric brain activity [34]. This assumption implies a residual function of dying neocortical neurons to generate alpha-(and theta-)activity under the influence of the amygdaloid nuclei (or other subcortical structures) for a short time span. Deafferentiation of cortical neurons from cortical grey matter in patients undergoing frontal lobotomy may also cause a BS-pattern [41,42]. Although deafferentiation of morphologically intact cerebral cortex is different from neocortical neuronal damage in anoxic encephalopathy, there may be severe and permanent functional undercutting of the cortex with diffuse deafferentiation of cortical neurons after prolonged cardiopulmonary arrest [43]. This may cause disinhibition of certain neocortical neurons or complex neuronal networks, creating increased excitation. More recent findings also indicate that disconnection of brain circuits may be involved in the generation of BS-EEG [43]. These assumptions also imply some kind of preserved excitability of the dying cerebral neurons to generate bursts of activity.
The mechanism connecting BS-EEG with generalized myoclonus in comatose survivors of CPR has also been discussed in the literature. Since myoclonus and bursts occur fairly synchronously, it was suggested that excitation of excitation of motoneurons in the brainstem and spinal cord occurs as in epileptic seizures [7,43]. This would imply a coexistence of severe deafferentiation of neocortical neurons as a possible mechanism of BS-EEG with intact efferent pathways projecting tom brainstem and spinal cord motoneurons. There is, however, no satisfactory answer regarding whether severe deafferentiation can really coexist with still functioning efferent pathways (43). A number of previous observations as well as our own document bursts of EEG activity without any visible movements or the presence of myoclonic jerks without associated bursts [7,45], BS-EEGs with complex sequences of limb movements [7], or eye movements [46], or tonic posturing during the suppression phase [45]. This has led to the assumption that some of the motor phenomena accompanying BS-EEG are caused by a release of brainstem circuits. All in all, it seems, that different motor phenomena accompanying BS-EEG may be due to different mechanisms [7].
Generalized myoclonus in comatose survivors of CPR, also called "myoclonic status epilepticus" [4,5] or "generalized status myoclonicus" [2], was repeatedly discussed in association with convulsive status epilepticus, and classified as a subgroup of convulsive status epilepticus with a poor prognosis [47,48]. Comatose surviviors of CPR with generalized myoclonus, however, differ in some important aspects from patients with convulsive status epilepticus:
(a) Anticonvulsant drugs commonly used in the treatment of convulsive status epilepticus, i.e. intravenous phenytoin, valproate, or benzodiazepine, are usually ineffective in acute posthypoxic myoclconus [2-7,49], personal observation].
(b) The control of generalized myclonus by propofol, previously reported in single patients [49,50] and seen in seven of our patients, did not improve the prognosis of these patients although it resulted in the cessation of epileptiform discharges.
(c) EEG-patterns in patients with generalized myoclonus, especially BS-EEG and alpha-coma-EEG, differ fundamentally from those seen in convulsive status epilepticus [1-8].
(d) Posthypoxic generalized myoclonus is a self-limited phenomenon usually ceasing within 1 to 2 days and associated with severe and extensive neocortical neuronal loss [2,4-6]. Although convulsive status epilepticus – at least lethal status – may also cause neuronal loss, these anatomical alterations defects have a largely different distribution being maximal in the hippocampus, and are less severe then in anoxic encephalopathy [51]. In general, convulsive status epilepticus is not associated with severe neuronal loss and most survivors of one episode are without sequelae that impair cerebral function.
Conclusion
Generalized myoclonus in comatose survivors after CPR still implies a poor prognosis despite improvement of the critical care of patients. It is a transient, self-limited phenomenon and reflects dysfunction of lethally damaged neurons [5]. As generalized myoclonus can be very disconcerting for relatives and staff, we recommend propofol, which often controls myoclonus (without improving the poor prognosis of these patients). Clinical, pathological, and EEG findings strongly indicate, that these patients die from severe hyoxic-ischemic damage rather than as a result of a "myoclonic status epilepticus". Abnormal EEG-patterns seen in these patients, especially BS-EEG and alpha-coma-EEG, may also be attributed to different forms of dysfunction of severely damaged neurons occurring at different stages of a dynamic process, finally leading to severe and extensive neuronal loss.
We determine the level of care to be provided 3 days after CPR. In patients with (i) unfavourable clinical signs (bilateral absence of the pupil light reaction and/or bilateral absence of the corneal reflex and/or loss of the vestibular-ocular reflex and/or no motor response on painful stimuli), (ii) an unfavourable EEG pattern (BS-EEG, generalized epileptiform discharges, alpha-coma-EEG), and (iii) elevated NSE serum levels, we restrict treatment to mechanical ventilation and intravenous fluids until they regain spontaneous breathing or die.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
FT conceived the study, examined all the patients, performed most of the EEG-recordings, participated in the analysis and interpretation of data, and drafted the manuscript
JJM participated in the analysis and interpretation of data, performed some EEG recordings, and helped to draft the manuscript
OS participated in the design of the study and in analysis and interpretation of data
TH participated in the analysis and interpretation of data and performed some of the EEG recordings
SH performed some of the EEG recordings and participated in the interpretation of data
JW participated in the design of the study and helped to draft the manuscript
SLW participated in the design of the study, helped to analyse the data and to draft the manuscript; and revised the manuscript critically
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank Ludwig Gutmann from Morgantown, West Virginia, for his careful review of the manuscript and for editing the English text. We owe him some beers. We also thank J.M. Guérit and B. Young for their comments during the review process, which gave us the opportunity to clarify some ambiguities of our paper.
==== Refs
Butenuth J Kubicki S Über die prognostische Bedeutung bestimmter Formen von Myoklonien und korrespondierender EEG-Muster nach Hypoxien Z EEG-EMG 1971 2 78 83
Celesia GG Grigg MM Ross E Generalized status myoclonicus in acute anoxic and toxic-metabolic encephalopathies Arch Neurol 1988 45 781 784 3390032
Krumholz A Stern BJ Weiss HD Outcome from coma after cardiopulmonary resuscitation: Relation to seizures and myoclonus Neurology 1988 38 401 405 3347343
Jumao-as A Brenner RP Myoclonic status epilepticus: A clinical and electroencephalographic study Neurology 1990 40 1199 1201 2116605
Young GB Gilbert JJ Zochodne DW The significance of myoclonic status epilepticus in postanoxic coma Neurology 1990 40 1843 1848 2123307
Wijdicks EFM Parisi J Sharbrough FW Prognostic value of myoclonic status in comatose survivors of cardiac arrest Ann Neurol 1994 35 239 243 8109907 10.1002/ana.410350219
Reeves AL Westmoreland BF Klass DW Clinical accompaniments of the burst-suppression EEG pattern J Clin Neurophysiol 1997 14 150 153 9165410 10.1097/00004691-199703000-00008
Thömke F Brand A Weilemann SL The temporal dynamics of postanoxic burst-suppression EEG J Clin Neurophysiol 2002 19 24 31 11896349 10.1097/00004691-200201000-00003
Arnoldus EPJ Lammers GJ Postanoxic coma: Good recovery despite myoclonus status Ann Neurol 1995 38 697 7574474 10.1002/ana.410380427
Morris HR Howard RS Brown P Early myoclonic status and outcome after cardiorespiratory arrest J Neurol Neurosurg Psychiatry 1998 64 267 268 9489546
Noachtar S Binnie C Ebersole J Mauguière F Sakamoto A Westmoreland B A glossary of terms commonly used by clinical electroencephalographers and proposal for the report from for the EEG findings Electroenceph clin Neurophysiol 1999 21 41
Bauer G Niedermeyer E, Lopes da Silva F Coma and brain death Electroencephalography: Basic Principles, Clinical Applications, and Related Fields 1999 Baltimore, Williams & Wilkins 459 475
Møller M Holm B Sindrup E Nielsen BL Electroencephalographic prediction of anoxic brain damage after resuscitation from cardiac arrest in patients with acute myocardial infarction Acta Med Scand 1978 203 31 37 626111
Synek VM Prognostically important EEG coma patterns in diffuse anoxic and traumatic encephalopathies in adults J Clin Neurophysiol 1988 5 161 174 3074973
Beydoun A Yen CE Drury I Variance of interburst intervals in burst suppression Electroenceph clin Neurophysiol 1991 79 435 439 1721570 10.1016/0013-4694(91)90162-W
Halett M Physiology of human posthypoxic myocluns Movement Disorders 2000 15 8 13 10755266
Lance JW Adams RD The syndrome of intention or action myoclonus as a sequel to hypoxic encephalopathy Brain 1963 86 111 136 13928398
Werhahn KJ Brown P Thompson PD Marsden CD The clinical features and prognosis of chronic posthypoxic myoclonus Movement Disorders 1997 12 216 220 9087980 10.1002/mds.870120212
Frucht S Fahn S The clinical spectrum of posthypoxic myoclonus Movement Disorders 2000 15 2 7 10755265
Roine RO Somer H Kaste M Viinikka L Karonen SL Neurological outcome after out-of-hospital cardiac arrest. Prediction by cerebrospinal fluid enzyme analysis Arch Neurol 1989 46 753 756 2742544
Prange HW Aue G Frauendorf H Reiber H Die neuronenspezifische Enolase als Prognosemarker bei zerebraler Hypoxie Intensivmed 1995 32 17 22
Schreiner AJ Pohlmann-Eden B Neuron-specific enolase and Protein-S-100b – Reliable markers for the enurologic outcome after hypoxic brain injury in humans Neurology 1996 46 A306
Fogel W Krieger D Veith M Adams HP Hund E Storch-Hagenlocher B Buggle F Mathias D Hacke W Serum neuron-specific enolase as early predictor of outcome after cardiac arrest Crit Care Med 1997 25 1133 8 9233737 10.1097/00003246-199707000-00012
Martens P Raabe A Johnsson P Serum S-100 and neuron-specific enolase for prediction of regaining consciousness after global cerebral ischemia Stroke 1998 29 2363 2366 9804649
Schoerkhuber W Kittler H Sterz F Behringer W Holzer M Frossard M Spitzauer S Laggner AN Time course of serum neuron-specific enolase: A predictor of neurological outcome in patients resuscitated from cardiac arrest Stroke 1999 30 1598 1603 10436107
Mori E Yamadori A Tsuruta H Ogawa MoriK Transient eye opening with EEG suppression-burst pattern in postanoxic encephalopathy Arch Neurol 1983 40 189 190 6830466
Harper SJ Wilkes RG Posthypoxic myoclonus (the Lance-Adams syndrome) in the intensive care unit Anaesthesia 1991 46 199 201 2014897
Binnie CD Prior PF Lloyd DSL Scott DF Margerison JH Eelectroencephalographic prediction of fatal anoxic brain damage after resuscitation from cardiac arrest Br Med J 1970 4 265 268 5475845
McCarthy GE Marshall DW Transient eyelid opening associated with postanoxic EEG burst-suppression pattern Arch Neurol 1981 38 754 756 7316841
Scollo-Lavizzari G Bassetti C Prognostic value of EEG in post-anoxic coma after cardiac arrest Eur Neurol 1987 26 161 170 3569370
Rothstein TL Thomas EM Sumi SM Predicting outcome in hypoxic-ischemic coma. A prospective clinical and electrophysiologic study Electroenceph clin Neurophysiol 1991 79 101 107 1713822 10.1016/0013-4694(91)90046-7
Bassetti C Bomio F Mathis J Hess CW Early prognosis in coma after cardiac arrest: a prospective clinical, electrophysiological, and biochemical study of 60 patients J Neurol Neurosurg Psychiatry 1996 61 610 615 8971110
Chen R Bolton C Young GB Prediction of outcome in patients with anoxic coma: A clinical and electrophysiologic study Crit Care Med 1996 24 672 678 8612421 10.1097/00003246-199604000-00020
Berkhoff M Donati F Bassetti C Postanoxic alpha (theta) coma: a reappraisal of its prognostic signficance Clinical Neurophysiology 2000 111 297 304 10680565 10.1016/S1388-2457(99)00246-1
Zandbergen EGJ de Haan RJ Koelman JHTM Hijdra A Prediction of poor outcome in anoxic-ischemic coma J Clin Neurophysiol 2000 17 498 501 11085553 10.1097/00004691-200009000-00008
Synek VM Synek BJL Theta pattern coma, a variant of alpha pattern coma Clin Electroenceph 1984 15 116 121
Synek VM Synek BJL Transition from alpha to theta pattern coma in fatal cerebral anoxia Clin Exp Neurol 1988 25 109 113 3267481
Janati A Archer RL Osteen PK Coexistence of ectopic rhythms and periodic EEG pattern in anoxic encephalopathy Clin Electroenceph 1986 17 187 194
Janati A Husain MM Moore DB Adametz JR Suppression-burst pattern associated with generalized epileptiform discharges and alpha-theta pattern coma Clin Electroenceph 1986 17 82 88
Bortone E Bettoni L Giorgi C Trabattoni GR Macia D Post-anoxic theta and alpha pattern coma Clin Electrophysiol 1994 25 156 159
Echlin FA Arnett V Zoll J Paroxysmal high voltage discharges from isolated and partially isolated human and animal cerebral cortex Electroenceph clin Neurophysiol 1952 4 147 164 13033793 10.1016/0013-4694(52)90004-7
Henry CE Scoville WB Suppression-burst activity from isolated cerebral cortex in man Electroenceph clin Neurophysiol 1952 4 1 22 14906270 10.1016/0013-4694(52)90027-8
Niedermeyer E Sherman DL Geocadin RJ Hansen HC Hanley DF The burst-suppresssion electroencephalogram Clin Electroencephalogr 1999 30 99 105 10578472
Steriade M Amzica F Contreras D Cortical and thalamic cellular correlates of electroencephalographic burst-suppression Electroencephalogr clin Neurophysiol 1994 90 1 16 7509269
Pourmand R Burst-suppression pattern with unusual clinical correlates Clin Electroencephalogr 1994 25 160 163 7813097
Nelson KR Brenner RP Carlow TJ Divergent-convergent eye movements and transient eyelid opening associated with an EEG burst-suppression pattern J Clin Neuroophthalmol 1986 6 43 46 2939113
Berg A Further evidence that for status epielpticus "one size fits all" doesn't fit Neurology 2002 58 515 516 11865124
Logroscino G Hesdorffer DC Cascino GD Annegers JF Bagliella E Hauser WA Long-term mortality after first episode of status epilepticus Neurology 2002 58 537 541 11865129
Wijdicks EFM Propofol in myclonus status epilepticus in comatose patients following cardiac resuscitation J Neurol Neurosurg Psychiatry 2002 73 94 95 12082068 10.1136/jnnp.73.1.94
Nagel M Mielke A Ferbert A Propofol beim posthypoxischen Status myoclonicus Nervenheilkunde 2003 22 A50
DeGiorgio CM Tomiyasu U Gott PS Treiman DM Hippocampal pyramidal cell loss in human status epilepticus Epilepsia 1992 33 23 27 1733757 10.1111/j.1528-1157.1992.tb02278.x
|
16026615
|
PMC1190185
|
CC BY
|
2021-01-04 16:28:53
|
no
|
BMC Neurol. 2005 Jul 18; 5:14
|
utf-8
|
BMC Neurol
| 2,005 |
10.1186/1471-2377-5-14
|
oa_comm
|
==== Front
BMC NeurosciBMC Neuroscience1471-2202BioMed Central London 1471-2202-6-491607899310.1186/1471-2202-6-49Research ArticleEffect of neutrophil depletion on gelatinase expression, edema formation and hemorrhagic transformation after focal ischemic stroke Harris Alex K [email protected] Adviye [email protected] Anna [email protected] Livia S [email protected] Maribeth H [email protected] Susan C [email protected] Program in Clinical and Experimental Therapeutics, College of Pharmacy, University of Georgia, Augusta, Georgia, USA2 Vascular Biology Center, Medical College of Georgia, Augusta, Georgia, USA3 Veteran's Affairs Medical Center, Medical College of Georgia, Augusta, Georgia, USA4 Department of Biostatistics, Medical College of Georgia, Augusta, Georgia, USA5 Department of Neurology, Medical College of Georgia, Augusta, Georgia, USA2005 3 8 2005 6 49 49 31 3 2005 3 8 2005 Copyright © 2005 Harris et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
While gelatinase (MMP-2 and -9) activity is increased after focal ischemia/reperfusion injury in the brain, the relative contribution of neutrophils to the MMP activity and to the development of hemorrhagic transformation remains unknown.
Results
Anti-PMN treatment caused successful depletion of neutrophils in treated animals. There was no difference in either infarct volume or hemorrhage between control and PMN depleted animals. While there were significant increases in gelatinase (MMP-2 and MMP-9) expression and activity and edema formation associated with ischemia, neutrophil depletion failed to cause any change.
Conclusion
The main finding of this study is that, in the absence of circulating neutrophils, MMP-2 and MMP-9 expression and activity are still up-regulated following focal cerebral ischemia. Additionally, neutrophil depletion had no influence on indicators of ischemic brain damage including edema, hemorrhage, and infarct size. These findings indicate that, at least acutely, neutrophils are not a significant contributor of gelatinase activity associated with acute neurovascular damage after stroke.
==== Body
Background
The matrix metalloproteinases (MMPs) are a family of some 23 zinc dependent proteases that, collectively, possess the ability to degrade nearly every component of the extra-cellular matrix [1-3]. The activity of the MMPs is tightly controlled through proteolytic activation of the zymogen forms and stoichiometric binding of tissue inhibitors of metalloproteinases (TIMPs). The MMPs play an important role in many physiological processes due to their inherent ability to remodel tissues [2,3]. However, in disease states such as vascular disease and stroke, the MMPs may become deleterious due to dysregulation and can result in tissue injury and inflammation. Specifically, the MMPs may be involved in the degradation of the basal lamina in reperfusion injury resulting in disruption of the blood brain barrier and hemorrhagic transformation [4].
Recently, several lines of evidence have demonstrated the involvement of the MMPs in cerebral ischemia. Studies in rat, mouse, and baboon models have shown that MMP-9 is up-regulated following transient focal ischemia [5-8]. Additionally, Asahi et al. have shown that MMP-9 knockout as well as MMP-9 inhibition reduces ischemic lesion volume [9]. However, others have shown that pharmacological inhibition of MMP-9 yields no change [10,11]. Lapchak et al. demonstrated that broad-spectrum MMP inhibition (BB-94) reduced the incidence of hemorrhage in tPA treated brains when administered shortly after the onset of ischemia [11]. In addition, Sumii et al. were able to show a reduction in hemorrhage severity in tPA treated animals given the same MMP inhibitor BB-94 [10].
Clearly the MMPs are involved in the pathology of cerebral ischemia and hemorrhagic transformation. However, there is still uncertainty as to the origin of MMP activity. Immunohistochemical studies from Asahi et al. demonstrated that MMP-9 is primarily up-regulated in the vascular spaces while others have shown that there is a concomitant staining of neutrophils in the areas of MMP-9 activation suggesting a role for the neutrophil in the up-regulation of MMP-9 [12-14]. Indeed, a recent study has shown that prevention of neutrophil infiltration significantly reduces MMP-9 up-regulation in an occlusion/reperfusion model of ischemia [15]. The source of MMP activity in focal cerebral ischemia is important to the development of therapies to target this mediator of neurovascular injury. It is still unknown whether neutrophils are an important source of MMPs in experimental hemorrhagic transformation.
The objective of the current study was to evaluate the role of the neutrophil in hemorrhagic transformation and edema formation in a hyperacute 3-hour occlusion/reperfusion model of focal ischemia. The central hypothesis was that depletion of neutrophils would reduce hemorrhage development due to prevention of the up-regulation of MMP-9.
Results
Neutrophil depletion
Control and neutrophil-depletion groups received normal rabbit serum and anti-PMN antibody, respectively, 24 hours prior to middle cerebral artery occlusion (MCAO) surgery. The dose of the anti-neutrophil antibody chosen (and this was batch-specific in our preliminary studies – not shown) was very effective in reducing the circulating neutrophils. In control animals that received normal serum, neutrophil count, expressed as mean ± SE of percent of total leukocyte count, was 11.8 ± 0.92% and as expected did not differ from baseline. In the depletion group, administration of anti-PMN antibody reduced the neutrophil count to 0.3 ± 0.11% from 9.4% at baseline (p < 0.0001 compared to serum). All animals included in the neutropenic group had >90% depletion of their neutrophils prior to the stroke surgery, as assessed by a blinded investigator using a hemocytometer (Fig. 1). The body weight of control animals was 288 ± 4 g at baseline, 292 ± 4 g prior to stroke and 237 ± 4 g at 24 hours. In the antibody-treated group, body weight was 300 ± 4 g at baseline, 292 ± 5 prior to stroke surgery, and 240 ± 5 g at 24 hours.
Gelatinase quantification
Ischemia for three hours followed by 24 h reperfusion resulted in a significant increase in the expression and activity of both MMP-2 and MMP-9 in the ipsilateral hemispheres (Fig. 2). Quantitative analysis of the gelatin zymography revealed consistent increases in the lower molecular weight form of MMP-9 (88 kDa). Although there was a faint band corresponding to 95 kDa MMP, the gelatinolytic activity in control vs neutrophil-depleted animals was similar. Protein levels of MMP-2 (Fig. 3) and MMP-9 (Fig. 4) were significantly increased in the ischemic hemisphere and neutrophil depletion did not affect MMP protein levels.
Infarct size, edema and hemorrhage assessment
There was no significant effect of neutrophil depletion on either infarct size or the development of hemorrhagic transformation in our model (Figure 5). In addition, cerebral edema formation was similar between untreated and neutrophil-depleted animals. Neurologic examination at 24 hours was not significantly different between the two treatment groups (not shown).
Discussion
Cerebral ischemia and reperfusion are known to induce large increases in MMP-9 protein and activity in the affected hemisphere and MMP-9 has been associated with hemorrhage formation in humans [16] and experimental animals [10]. Experimental inhibition of the MMPs has been associated with decreased hemorrhage formation and improved stroke outcome [11]. Since neutrophils are a known source of MMP-9, this study was designed to determine whether neutrophil depletion, prior to cerebral artery occlusion, would decrease MMP expression and reduce hemorrhagic transformation.
Despite the fact that neutrophil depletion was essentially complete prior to the focal cerebral ischemia in our experiment, we detected an increase in gelatinolytic activity corresponding to 86–88 kDa MMP-9. There was a faint band for the 95 kDa proMMP-9, and the activity did not differ between the ischemic and non-ischemic hemispheres. Immunoblotting experiments confirmed the results of gelatin zymography and showed that specific up-regulation of the MMP-9 protein is responsible for increased gelatinolytic activity. Justicia et al. reported that, in a 1 h MCAO model in which the neutrophil depletion or antagonism was achieved by vinblastine administration or neutralizing antibodies against VCAM-1, respectively, there was no effect on the 88 kDa MMP-9 expression but, increases in 95 kDa MMP-9 activity were prevented. Our results are in agreement with this observation that neutrophils are not the source of 88 kDa MMP-9. However, we do not detect any increase in 95 kDa MMP-9. This difference may be due to the duration of ischemia (1 h vs 3 h) and the extent of ischemic damage [15]. We have previously shown that hemorrhagic transformation occurs in various different regions of the injured hemisphere including: preoptic area, striatum and the lateral cortex in this model and that development of hemorrhagic transformation is related to the duration of occlusion, with no hemorrhage when reperfusion occurs after only 1 hour [17,18].
In this model of focal cerebral ischemia and hemorrhagic transformation, where the MCA was occluded for 3 hours prior to reperfusion, no significant effect of neutrophil depletion could be shown on either neurovascular damage or neurologic function. Also, from our results, it is clear that neutrophils are not an important contribution to the increased MMP-9 expression in the 24 hours after this injury. It is possible that the contributions of neutrophils are greater when the injury to the brain is less profound than in our model. Several different investigators have demonstrated that preventing the adhesion of neutrophils [19] after ischemia and reperfusion reduces the ultimate injury. Maier et al. reported that in superoxide dismutase (SOD) knock-out animals that are more susceptible to ischemic damage, neutrophils are not the source of MMP-9 contributing to blood brain barrier breakdown providing further support for the results of this present study [20].
Activated neutrophils release free radicals and proteolytic enzymes such as MMP-8 (neutrophil elastase), in addition to activating cytokines, which further the recruitment of leukocytes to the site of injury. We initially thought that neutrophil adherence and activation may be an early contributor to the development of microvascular injury and hemorrhagic transformation after cerebral ischemia but this turned out to be not the case [21]. It is still possible that neutrophils are involved in the destruction of the basal lamina, but it is likely to occur later in the process, after neutrophil recruitment is maximal.
Conclusion
Although neutrophils have been shown to contain MMP-9, release of MMP-9 by neutrophils is probably not the mechanism of the early microvascular damage leading to edema formation and hemorrhagic transformation in our model. In addition, neutrophil adhesion is not necessary for the increase in activation of MMP-9 after ischemia. Although neutrophils may contribute to the ultimate degree of neurologic damage after ischemia and reperfusion in the brain, they are not necessary for the development of hemorrhagic transformation.
Methods
Neutrophil depletion
Neutrophil depletion was accomplished by intravenous administration of a polyclonal rabbit anti-rat polymorphonuclear neutrophil (PMN) antiserum (Accurate Chemical and Scientific Corporation, Westbury, NY) 24 hours prior to ischemia [22]. Animals were randomized to receive either the PMN antiserum (0.3 mL diluted in 1 mL saline) or an equal volume of normal rabbit serum (control) 24 hours prior to surgery. Neutrophil counts, as a percentage of total leukocyte count, were determined at baseline and 23 hours post administration to determine the level of depletion.
Animal preparation/physiological monitoring
The Care of Experimental Animal Committee of Medical College of Georgia approved the protocol. Male Wistar rats, from the Charles River Breeding Company (Wilmington, MA, USA) within a narrow range of body weight (250–300 g) were used. All animals were anesthetized with 2% isoflurane via inhalation. Cerebral ischemia was induced using the intraluminal suture MCAO model [23]. The MCA was occluded with a 19–21 mm 4-0 surgical nylon filament, which was introduced from the external carotid artery lumen into the internal carotid artery to block the origin of the MCA. The suture was removed after 3 hours of occlusion. We have shown previously that hemorrhagic transformation is related to the duration of occlusion in this model [17,18]. Prior to, and after reperfusion, each rat was evaluated neurologically using the Bederson Scale [24]. After 21 hours of reperfusion, the animals were anesthetized with ketamine 44 mg/kg and xylazine 13 mg/kg I.M. (cocktail), perfused with saline, sacrificed, and the brain tissue was removed.
Brain preparation and homogenization
The brains were sliced into seven 2-mm thick slices in the coronal plane and stained with a 2% solution of 2, 3, 5-triphenyltetrazolium chloride (TTC) (Sigma Chemical Co., USA) for 15–20 minutes. Images of the stained sections were taken. Grossly visible infarction zones were quantified using image analysis software (Zeiss- KS 300) in pixel2 [25]. Because brain edema can affect the accuracy of infarct assessment [26], the measurement of the corrected infarct size was taken [27,28]. The ischemic and non-ischemic hemispheres of the slices for the ELISA assay were separated and processed individually, using the non-ischemic side as a control. Each slice was homogenized, vortexed and sonicated. Quantification of the hemoglobin was accomplished using a direct ELISA method, as previously described [29].
For zymography and western blotting, brains were placed in a chilled coronal matrix and 2-mm sections were removed from the core of the infarct. The slices were collectively placed, by hemisphere, into labeled tubes, flash frozen, and stored at -80°C until use. Samples were homogenized according to the methods of Heo et al. [7]. Briefly, samples were placed in 300 μL cold working buffer consisting of 50 mmol/L Tris-HCl (pH 7.5), 75 mmol/L NaCl, and 1 mmol/L PMSF, homogenized with a glass homogenizer and centrifuged at 4°C for 20 minutes at 9000 rpm. The supernatant was collected and stored at -80°C until use. Protein concentrations were determined using the Bradford method.
Gelatin zymography
On the day of the experiment, samples (20 μg protein/sample) were loaded onto 10% gelatin zymogram gels (Bio-Rad) and separated under nonreducing conditions [30]. The gels were then rinsed twice in 2.5% Triton X-100 and incubated for 24 hours in substrate buffer containing 21 mM Tris·HCl, 10 mM CaCl2, 0.04% NaN3. Gels were then stained by Coomassie blue R-250 followed by destaining in 55% methanol and 7% acetic acid. Lytic activity was viewed as clear bands on a dark blue background and was quantitated by densitometric analysis (Gel-Pro v 3.1, Media Cybernetics, Carlsbad, CA.).
Edema quantitation
Brain edema was measured as a difference in area between the ischemic and non-ischemic hemispheres. The brain section analyzed was the one corresponding to the area of the ischemic damage in four consecutive 2-mm slides. The brain slides were analyzed separately and the areas and differences combined for whole hemi-section comparison. The area analysis was done using Zeiss KS-300.
Statistics
A Wilcoxon rank sum test was performed to make group comparisons (control vs. depletion) in MMP activity and protein expression for the difference between the ischemic and non-ischemic sides of the brain and group comparisons for neutrophil depletion, edema, infarct size and hemorrhagic transformation. A Wilcoxon signed rank test was used to make within rat comparisons between ischemic and non-ischemic sides of the brain for MMP activity and protein expression. Statistical significance was determined at p < 0.05 and SAS 8.2 was used for all analyses.
Authors' contributions
AKH performed all the zymography and western blotting and contributed to preparation of the manuscript.
AE contributed to the optimization of zymograms, data interpretation and preparation of the manuscript.
AK did all the animal preparation, MCAO, infarct and hemoglobin analysis.
LM completed the edema analysis and contributed to the manuscript preparation.
MHJ performed the statistical analysis.
SCF coordinated all aspects of this work and prepared the manuscript.
All authors approved the final version of the manuscript.
Acknowledgements
This work was supported by grants from NINDS (1U01NS43127-01 and RO1NS044216-01) and the American Heart Association Southeast Affiliate to SCF, American Diabetes Association Research Award and NIH (HL076236-01) awards to AE and American Heart Association Southeast Affiliate predoctoral fellowship to AKH.
Figures and Tables
Figure 1 Neutrophil counts at baseline and at 24 hours after normal rabbit serum (control) or anti-PMN antibody (PMN-depleted) administration. Anti-PMN antibody resulted in a consistent decrease in circulating neutrophils to less than 10% of baseline counts whereas there was no change in neutrophil counts in control animals that received the normal rabbit serum. *p < 0.001 control (n = 8) vs after PMN-depletion (n = 10).
Figure 2 MMP activity is increased in focal cerebral ischemia. (A) Representative zymogram showing changes in MMP activity in the ischemic (I) and non-ischemic (N) hemispheres. (B) Densitometric analysis of lytic bands indicates increases in MMP-2 and MMP-9 activity in ischemic hemispheres from control and PMN-depleted animals. Anti-PMN antibody treatment has no affect on MMP activity. (*p < 0.05 vs contralateral hemisphere, control n = 6, PMN n = 4).
Figure 3 MMP-2 protein expression is increased in ischemia. (A). Representative immunoblot demonstrating changes in MMP-2 protein expression in ischemic hemispheres from control and PMN-depleted animals. (B) Densitometric analysis of immunoreactive bands indicates that MMP-2 protein is increased in ischemia and is unchanged by neutrophil depletion. (*p < 0.05 vs contralateral hemisphere, saline n = 6, PMN n = 4).
Figure 4 MMP-9 protein expression is unchanged by neutrophil depletion. (A). Representative immunoblot showing immunoreactive MMP-9 bands in both ischemic (I) and non-ischemic (N) hemispheres. (B) Densitometric analysis indicates an ischemia induced elevation of MMP-9 expression that is unaffected by PMN depletion. (*p < 0.05 vs contralateral hemisphere, control n = 6, PMN n = 4).
Figure 5 Infarct size, edema formation and hemoglobin excess differences between control and PMN-depleted animals. (A). Infarct size, as measured by TTC staining, indicated no differences between depletion and control groups. (B). Likewise, PMN depletion had no affect on the development of hemorrhagic transformation. (C) Edema determined as the difference in area between the ischemic and non-ischemic hemispheres was similar in neutrophil-depleted and control animals (n = 8 control, n = 10 PMN).
==== Refs
Nagase H Woessner JF Matrix metalloproteinases J Biol Chem 1999 274 21491 21494 10419448 10.1074/jbc.274.31.21491
Visse R Nagase H Matrix metalloproteinases and tissue inhibitors of metalloproteinases: structure, function, and biochemistry Circ Res 2003 92 827 839 12730128 10.1161/01.RES.0000070112.80711.3D
Smeglin A Frishman WH Elastinolytic matrix metalloproteinases and their inhibitors as therapeutic targets in atherosclerotic plaque instability Cardiol Rev 2004 12 141 150 15078583 10.1097/01.crd.0000105000.46909.81
Hammann GF Okada Y Fritge R del Zoppo GJ Microvascular basal lamina antigens disappear during cerebral ischemia and reperfusion Stroke 1995 2120 2126 7482660
Rosenberg GA Navratil M Barone F Feuerstein G Proteolytic cascade enzymes increase in focal cerebral ischemia in rat J Cereb Blood Flow Metab 1996 16 360 366 8621740 10.1097/00004647-199605000-00002
Gasche Y Fujimura Y Copin J Kawase M Masengale J Chan PH Early appearance of activated MMP-9 after focal cerebral ischemia in mice J Cereb Blood Flow Metab 1999 19 1020 1028 10478654 10.1097/00004647-199909000-00010
Heo JH Lucero J Abumiya T Koizol JA Copeland BR del Zoppo GJ Matrix metalloproteinases increase very early during experimental focal cerebral ischemia J Cereb Blood Flow Metab 1999 19 624 633 10366192 10.1097/00004647-199906000-00005
Mun-Bryce S Rosenberg GA Matrix metalloproteinases in eerebrovascular disease J Cereb Blood Flow Metab 1998 18 1163 1172 9809504 10.1097/00004647-199811000-00001
Asahi M Asahi K Jung J del Zoppo GJ Fini E Lo EH Role for matrix metalloproteinase 9 after focal cerebral ischemia: Effects of gene knockout and enzyme inhibition with BB-94 J Cereb Blood Flow Metab 2000 20 1681 1689 11129784 10.1097/00004647-200012000-00007
Sumii T Lo EH Involvement of matrix metalloproteinase in thrombolysis-associated hemorrhagic transformation after embolic focal ischemia in rats Stroke 2002 33 831 836 11872911 10.1161/hs0302.104542
Lapchak PA Chapman DF Zivin JA Metalloproteinase inhibition reduces thrombolytic (tissue plasminogen activator)-induced hemorrhage after thromboembolic stroke Stroke 2000 31 3034 3040 11108768
Asahi M Wang X Mori T Sumii T Jung J Moskowitz MA Fini E Lo EH Effects of matrix metalloproteinase-9 gene knock-out on the proteolysis of blood-brain barrier and white matter components after cerebral ischemia J Neurosci 2001 21 7724 7732 11567062
Planas AM Justicia C Sole S Friguls B Cervera A Adell A Chamorro A Certain forms of matrix metalloproteinase-9 accumulate in the extracellular space after microdialysis probe implantation and middle cerebral artery occlusion/reperfusion J Cereb Blood Flow Metab 2002 22 918 925 12172377 10.1097/00004647-200208000-00003
Romanic AM White RF Arleth AJ Ohlstein EH Barone F Matrix metalloproteinase expression increases after cerebral focal ischemia in rats: inhabition of matrix metalloproteinase-9 reduces infarct size Stroke 1998 29 1020 1030 9596253
Justicia C Panes J Sole S Cervera A Deulofeu R Chamorro A Planas AM Neutrophil infiltration increases matrix metalloproteinase-9 in the ischemic brain after occlusion/reperfusion of the middle cerebral artery in rats J Cereb Blood Flow Metab 2003 23 1430 1440 14663338 10.1097/01.WCB.0000090680.07515.C8
Montaner J Alvarez-Sabin J Molina C Angles A Abilleira S Arenillas J Monasterio J Matrix metalloproteinase expression is related to hemorrhagic transformation after cardioembolic stroke Stroke 2001 2762 2767 11739970
Fagan SC Nagaraja TN Fenstermacher JD Zheng J Johnson M Knight RA Hemorrhagic transformation is related to the duration of occlusion and treatment with tissue plasminogen activator in a nonembolic stroke model Neurol Res 2003 25 377 382 12870264 10.1179/016164103101201526
Fagan SC Garcia JH Hemorrhagic transformation in focal cerebral ischemia: influence of time to artery reopening and tissue plasminogen activator Pharmacotherapy 1999 19 139 142 10030763 10.1592/phco.19.3.139.30932
Chopp M Li Y Jiang N Zhang RL Prostak J Antibodies against adhesion molecules reduce apoptosis after transient middle cerebral artery occlusion in rat brain J Cereb Blood Flow Metab 1996 578 584 8964796 10.1097/00004647-199607000-00007
Maier CM Hsieh L Yu F Bracci P Chan PH Matrix metalloproteinase-9 and myeloperoxidase expression: quantitative analysis by antigen immunohistochemistry in a model of transient focal cerebral ischemia Stroke 2004 35 1169 1174 15060315 10.1161/01.STR.0000125861.55804.f2
Fagan SC Hess DC Hohnadel EJ Pollock DM Ergul A Targets for vascular protection after acute ischemic stroke Stroke 2004 35 2220 2225 15284446 10.1161/01.STR.0000138023.60272.9e
Sir O Fazal N Choudry MA Goris RJA Gamelli RL Sayeed MM Role of neutrophils in burn-induced microvascular injury in the intestine Shock 2000 113 117 10947152
Knight RA Barker PB Fagan SC Li Y Jacobs MA Welch KMA Prediction of impending hemorrhagic transformation in ischemic stroke using magnetic resonance imaging in rats Stroke 1998 29 144 151 9445344
Bederson JB Pitts LH Tsuji M Nishimura MC Davis RL Bratowski J Stroke 1986 472 476 3715945
Terai K Suzuki M Miyata K Amount of bleeding and hematoma size in the collagenase-induced intracerebral hemorrhage rat model Neurochemical Research 2003 28 779 785 12716030 10.1023/A:1022826220469
Lin TN He YY Wu G Khan M Hsu CY Effects of brain edema on infarct volume in a focal cerebral ischemia model in rats Stroke 1993 24 117 121 8418534
Yang Y Li Q Miuyashita H Howlett W Siddiqui M Shauaib A Usefulness of postischemic thrombolysis with or without neuroprotection in a focal embolic model of cerebral ischemia J Neurosurg 2000 92 841 847 10794299
Swanson RA Morton MT Tsao-Wu G Savalos RA Davidson C Sharp FR A semiautomated method for measuring brain infarct volume J Cereb Blood Flow Metab 1990 10 290 293 1689322
Hilali H Simpkins A Hill W Waller J Knight R Fagan S Single slice method for quantification of hemorrhagic transformation using direct ELISA Neurol Research 2004 93 98 10.1179/016164104773026606
Ergul A Portik-Dobos V Giulumian AD Molero MM Fuchs LC Stress upregulates arterial matrix metalloproteinase expression and activity via endothelin A receptor activation Am J Physiol Heart Circ Physiol 2003 285 H2225 32 12842810
|
16078993
|
PMC1190186
|
CC BY
|
2021-01-04 16:03:48
|
no
|
BMC Neurosci. 2005 Aug 3; 6:49
|
utf-8
|
BMC Neurosci
| 2,005 |
10.1186/1471-2202-6-49
|
oa_comm
|
==== Front
BMC Oral HealthBMC Oral Health1472-6831BioMed Central London 1472-6831-5-71604280610.1186/1472-6831-5-7Research ArticlePilot survey of oral health-related quality of life: a cross-sectional study of adults in Benin City, Edo State, Nigeria Okunseri Christopher [email protected] Amit [email protected] R Iván [email protected] Colman [email protected] Department of Clinical Sciences, School of Dentistry, Marquette University, Milwaukee, P.O. Box 1881, Wisconsin, 53201. USA2 Department of Dental Informatics, Temple University School of Dentistry, Philadelphia, USA3 Department of Periodontology and Public Health, Faculty of Dentistry, Prince Philip Dental Hospital, University of Hong Kong, Hong Kong2005 25 7 2005 5 7 7 22 11 2004 25 7 2005 Copyright © 2005 Okunseri et al; licensee BioMed Central Ltd.2005Okunseri et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Oral health studies conducted so far in Nigeria have documented prevalence and incidence of dental disease using traditional clinical measures. However none have investigated the use of an oral health-related quality of life (OHRQoL) instrument to document oral health outcomes. The aims of this study are: to describe how oral health affects and impacts quality of life (QoL) and to explore the association between these affects and the oral health care seeking behavior of adults in Benin City, Edo State, Nigeria.
Methods
A cross-sectional survey recruited 356 adults aged 18–64 years from two large hospital outpatient departments and from members of a university community. Closed-ended oral health questionnaire with "effect and impact" item-questions from OHQoL-UK© instrument was administered by trained interviewers. Collected data included sociodemographic, dental visits, and effects and impact of oral health on QoL. Univariate and bivariable analyses were done and a chi-square test was used to test differences in proportions. Multivariable analyses using ANOVA examined the association between QoL factors and visits to a dentist.
Results
Complete data was available for 83% of the participants. About 62% of participants perceived their oral health as affecting their QoL. Overall, 82%, 63%, and 77% of participants perceived that oral health has an effect on their eating or enjoyment of food, sleep or ability to relax, and smiling or laughing, respectively. Some 46%, 36%, and 25% of participants reported that oral health impact their daily activities, social activities, and talking to people, respectively. Dental visits within the last year was significantly associated with eating, speech, and finance (P < 0.05). The summary score for the oral health effects on QoL ranged from 33 to 80 with a median value of 61 (95% CI: 60, 62) and interquartile range of 52–70. Multivariable modeling suggested a model containing only education (F = 6.5, pr>F = 0.0111). The mean of effects sum score for those with secondary/tertiary education levels (mean = 61.8; 95% CI: 60.6, 62.9) was significantly higher than those with less than secondary level of education (mean = 57.2; 95% CI: 57.2, 60.6).
Conclusion
Most adults in the study reported that oral health affects their life quality, and have little/no impact on their quality of life. Dental visits within the last year were associated with eating, speech, and finance.
==== Body
Background
Although common oral diseases are not life threatening, their outcomes may influence the overall wellbeing of individuals and populations. Oral health-related quality of life (OHRQoL) characterizes a person's perception of how oral health influences an individual's life quality and overall well being. This concept has received a lot of attention in the past two decades from sociologists, psychologists and the health professions, [1-15] with different instruments been developed to measure quality of life (QoL) and OHRQoL.
Cohen & Jago [2] first recognized the lack of data on the psycho-social impact of oral health problems. [3] To address this, several authors developed socio-dental indicators to measure the social impact of oral health problems. [5-10] In addition, other generic and disease-specific measures were developed based on the conceptual framework of the World Health Organization's (WHO) International Classification of Impairments, Disabilities and Handicaps (ICIDH). [1,6-16] However, some concerns have been raised about the instruments so far developed, because of their use of older adults in testing the reliability and validity of the instruments, the use of non-random samples, and some have mostly professionally dominated opinions. [13,15] Other concerns also include measuring of positive and or negative effects related to QoL and the varied number of item questions or domains in each instrument. [13,15]
Most of the OHRQoL instruments developed so far measure either the "effect" or the "impact" of oral health on life quality and others measure the "effect and "impact" together. The "effect" dimension examines the physical, psychological and social effects of oral health attributes, and the "impact" dimension examines the impact of oral health attributes on daily activities, chewing ability and talking to people. It also examines the impact of the effects on individuals' overall quality of life. This "effect" and "impact" domains of oral health are better assessed using OHRQoL measures rather than the traditional clinical disease status measures. Slade & Spencer [8] and Adulyanon et al. [12] instruments for the most part focused on the negative effects of how oral health affects quality of life, but that developed by McGrath & Bedi [13] included the positive "effects" dimensions which reflected the concept of health beyond the mere absence of disease-impairment-disability-handicap. [13] Developing this idea further, Locker [17] suggested an extension of the ICIDH scope to include certain feeling states (e.g., pain and psychological discomfort) which are prominent consequences of oral disease.
The instrument (OHQoL-UK©) developed by McGrath & Bedi used a random probability sampling method. [13,15] It is based on the public's perceptions in the United Kingdom of how oral health affects life quality. [13,15] It consists of 16 key questions relating to 16 key areas of oral health-related quality of life, such as comfort, speaking, and social life, and each of the 16 key questions are also rated for their 'impact' on overall quality of life. [13,15] OHQoL-UK© has been tested for reliability and validity and found to be a valid and reliable measure for assessing OHRQoL, and have also been reported to have good psychometric properties. [13]
OHQoL-UK© and other oral health-related quality of life instruments have been used to explore a relationship between sociodemographic factors in different populations, [18-20] from different countries including Tanzania, Greece, Thailand, Germany, Syria, Egypt, Saudi Arabia, and Uganda. [12,18-22] This has lead to a paradigm shift from the use of only traditional assessment of oral health with a focus on disease to a more comprehensive community measure of health service provision. [1] This shift gives healthcare providers the opportunity to move from the concept of just treating disease, to a holistic model of caring for the patient as a productive member of the society under the "socio-environmental-medical model" of caregiving that encompasses a broader definition of oral health.
Studies show that OHRQoL is related to age, gender, and socioeconomic factors. [6,22] A study of secondary school students conducted in Nigeria found that participants perceived their teeth to be important for their appearance [23] and self esteem. [24] Overall, their perception of the importance of dental health was similar to those reported from the United States. [25] In a recently published study conducted in Nigeria we demonstrated that being younger, being female, and being employed were associated with visiting a dentist in the past 12 months. [26] Other studies have documented prevalence of dental caries and periodontal disease in Nigeria, [27] and described oral health care practice among physicians, [28] as well as oral health knowledge and attitudes of Nigerian school teachers. [29]
Despite these studies there is a paucity of information on how oral health affects and impacts quality of life of persons from sub-Saharan African countries (e.g., Nigeria) which have multiple tribes, varied cultural beliefs, high levels of unemployment and poverty. The specific aims of this study were: 1) to describe the effect and impact of OHRQoL factors, and 2) to explore the association between these effects and oral health care seeking behavior of adults in Benin City, Edo State, Nigeria. This study used oral health related quality of life measures patterned after OHQoL-UK© [13]. The questions of how oral health is related to quality of life were described in two dimensions: "effects" and "impacts". The effect dimension included three domains (physical effects, psychological effects and social effects), and the impact dimension included three domains (impact on daily activities, chewing ability and talking to people). We believe that this study will fill a gap in OHRQoL on Benin City, Edo State, Nigeria and will serve as an impetus for more research in this area.
Methods
Sample selection
This study was conducted in Benin City a town which has a population of 2.2 million. The city is a major commercial center that serves as the gateway between the northern, western, and eastern parts of Nigeria, and is home to a substantial number of people from all the major Nigerian tribes/ethnic groups. One of the four Nigerian dental schools is located in Benin City. Anecdotal evidence suggests that the general and teaching hospitals located in the city treat the most patients compared to all private clinics in the state put together. The individuals using these hospitals come from all levels of the socioeconomic fabric of the society. The teaching hospital in Benin City is also adjacent to a large university community. Individuals for the most part pay for dental services out of pocket on a fee-for-service basis at government-owned and privately-owned dental clinics.
Four hundred and twenty six persons were recruited to participate in this study, of which three hundred and fifty-six (83%) had complete usable information. Participants aged 18–64 years were recruited from two large outpatient medical care facilities (University of Benin Teaching Hospital and Central Hospital), and from the adjacent university community. Three interviewers were trained by one of the authors (CO). The interviewers conducted face-to face interviews with the adult participants at the waiting area of the medical outpatient clinics over a 5-week period in the summer of 1999. On average, it took 10 minutes of contact time between the interviewer and the participant in the outpatient waiting area to complete one questionnaire. The importance of collecting this data was explained to participants and their participation was strictly voluntary with no incentives offered.
Data collection
The closed-ended questionnaire was prepared in English and consisted of the 16 key questions of OHRQoL identified in the OHQoL-UK© by McGrath et al. 2000. [13,15] The questionnaire was pre-tested among a group of medical hospital outpatients and university students before it was administered to the study participants. The questions of how oral health is related to quality of life was patterned after OHQoL-UK© [13,15,22,30,31] and described in two dimensions "effects" and "impacts".
The effect dimension included three domains (physical effects, psychological effects and social effects), and the impacts dimensions included three impact item questions (impact on daily activities, chewing ability and talking to people). The "impact" item questions used in this study included only 3 item questions from the original OHQoL-UK©, and was analyzed separately from the "effect" portion of the instrument. Participants were interviewed using closed-ended questions, such as; "What affect does your oral health have on your eating or enjoyment of food"? Possible responses on the "effect" were: "Very Good, Good, None, Bad, Very Bad". For example, a question on impact was: "Have problems with your teeth or gums affected your daily activities such as your work or hobbies? Possible responses were: all of the time, most of the time, some of the time, little of the time, none of the time. Each item was scored on a Likert scale from 1 to 5, with a "very bad effect" scored as 1, very good effect as 5, and no effect as 3. The sum of individual item responses were added together to generate an overall OHQoL-UK© score with possible values ranging from 16–144. Additionally, the sum of the responses to items in each domain (physical, psychological, and social) produced subdomain scores. Other data collected were age, self-reported oral health problems, dental visits, gender, ethnicity, number of teeth they possessed, and educational status.
Data analysis
Data from the paper questionnaires were entered into a computer using SPSS v10.0 for Windows [32] and later converted to SAS® data sets (SAS® V 8.2 Cary, NC, USA) for all analyses. [33] The variables available for this study were sex, age, educational level, employment status, tribe/ethnicity, and last dental visit. Education was categorized into two groups: primary education, and secondary/tertiary education. Oral health care utilization variable was derived from the question: how long ago was it since your last visit to the dentist? The possible responses to this question were: "within the last twelve months", "between twelve and thirty six months" and the last option was "never been to a dentist". We dichotomized the variable at the 12-month time point.
Univariate analyses were conducted for all variables, and all missing/out of range values were verified against the paper questionnaire for accuracy and data entry errors corrected. We evaluated bivariable associations of available variables with visit to a dentist in the past 12 months and sex using Chi-square tests. Statistical significance was inferred at P < 0.05. For multivariable analyses, visit to dentist in the past one year (yes/no); sex; education (primary, and secondary/tertiary); employment (yes/no) were dichotomized, whereas age was categorized into three levels –18–24 years, 25–34 years, 35+ years; and ethnicity had four levels – Edo, Ibo, Yoruba, Others (the first three being major Nigerian tribal/ethnic groups). First, we assessed whether the differences arising out of the bivariate analyses remained after adjusting for confounding by sociodemographic factors. For this, we used logistic regression models for modeling the association between visits to a dentist in the past year and the OHRQoL measured attributes. Thereafter we assessed in the same way whether the differences between the two sexes remained upon adjustment for confounding by sociodemographic factors.
In another set of multivariable analyses, we assigned positive integer values to the effect responses (from 1 = very bad to 5 = very good) to derive a sum score for effects. Thereafter, we used this sum score of effects as a continuous variable. We evaluated differences between explanatory factor groups using ANOVA models studying main effects only in SAS® employing PROC GLM. For pair-wise comparisons, we employed Scheffe's test, controlling for Type I errors in post hoc testing of differences in group means.
Results
Eighty six percent of the participants were below 35 years of age; 55% were women; and 88% had secondary/tertiary education. Most participants (71%) reported that they will only visit a dentist when they need treatment and 21% reported that they had current dental problems, but choose to delay getting the required treatment. Some 88% of participants reported that they could not afford dental treatment, and 89% reported that they were not ready to spend money on dental treatment. Overall, 62% reported that they perceived their oral health to be good; 35% as moderate and 3% as bad. The summary score for the oral health effects on quality of life for the participants ranged from 33 to 80 with a median value of 61 (95%CI: 60, 62) and interquartile range of 52 – 70.
Table 1 shows the response distribution for the effect dimension of OHRQoL. Overall, more than 17% of participants reported a good or very good effect of oral health related issues on their quality of life on each of the domains (physical, psychological and social). About 18–47% participants reported that oral health issues did not have any effect on different aspects within each domain (Table 1). In general, the proportion of participants reporting 'no effect' was substantially lower than those reporting a good or very good effect. The only two exceptions to this pattern was the effect of oral health issues on finances and on work. For effect on finances, 18% participants reported bad, 0.8% very bad effect; 44% participants reported no effect; and about 19%, and 17% reported good or very good effect respectively. About 47% of participants reported no effect of oral health on work 26% and 22% reported a good or very good effect respectively.
Table 1 Attributes of the "effect" dimension of oral health related quality of life of study populations (n = 356).
Response (percent) to "effect" questions
Domain Attribute (effect on) Very bad Bad None Good Very good
Physical Eating 1.4 8.7 18.4 34.4 37.2
Appearance 0.6 3.1 23.5 37.2 35.8
Speech 0.6 2.8 26.5 34.9 35.2
General health 0.28 3.9 22.1 45.5 28.2
Breath 0.6 6.4 29.6 36.0 27.4
Comfort/Relaxation 0.3 8.9 31.2 32.7 26.8
Psychological Sleep 1.4 4.8 36.6 27.4 29.9
Confidence 0.3 6.7 29.6 32.1 31.3
Worry 1.1 3.6 41.9 29.3 24.0
Mood 0.6 6.15 44.4 27.9 21.0
Personality 1.4 4.2 32.4 32.1 29.9
Social Social life 0.6 4.2 29.3 34.9 31.0
Romantic relationships 0.8 3.6 34.1 29.3 32.1
Smiling 0.0 6.4 23.5 31.6 38.6
Work 0.6 4.5 47.2 26.0 21.8
Finance 0.8 18.4 44.1 19.3 17.3
Overall, more men reported good or very good effects of oral health on different quality of life attributes compared to women, though these were not statistically significant (Table 2). However, more women (67%) reported a good/very good effect of oral health on sleep compared to men (45%). This association remained significant even after adjusting for age, ethnicity, employment and education. The adjusted odds ratio of reporting a good/very good effect by women was 2.24 (95% CI: 1.41, 3.57) compared to men.
Table 2 Attributes of the "effect" dimension of oral health related quality of life between sexes of study populations (n = 356).
Response (percent) to "effects" questions
Men (n = 161) Women (n = 195)
Domain Attribute (effect on) Very bad/Bad None Very good/Good Very bad/Bad None Very good/Good
Physical Eating 8.1 18.6 73.3 11.7 18.3 70.1
Appearance 3.7 21.1 75.2 3.6 25.4 71.1
Speech 3.7 24.2 72.0 3.0 28.4 68.5
General Health 3.7 24.2 72.0 4.6 20.3 75.1
Breath 6.8 28.6 64.6 7.1 30.5 62.4
Comfort/Relaxation 8.1 36.0 55.9 10.2 27.4 62.4
Psychological Sleep 5.6 49.1 45.3 6.6 26.4 67.0a
Confidence 9.3 24.8 65.8 5.1 33.5 61.4
Worry 6.2 42.9 50.9 3.6 41.1 55.3
Mood 6.2 46.6 47.2 7.1 42.6 50.3
Personality 5.0 32.3 62.7 6.1 32.5 61.4
Social Social life 5.0 24.8 70.2 4.6 33.0 62.4
Romantic Relationship 5.0 35.4 59.6 4.1 33.0 64.9
Smiling 6.2 24.8 68.9 6.6 22.3 71.1
Work 5.6 46.0 48.4 4.6 48.2 47.2
Finance 19.9 49.1 31.1 18.8 40.1 41.1
aStatistically significantly different between men and women (P < 0.05)
Table 3 shows participants' response to effect questions classified by previous visits to a dentist. A substantial proportion of those who had never visited a dentist reported good or very good effect of OHRQoL especially for the effect on eating, speech, worry, mood, and finance. This was statistically significant. The differences for effects on appearance, comfort and relaxation, work, and finance were substantial and came very close to being statistically significant (p-values ranged between 0.05–0.07). Upon multivariable adjustment for age, sex, employment, ethnicity and education, the effects on eating, relaxation, and worry remained statistically significant.
Table 3 Attributes of effect attributes of oral health related quality of life of participants with and without a dental visit.
Response (percent) to effects dimension
Never Visited Dentist n = 79 Last visit >1 year ago n = 185 Last visit within 1 year n = 92
Domain Attribute (effect on) Very bad/Bad None Very good/Good Very good/Good Very good/Good
Physical Eating 6.2 15.0 78.8 73.7 60.9 b
Appearance 3.8 15.0 81.2 71.0 69.6 c
Speech 2.5 17.5 80.0 68.3 65.2 b
General Health 3.8 20.0 76.2 75.8 67.4
Breath 7.5 23.8 68.7 66.1 53.3
Comfort/Relaxation 2.5 35.0 62.5 62.4 51.1 c
Psychological Sleep 5.0 27.5 67.5 53.2 56.5
Confidence 10.0 15.0 75.0 55.4 69.6
Worry 5.0 33.8 61.2 53.2 46.7 b
Mood 8.8 33.7 57.5 48.4 42.4 b
Personality 7.5 23.8 68.7 61.3 57.6
Social Social life 7.5 25.0 67.5 67.2 62.0
Romantic Relationship 7.5 33.8 58.7 64.0 58.7
Smiling 7.5 21.2 71.2 69.4 70.7
Work 5.0 38.8 56.2 46.2 43.5 c
Finance 0.0 53.8 46.2 34.4 32.6 b
bStatistically significantly different compared to those reporting very good/good, (P < 0.05) but never visited a dentist; cClose to being significant at the 0.05 level.
Table 4 shows the response to impact dimension of oral health on QoL among the participants in the study. Overall, a large proportion of participants reported that oral health had no impact on their daily (54%) or social activities (64%) or in talking to other people (75%). The response pattern was similar between men and women except that a substantial proportion (48%) of women reported 'little" effect on daily activities compared to 34% men.
Table 4 Attributes of the "impact" attributes of oral health related quality of life of study populations (n = 356).
Response (percent) of impact dimension
Group Domain/attributes Extreme Great Moderate Little none
Total (n = 356) Daily activities 1.1 2.8 0.8 41.7 53.6
Social activities 0.8 3.6 1.7 30.5 63.7
Talking to people 0.6 2.8 0.8 20.4 75.4
Men (n = 161) Daily activities 1.2 3.1 1.2 34.2 60.2
Social activities 0.6 3.7 2.5 27.9 65.2
Talking to people 0.6 3.7 0.6 21.7 73.3
Women (n = 195) Daily activities 1.0 2.5 0.5 47.7 48.2
Social activities 1.0 3.0 1.0 32.5 62.4
Talking to people 0.5 2.0 1.0 19.3 77.2
The overall multivariable ANOVA model that included age, visits to a dentist, sex, education, ethnicity and employment, did not suggest statistically significant differences between groups (F = 1.22 Pr>F = 0.2788). However, a model including sex and education only suggested between group differences (F = 3.49, Pr>F = 0.0315). In this model, type-3 p-values (variables added last test) for education and sex were 0.0109 and 0.4927 respectively. Therefore we finalized our ANOVA model to one containing education only (F = 6.52, pr>F = 0.0111). The mean of effects sum score for those with secondary/tertiary education (mean = 61.8; 95% CI: 60.5, 62.9) was significantly higher than those with less than secondary/tertiary level of education (mean = 57.2; 95% CI: 57.2, 60.6).
Discussion
To the best of our knowledge, this study is the first attempt at providing some insights into how adults in Nigeria perceive the effect of oral health on their QoL. However, this study may have limitations that may influence its interpretation and generalizability arising from the use of a convenience sample that does not represent the Nigerian adult population. An earlier version of the OHQoL-UK© [13] instrument was used for this study. The effects sections of both instruments were the same, but the impact section of the current OHQoL-UK© had more questions than the version used in this study. However, the main analyses in this study were made around the effects attributes of OHQoL-UK©.
Most of the study participants had secondary/tertiary education. This could have occurred because one of the study sites was within a university teaching hospital community. Anecdotal evidence suggests that there is difficulty in getting participants with primary education to participate in oral health questionnaire survey in these communities, especially when they perceive that they are not particularly at risk for dental disease. A study conducted in Nigeria assessing the association of socio-demographic factors and edentulism in an adult population had mostly tertiary educated participants. [34] The authors stated that one of the possible reasons for the high numbers of tertiary educated participants in their study could be because these groups of people are more informed about their oral health needs and are also more likely to seek dental treatment. [34]. In addition, tertiary-educated persons are expected to be able to afford dental service, should have better access to adequate dental care, and have better than average oral health habits. [34]
Results from this study were similar to what was reported from the study on the perception of dental esthetics between students from Nigeria and the United States of America. [25] Most participants in this study felt that their oral health had an effect, mostly a good or very good effect on their QoL, similar to earlier studies from developed countries. [34,35] Studies conducted in Nigeria have also reported that there appear to be a shortage in the number of practicing dentists, [36] and the involvement of physicians untrained in oral health care providing dental services. [28] Another study reported poor oral care and poor oral health awareness/knowledge in their study population. [29] Despite, these results study participants still rated their oral health as having an affect on their Qol.
Intriguingly, a larger proportion of participants who had never visited a dentist rated the effects of oral health on their QoL as very good/good compared to those who had visited a dentist either within the past year or earlier, or both. A possible explanation for this paradoxical observation could lie in the oral healthcare seeking behavior of Nigerian adults. If people generally visit a dentist only if there is a severe oral condition requiring immediate attention as reported in our earlier study [26] then those visiting a dentist would be an orally less healthy group compared to those who never visited a dentist. This latter group would have a better self-perceived oral health status, and consequently report higher OHRQoL attributes. This phenomenon, which we term "healthy person non-visitor effect", is perhaps similar in conceptual moorings to the well-known "healthy volunteer effect".
Observations from this study are suggestive of the "healthy person non-visitor effect" because those who have never been to a dentist consider themselves not in need of oral health care and perceive their oral health to be better and therefore do not see the need to visit a dentist. The fact that a large proportion of participants reported that their visiting a dentist depends upon a perceived need for treatment also lends credence to this idea. Another possible explanation for reporting very good/ good effect of oral health on QoL by those who have never visited a dentist may be derived from how they prioritize oral health in comparison to general health. Such self-reports could have resulted from treatment-need which is based on oral health care seeking behavior that imparts a false perception of good oral health status.
If this were to be true then a visit to a dentist would tend to lower self-rated oral health status because the person would become aware of oral problems which they were originally unaware of and could lead to fewer of such participants rating their oral health or its effects highly. In either case, this self rating of effect of oral health on QoL appears to be distal to fundamental attitudes guiding oral health seeking behavior. This study suggests a situation of low oral health awareness and a treatment need based health care seeking behavior. Okunseri et al. [26] reported that majority (88%) of the study participants could not afford dental treatment and 89% were not ready to spend money on it. [26] It therefore appears that the general oral health perception could be reported as "good", implying no self-perceived need for dental treatment, by default.
To support this argument, results from the study of the impact dimensions of oral health on QoL shows that more than 90% participants reported little or no impact of oral health in their daily activities, social activities or in their ability to talk to people. It has been reported that low dental care utilization was determined by age, sex and employment in the same group of participants suggesting that normative oral health care needs could be much higher than perceived needs. [26] In a study population of predominantly young persons with low oral healthcare utilization, high self perceived oral health status, poor oral health seeking behaviors observed here, the low impact attributed to oral health can be interpreted as resulting more from cultural and attitudinal factors than from an inherently healthy cohort.
Despite several limitations mentioned in this study, we have been able to describe important attributes of oral health seeking behavior and oral health quality of life factors of adults living in Benin City, Edo State, Nigeria. This study also identified some potential health care-seeking issues that might be important when considering how to promote oral health awareness and oral healthcare seeking behavior. The study could also be used by policymakers as a framework to develop appropriate oral health strategies to improve and maintain the oral healthcare of adults.
Conclusion
This study shows that while participants reported very good/good effects of oral health on their quality of life, they also reported that oral health had little/no impact on their QoL. This, along with low oral healthcare utilization, and treatment-oriented healthcare seeking behavior could reflect cultural attributes of the population.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
CO: conceived of the study, participated in study design, carried out the study, participated in the statistical analysis, writing and in the reviewing and in responding to all reviewers' queries.
AC: set up manuscript idea, performed the statistical analysis, participated in the writing, reviewing and responding to reviewers queries.
IL: participated in the writing, critique and reviewing
CM: conceived of the study, participated in study design, writing, critique and reviewing
All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
==== Refs
Slade SD Sanders A ICF Australian User guide V1.0: The ICF and oral health accessed November 19, 2004
Cohen L Jago J Towards the formulation of sociodental indicators International Journal of Health Services 1976 6 681 7 971976
Allen PF Assessment of oral health related quality of life Health and Quality of Life outcomes 2003 1 40 14514355 10.1186/1477-7525-1-40
Reisine S Dental disease and work loss J Dent Res 1984 63 1158 1161 6589281
Cushing AM Sheiham A Maizels J Developing socio-dental indicators-the social impact of dental disease Community Dent Health 1986 3 3 17 3516317
Atchison KA Dolan TA Development of the Geriatric Oral health assessment Index J Dent Educ 1990 54 680 87 2229624
Strauss RP Hunt RJ Understanding the value of teeth to older adults: influences on quality of life JADA 1993 124 105 10 8445136
Slade GD Spencer AJ Development and evaluation of the oral health impact profile Community Dent Health 1994 11 3 11 8193981
Locker D Miller Y Evaluation of subjective oral health status indicators Journal of Public Health Dentistry 1994 54 167 76 7932353
Leao AT Sheiham A The development of a socio-dental measure of dental impacts on daily living Community Dent Health 1996 13 22 26 8634892
Kressin N Spiro A 3rdBossé R Garcia R Kazis L Assessing oral health-related quality of life: findings from the normative aging study Med Care 1996 43 416 427 8614164 10.1097/00005650-199605000-00004
Adulyanon S Vourapukjaru J Sheiham A Oral impacts affecting daily performance in a low dental disease Thai population Community Dent Oral Epidemiol 1996 24 385 9 9007354
McGrath C Bedi R An evaluation of a new measure of oral health related quality of life OHQoL-UK(W) Community Dent Oral Epidemiol 2001 18 138 143
Badley EM The ICIDH: format, application in different settings and distinction between disability and handicap Int Disabili Stud 1987 9 122 5
McGrath C Bedi R Gilthorpe MS Oral health related quality of life-views of the public in the United Kingdom Community Dent Health 2000 17 3 7 11039623
WHO (World Health Organization) International classification of impairments, disabilities and handicaps 1980 Geneva, WHO
Locker D Slade GD Concepts of oral health, disease and quality of life Measuring Oral health and Quality of Life 1977 University of North Carolina
Soe KK Gelbier S Robinson PG Reliability and Validity of two oral health related quality of life measures in Myanmar adolescents Community Dent Health 2004 21 3006 311
Masalu J Åstrøm AN Applicability of an abbreviated version of the oral impacts on daily performances (OIDP) scale for use among Tanzanian students Community Dent Oral Epidemiol 2003 31 7 14 12542427 10.1034/j.1600-0528.2003.00011.x
Tsakos G Marcenes W Sheiham A Cross-cultural differences in oral impacts on daily performance between Greek and British older adults Community Dent Health 2001 18 209 21 11789697
John MT Koepsell TD Hujoel P Miglioretti DL LeResche L Micheelis W Demographic factors, denture status and oral health-related quality of life Community Dent Oral Epidemiol 2004 32 125 32 15061861 10.1111/j.0301-5661.2004.00144.x
McGrath Translation and validation of an Arabic version of the UK oral health related quality of life measure (OHQoL-UK) in Syria, Egypt and Saudi Arabia Community Dent Health 2003 20 241 245 14696744
Onyeaso CO Aderinokun GA The relationship between dental aesthetic index (DAI) and perceptions of aesthetics, function and speech amongst secondary school children in Ibadan, Nigeria Int J Paediatr Dent 2003 13 336 41 12924989 10.1046/j.1365-263X.2003.00478.x
Onyeaso CO An assessment of relationship between self-esteem, orthodontic concern, and Dental Aesthetic Index (DAI) scores among secondary school students in Ibadan, Nigeria Int Dent J 2003 53 79 84 12731694
Otuyemi OD Ogunyinka A Dosumu O Cons NC Jenny J Kohout FJ Jakobsen J Perceptions of dental aesthetics in the United States and Nigeria Community Dent Oral Epidemiol 1998 26 418 20 9870542
Okunseri C Born D Chattopadhyay A Self-Reported Dental Visits among Adults in Benin City, Nigeria International Dent J 2004 54 450 456
Kubota K Hollist NO Olusile AO Yonemitsu M Minakuchi S Watanabe H Ohsawa K Ohnishi M Ono Y Ajayi-Obe SO Joint epidemiological longitudinal dental survey in Nigeria, especially in comparison with that of Japanese Bull Tokyo Med Dent Univ 1993 40 59 78 8462119
Sa'adu ZO Abdulraheem IS Oral health care practice and socio-demographic findings among the physicians in Ilorin, Nigeria Niger J Med 2003 12 211 6 14768196
Sofola OO Agbelusi GA Jeboda SO Oral health knowledge, attitude and practices of primary school teachers in Lagos State Niger J Med 2002 11 73 6 12221964
McGrath C Bedi R Measuring the impact of oral health on quality of life in Britain using OHQoL-UK(W) J Public Health Dent 2003 63 73 77 12816136
McGrath C Bedi R Population based norming of the UK oral health related quality of life measure (OHQol-UK©) Br Dent J 2002 9 521 524 12572737
Statistical Package for the Social Science (SPSS) 1999 Release 10.0 for Windows, Chicago
Stokes Maura E Davis Charles S Koch Gary G Categorical Data Analysis Using the SAS® System 2000 Second Cary, NC: SAS institute Inc
US Public Health Service, National Center for Health Statistics Current estimates from the National Health Interview Survey, 1993 DHHS Publication No (PHS) 95-1518, Series 10 No 190 1994 Washington DC, Government Printing Office
US Public Health Service, National Center for Health Statistics Use of dental services and dental health. United States, 1986 DHHS Publication No (PHS) 88-1593, Series 10 No 165 1988 Washington DC, Government Printing Office
Ndiokwelu E Primary health care approach. Its relevance to oral health in Nigeria Odontostomatol Trop 2002 25 29 32
|
16042806
|
PMC1190187
|
CC BY
|
2021-01-04 16:29:57
|
no
|
BMC Oral Health. 2005 Jul 25; 5:7
|
utf-8
|
BMC Oral Health
| 2,005 |
10.1186/1472-6831-5-7
|
oa_comm
|
==== Front
BMC Palliat CareBMC Palliative Care1472-684XBioMed Central London 1472-684X-4-41608351310.1186/1472-684X-4-4SoftwareCareSearch: finding and evaluating Australia's missing palliative care literature Tieman Jennifer J [email protected] Amy P [email protected] Belinda S [email protected] David C [email protected] Department of Palliative and Supportive Services, Flinders University, Adelaide South Australia, Australia2 Southern Adelaide Palliative Services, Repatriation General Hospital, Daw Park, South Australia, Australia3 Division of Medical Oncology, Department of Medicine, Duke University Medical Centre, Durham, North Carolina, USA2005 8 8 2005 4 4 4 11 3 2005 8 8 2005 Copyright © 2005 Tieman et al; licensee BioMed Central Ltd.2005Tieman et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Palliative care is an evolving specialty with a growing evidence base. However, evidence is less accessible than it could be with a lower than average conversion of conference abstracts to articles in peer-reviewed journals and the need for more accessible tools to support evidence-based practice (EBP) in palliative care. The CareSearch project involved identifying, collecting and evaluating Australia's "grey" palliative care literature and identifying international published literature missing from the electronic indexing systems. The literature was then catalogued and made publicly available through the CareSearch website.
Results
To date over 2,500 items have been included in the CareSearch database and can be accessed and searched through a publicly available website. Nearly 2,000 items are conference abstracts and 178 are theses or government, organisational and planning documents. A further 410 items relate to articles from palliative journals that are not indexed on a major bibliographic database. The website also provides tools and facilities to support palliative care practice and research.
Conclusion
CareSearch is a new evidence resource for palliative practitioners, educators and researchers. The palliative community now has access to a more comprehensive literature base as well as a resource that supports the integration of knowledge into practice. This specialised data repository enables users to access information on the body of work that has shaped palliative care development and prevents the potential loss or duplication of research work. It also provides a template for other emerging disciplines to use in capturing their literature and evidence.
==== Body
Background
Palliative care has only been a distinct academic discipline in Australia since the 1980s, with an emerging publication base. Defining the work already done that contributes to evidence-based practice and research directions is time consuming and difficult. This problem is shared with many clinical disciplines especially those that are relatively young such as rehabilitation, geriatric medicine, sports medicine and sexual health.
The role of evidence in the practice of health care is built on a number of concepts relating to structured use of information, an evaluation of material and the use of best available evidence. Sackett and colleagues (1996) define evidence-based practice as follows:
Evidence based medicine is the conscientious, explicit, and judicious use of current best evidence in making decisions about the care of individual patients. The practice of evidence based medicine means integrating individual clinical expertise with the best available external clinical evidence from systematic research (P71) [1].
As an emerging field, the early literature was not always captured within the peer-reviewed literature and is still difficult to access. There is a perception among researchers, clinicians and funders that the amount of information available in an accessible form does not reflect the research and planning that has occurred. Poor accessibility is worsened by:
• Presentation of work at conferences without preparation of a subsequent peer reviewed manuscript;
• Researchers who do not communicate their work in any forum;
• The breadth of clinical scope for good palliative care [2];
• Slow emergence of specific peer-reviewed publications as a new discipline develops;
• Inconsistent cataloguing of the palliative care literature in major electronic bibliographic databases; and,
• Difficulties in searching for palliative care literature due to sub-optimal search strategies.
These problems make the palliative care evidence pool appear less comprehensive than it is.
For example, the journal Palliative Medicine began in 1987 but was not indexed in MEDLINE or EMBASE until 1992; Death Studies was indexed in MEDLINE until 1980 and then not reliably indexed again until 1997 while it was variably indexed in EMBASE from 1984 onwards. After indexing on electronic bibliographic databases, the variability in indexing can be seen when citations accessed in major electronic bibliographies are compared with hand searches [3,4].
For a field such as palliative care, issues around evidence are made more difficult by the multidisciplinary and multispecialty nature of the clinical care. The list of relevant journal resources spans all specialties and sub-specialties, just as prognostication and care of patients with life-limiting illness span all diseases. Limiting the palliative care resources to those published in palliative medicine sub-specialty journals severely limits the breadth of information available to inform best practice. Articles relevant to palliative care can be found in sources from gastroenterology, respiratory, cardiology, surgery, nursing and paediatrics specialty journals to the health education, health promotion and occupational therapy literature. Relevant information can also be found in the literature of non-bioclinical disciplines such as pastoral care and social work. This creates additional hurdles in locating and evaluating pertinent literature [5].
Grey literature is defined as "that which is produced on all levels of government, academics, business and industry in print and electronic forms but which is not controlled by commercial publishers" [6]. It includes materials such as dissertations, census and statistical data, reports of research (completed and uncompleted), and technical reports. It does not mean that grey literature is not peer reviewed and indeed it may have undergone extensive review. Citations of these works are usually left out of the major bibliographic databases.
This "missing" literature and grey literature is particularly significant in emerging fields such as palliative care where government-funded scoping studies, organisation-sponsored reports and other commissioned work often most accurately reflect the evolving state of current knowledge [5,7,8].
The complications of searching the palliative care literature – including incomplete indexing, multiple disciplines, and the grey literature – are amplified by the problems inherent to searching in any subject area. Reliance on a single electronic bibliographic database can reduce the ability to identify relevant articles. Various studies have shown that it is important to identify and search the bibliographic databases most relevant to the topic to improve the recall of target studies [9,10].
Electronic searching alone also reduces the results attained. Savoie and colleagues [11] demonstrated that extended searching beyond the major databases improved the identification of randomised controlled trials when compared to bibliographic index searching alone. Searching specialised databases like CareSearch and trial registers were the most important additional strategies after searching electronic bibliographic databases.
In order to be identified through any search, a basic requirement is that the author must have published the work in some forum. A recent review of studies dealing with acceptance rates and subsequent publication rates for biomedical abstracts found a conversion rate, from presentation to publication in full, of around 45% [12]. There is the perception that many researchers in the discipline of palliative care have presented their work at conferences, but not subsequently submitted manuscripts for peer-reviewed publication. Preliminary estimates from a review of CareSearch abstracts suggest a conversion rate of less than 20%. This problem is not confined to palliative care; the National Institutes of Health (NIH) found that only 80% of the studies they funded were published [13].
While the lack of publication may reflect that many presentations were not based on rigorous studies and hence did not meet publication criteria, for an evolving field the loss of conceptual work or preliminary studies remains a significant loss if it fails to inform the development of future work. Further, at times the best evidence will be grey literature as it is the only evidence that exists.
Publication bias, whether institutionalised or self-regulated by authors, has significant effects on the assessment of interventions. Exclusion of this literature (i.e. studies that are not published, have a limited distribution or are not included in bibliographic retrieval systems) can lead to a significant overestimation of an intervention's effect [14,15].
Systematic reviews can only claim to be complete if the process involves systematic searching of published data including conference abstracts [11]. The savings of having one source for searching for these items are manifest, if such a collection continues to be updated.
In an effort to address some of these problems outlined above as they relate to palliative care, the Rural Health and Palliative Care Branch of the Australian Department of Health and Ageing provided $580 000 over 5 years from the National Palliative Care Program to fund the Evidence Based (CareSearch) Project. Flinders University was contracted by the Australian Government to undertake the work. The University's Department of Palliative and Supportive Services has managed and executed the project.
The project has three specific aims:
• To capture and collate Australia's "missing" palliative care evidence and the international "missing" published literature in palliative care;
• To make this literature accessible to inform best practice; and
• To promote evidence-based practice (EBP) in palliative care through an electronic cyber-community.
The project is commonly called "CareSearch", reflecting its virtual home at the CareSearch website [16] and its focus on identification of literature relevant to palliative care.
Implementation
Project management
A National Reference Group of thirteen clinicians and researchers volunteer their time to oversee the project. They contribute their substantial skills in EBP, palliative care, general clinical care, information systems and research evaluation. The National Reference Group functions as an Editorial Board meeting twice a year to advise on directions and policy issues. The Project Team translates these broad directions into operational activities. The Project Team comprises the three part-time project staff members who carry out the work and the two investigators who act as a local resource and as advisors. The National Reference Group also assists by evaluating conference abstracts and by providing reviews for the "Hot Picks" section.
Project parameters
The National Reference Group has provided advice on the issues and processes around the inclusion and assessment of materials selected for the literature databases. The considerations around the inclusion process recognised that often there is no randomised controlled trial or "gold standard" in the literature to address the clinical questions in palliative care and that clinicians may need to consider the "evidence pyramid" and look for the best available evidence. For some areas in palliative care there may not be any "formal" evidence available to support clinical judgement and as is the case with other areas of clincial practice, even may not relate to "this" patient. Evidence issues are compounded by the multidisciplinary nature of palliative care that draws on different methodological approaches and research paradigms. As a result the primary emphasis of the project was on identifying sources of information, evidence and commentary that could add to an evidence base given the broadest definition of "best available evidence".
In conjunction with the CareSearch Project Team, the National Reference Group identifies literature sources and developed a schema to organise and evaluate the collected materials and data. Information sources include those not available through the major electronic bibliographies such as abstracts from conference proceedings; reports and other literature from federal, state, and territory government departments and from palliative organisations; theses and treatises; and missing published literature representing non-indexed articles from palliative care journals (before the date of first citation and unpredictable omissions since that date). Collection and evaluation of material is an ongoing function of the project.
Abstracts and conference proceedings
The National Reference Group identifies conferences held in Australia since 1980 that may include issues relevant to palliative care. The sponsoring organisation for each conference is contacted and permission sought to collect a copy of the conference proceedings or abstract book, copy pertinent abstracts, and add them to the CareSearch library. Once the abstract books are obtained, two National Reference Group members separately review each book and select relevant abstracts about the care of people with advanced life-limiting illness regardless of diagnosis. CareSearch continues to collect and evaluate new conference abstracts from each relevant conference or scientific meeting presented around Australia.
For level of evidence evaluation, a purpose driven proforma was developed [See additional file 1]. Using the proforma, evaluators record information about the study, identify key words, and decide whether data are presented in the abstract. The evaluation schema also looks at study question, study design, validity of the conclusions and generalisability of the results to clinical practice. For many abstracts the conference abstract evaluation is limited due to the limited information provided in the abstract. Evaluators must also exercise judgement with regard to indicating what could be the best study design to address the topic being covered by the conference presentation. However, this schema provides broad guidance for users regarding the quality of the work reported in the abstract and hence provides a caution about its possible application in practice.
Two reviewers independently rate each abstract and give it a final A, B or C classification. If the reviewers do not agree on the classification, a blinded third reviewer rates the abstract and the final classification is that which receives two of three nominations. The abstract and reviews are stored in the CareSearch database. If an author disagrees with the final classification, he or she can contact CareSearch directly to resolve the issue; the classification information will be removed from the website and the abstract will be sent out to a new pair of National Reference Group members for review.
Reports & treatises
Listings of state, territory and federal departments with responsibility for palliative care activities were prepared together with a listing of national organisations with palliative interests and universities within Australia. All of these contacts have been approached in writing with follow-up phonecall for contributions to the CareSearch database. They have been informed of the nature of the project and permission has been obtained to provide the document abstracts or executive summaries on the CareSearch website. These organisations nominate documents for inclusion in CareSearch and the National Reference Group then reviews these selections for relevance. These documents are not further evaluated.
Academic institutions offering higher degrees in areas associated with palliative care were identified. The project then contacted the academic institution to nominate any Masters or PhD theses or academic treatises that are palliative in content. The nominated items are reviewed by the National Reference Group for relevance. The documents are entered into the literature database without further evaluation.
Non-indexed journal articles
The National Reference Group formed a consensus opinion on twelve key journals for palliative care from an initial list of 51 journals seen to be relevant to palliative care. The journals were handsearched from the first published volume until July 2002 to identify all articles describing original research or significant reviews in the field. Ovid MEDLINE, EMBASE, CINAHL and PsycINFO were then searched for all of the identified articles. Reviews since July 2002 are ongoing. The "missing" published palliative care literature is defined as those articles published (as evidenced by handsearching) but not indexed in the major biomedical bibliographic databases. The publishers of the palliative care journals have been contacted, and if they agree, abstracts for the "missing" articles are added to the CareSearch library. A listing of the palliative care journals and those who agree to have their abstracts included is available on the CareSearch website.
Database and website
A structured database houses all CareSearch data with the results available on a user-friendly public-access website. The website ensures systematic delivery of the key outcomes for all people who need to access this work. The database was built using Microsoft Access 2000 (Microsoft Corporation, Seattle, Washington, USA) and an interactive website was designed using Microsoft Visual Basic .NET and Microsoft SQL Server 2000. The website can be easily updated by project administrative staff using a content management system. The website layout and database access is being improved based on comments solicited from members of the National Reference Group and website users.
The primary function of the website is to house the CareSearch library and permit user access. The search engine allows unrestricted searches by any word (not limited to key words) and specific searches by author, database (e.g. conference abstract) or year. Boolean logic is used to refine searches.
Results
Abstracts and conference proceedings
To create the conference abstracts database, 25 conference organisations were approached in the initial investigation. All organisations supplied books. One hundred and eleven books of conference proceedings were reviewed and 1,690 conference abstracts were assessed as being relevant to palliative care. Additional abstracts have been added as part of the annual update cycle and as other relevant conferences are identified. Over 20% of these conference abstracts to date have been formally evaluated.
Other literature
To date, 100 government and organisation sponsored documents have been located and included in the database. In addition, 78 theses from 14 Australian universities have been catalogued.
Non-indexed articles
Twelve palliative care journals were reviewed from initial publication to July 2002. These journals were:
• Palliative Medicine
• American Journal of Hospice and Palliative Care
• Journal of Pain and Symptom Management
• Journal of Palliative Medicine
• Journal of Palliative Care
• International Journal of Palliative Nursing
• Progress in Palliative Care
• European Journal of Palliative Care
• Death Studies
• Supportive Care in Cancer
• Hospice Journal
• Psycho-Oncology
A total of 8,398 items were identified, with 7,557 (90%) indexed in one of the four main bibliographic databases. Of the 841 non-indexed items, 410 (49%) have been identified as research or commentary and been included in the CareSearch database. The other 431 items were not included as they were not deemed to be articles. They comprised excluded items such as book reviews, video reviews or conference summaries.
Other evidence based resources
Based upon feedback from National Reference Group members and users, CareSearch is developing and updating multiple resources that support and foster EBP in palliative care. Current resources include a monthly "Hot Pick" literature review of a recent publication, a research resource area including a specialised platform to promote data management in multi-site clinical studies, palliative care audit tools, and a "search strategy generator" that facilitates efficient MEDLINE searching and generates up to date resources saving the website administrator the need to constantly review the currency of materials.
Members of the National Reference Group have agreed to provide reviews for the Hot Picks. They select an article from the literature that they believe is significant to the practice of palliative care and provide a written review highlighting its relevance.
There are currently eleven searches on palliative care topics available on the website. A further ten will be written in the coming year. Additional pages for the Services Models and Clinical Practices section will also be added in the coming year.
The CareSearch Research Platform has been developed to support research work within palliative care by providing access to a tool that
• enables the online design of data collection forms and questionnaires;
• allows for web-based and email-based form completion through the CareSearch website;
• enables data entry from multiple sites with a single co-ordinating research site;
• provides for basic reporting of results with features such as percentages, graphs, and tables; and
• allows export of data to other programs such as Excel, Access or SPSS.
With a user-friendly interface, the tool encourages beginning researchers and small services who may not have access to statistical and research resources as well as supporting larger multi-site research activities. The capacity to webhost the studies to allow online data entry from multiple sites is particularly beneficial to agencies that do not have access to such resources. This platform has already been used to collect data for several projects including an international multisite clinical trial. Given specific challenges in participant accrual in palliative studies a tool available internationally for multisite research is critical to generate better quality evidence in clinical areas.
The project has only just begun to look at the issues associated with the development of online communities that could play a significant role in knowledge dissemination and translation. New project directions are being informed by the developing body of literature relating to communities of practice and knowledge translation [17-20]. Bulletin boards and targeted resources and tools for special interest groups will be introduced in the coming year.
Database and website
In total over 2,500 items that are missing from the formally indexed palliative care literature have already been located, evaluated, catalogued and combined into the CareSearch database. This continues to grow. Access to the database items is through the CareSearch website [16]. There has been consistent expansion in the use of the website since it became available in March 2004 (See Table 1). The CareSearch website tracks usage through the LiveSTATS.XSP log analyser (LiveStats Service Provider Edition 6.2). While care needs to be exercised in the interpretation of website statistics, they provide a useful indicator of site usage. The Visits report shows the number of visitor sessions to CareSearch during a specified period. A visit refers to a series of requests from a uniquely identified client. Website statistics show a three-fold increase in the number of visits to the site in the eleven months to January 2005, with a high level of repeat use, suggesting that users find the material relevant and useful. The increasing number of distinct internet service providers (ISPs) and companies visiting the site suggests that new users are able to locate the website.
Table 1 Visitor statistics: CareSearch from March 2004 – January 2005
Month Total visits Distinct ISPs or companies visiting the site
March 04 413 77
April 04 360 90
May 04 636 144
June 04 925 153
July 04 1047 173
August 04 1137 216
September 04 1190 252
October 04 868 221
November 04 1065 302
December 04 1381 265
January 05 2051 275
Discussion
The volume of data and materials identified supports the initial perception that there was a large "missing" literature not previously collated. CareSearch provides access to this "missing" non-indexed palliative care literature, complements the existing bibliographic databases and extends the coverage of palliative materials. Literature formerly only available through extensive handsearching, which could be of great significance in systematic reviews and meta-analyses, is now electronically available [21]. Over 2,500 items are stored in the CareSearch library. Inadequate access to this "missing" literature allowed a knowledge deficit and left the research community at risk of repeating completed work rather than building on existing knowledge. CareSearch is a critical platform for rectifying these problems, facilitating EBP and highlighting future research directions. The website's structure and features support EBP and the development of critical skills for identification and appraisal of information through access to search strategies, the inclusion of evidence hierarchies, audit tools and current literature summaries.
The palliative care community now has access to a more comprehensive literature base as well as a resource that supports integration of knowledge into practice. As much of the collected material falls within the concept of "grey literature" it may not have been subject to the same peer-reviewed process as published literature and may therefore be more variable in terms of evidence standards. However, this material captures not only the developmental history of the discipline, but also some of the knowledge that is missing from the formal peer-reviewed bibliographic databases for many different reasons and which can add value to future research and current practice.
For the discipline itself, CareSearch has systematically demonstrated that there is an Australian evidence base contributing to palliative practice and highlights this output in a public forum. The project has also provided a template and model for other emerging disciplines to capture their research and commentary to support the progress of the field.
The CareSearch approach to comprehensive coverage of the Australian palliative care information also facilitates identification of gaps in knowledge. This can promote thoughtful research questions and more appropriate use of existing data to support developments in clinical research including data for new studies' power calculations and recruitment, clinical care, service planning, funding and practice change. There is also the capacity to build linkages between researchers and encourage collaborative Australian projects building on existing local data.
The non-indexed published literature portion of the library reflects worldwide input. As the CareSearch team seeks to expand coverage more widely outside of the Pacific region, it will need to partner with other organisations and interested parties to build an understanding of the total state of current knowledge in palliative care including gaps and future directions.
The role of online literature and evidence in supporting clinical practice and research within Australia has been highlighted in the recent evaluation study of the Clinical Improvement Access Program [22]. This study found patterns of use of online resources that increased with patient admissions. This pattern coupled with self-report of use by clinicians showed that health professionals were using online evidence to support clinical decision-making. Clinicians, educators, researchers and healthcare planners in palliative care now have similar benefits through access to assembled and evaluated clinical and service resources.
Unpublished abstracts and other literature contribute to the knowledge base and need to be considered within the wider context of evidence. Randomised controlled trials are rare in palliative care and it is important to keep in mind that best evidence is the best established information available to answer the question at hand [1,23]. To date, less rigorous studies have figured more prominently in palliative care decision making than in other disciplines because that is what is available. Highlighting the current state of evidence through projects like CareSearch may support the arguments for further research [5].
To support continuing clinical and practice improvements, it is important that the existing databases retain their currency. The databases must continue to capture the new evidence and material as they are released to ensure that CareSearch is a "living" anthology. The website will also need to expand to incorporate the identified and emerging needs of the palliative care community with regard to information and evidence. The National Reference Group and the local CareSearch Project Team have already identified a number of future directions, including:
• Exploring mechanisms to evaluate and refer to websites themselves given their increasing influence as an information source;
• Increasing the reach of the project by actively engaging the wider palliative community with the contents and philosophy of the CareSearch website;
• Including abstracts from palliative related journals that are not indexed in the common electronic bibliographic systems;
• Building the critical appraisal capacity of the palliative care community by widening participation in the evaluation of conference abstracts;
• Validating search filters to support effective searching for palliative care information within the general medical and health literature;
• Developing descriptors to enable generalisability of results across palliative patient populations;
• Investigating the use of the website as a benchmarking and audit service for palliative care services;
• Increasing the usefulness of the Research platforms by creating template surveys and including formatted tools for use in research;
• Partnering with palliative care communities around the globe to foster evidence-based palliative care internationally; and
• Using CareSearch as a template for other disciplines to develop their research and evidence databases.
For other groups who may be interested in using this approach to consolidate the knowledge and research materials for their discipline, it is important to be realistic about the time and efforts required to source and assess the materials. This type of project relies heavily on the goodwill and expertise of those involved in a volunteer capacity. It also requires substantial expertise and management skills to develop not only the internal processes but also carry out the negotiations with institutions, agencies and technology partners and meet the various legislative requirements including copyright, privacy and intellectual property.
Conclusion
The palliative care literature readily available through the bibliographic indexing system may be less comprehensive than the actual pool of research, review and opinion. By identifying possible missing data and reviewing these materials, the Evidence-Based (CareSearch) Project has brought together not only Australia's palliative care literature but also key concepts relating to palliative care practice, planning and research. By embedding access to missing evidence resources within a user-friendly accessible structured framework, the CareSearch website actively promotes contemporary EBP within the palliative care community.
Availability and requirements
• Project name: Evidence based (CareSearch) Project
• Project home page:
• Operating system(s): Windows 2003
• Programming language: MS VB .Net, MS C#, ASP
• Other requirements: Client – Internet Explorer 5.5 or later, Server – Minigen Content Manager
• License: Not Applicable
• Any restrictions to use by non-academics: None except for the Research Platform which requires registration with the CareSearch project
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
DC and AA are co-investigators for the project. DC, AA, JT and BF were involved in the location and evaluation of the materials for inclusion and in the development of the websites and its resources. All authors have had intellectual and editorial input to this manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Supplementary Material
Additional File 1
Conference Abstract Evaluation. Additional file 1 contains the schedule used by reviewers for the evaluation of conference abstracts entered into the CareSearch database. It also details the instructions and procedures for abstract evaluators and lists the review areas for the conference abstract evaluation.
Click here for file
Acknowledgements
The authors graciously thank Ms. Rita Evans, Ms. Julie Mueller, and Mr. Steve Dunlop and the Palliative Care Branch of the Australian Department of Health and Ageing. We also thank the members of the CareSearch National Reference Group including Professor Sanchia Aranda, Associate Professor Paul Glare, Professor Carol Grbich, Ms. Meryl Horsell, Professor Linda Kristjanson, Associate Professor Geoff Mitchell, Professor Paddy Phillips, Dr. Odette Spruyt, and Professor Patsy Yates, and the previous CareSearch project manager, Mr. Timothy Butler.
Direct costs of this study were provided through a grant from the Rural Health and Palliative Care Branch of the Australian Department of Health and Ageing (Canberra, Australia), under the National Palliative Care Strategy. Dr. Abernethy's salary is generously provided through a Clinical Scientist Development Award from the Doris Duke Charitable Foundation of New York, New York, USA. The design, conduct, analysis and write-up of the study were performed independently from any funding or sponsoring agency.
==== Refs
Sackett DL Rosenberg WM Gray JA Haynes RB Richardson WS Evidence based medicine: what it is and what it isn't BMJ 1996 312 71 72 8555924
Currow DC Abernethy AP Quality palliative care: practitioners' needs for dynamic lifelong learning J Pain Symptom Manage 2005 29 332 334 15857734 10.1016/j.jpainsymman.2005.02.005
Abernethy AP Currow D Butler T Searching electronic bibliographic databases for palliative care articles 5th Conference and Scientific Meeting of the Australian and New Zealand Society of Palliative Medicine 2002 Townsville, Australia
McDonald S Lefebvre C Antes G Galandi D Gotzsche P Hammarquist C Haugh M Jensen KL Kleijnen J Loep M Pistotti V Ruther A The contribution of handsearching European general health care journals to the Cochrane Controlled Trials Register Eval Health Prof 2002 25 65 75 11868446 10.1177/0163278702025001005
Higginson IJ Evidence based palliative care. There is some evidence - and there needs to be more BMJ 1999 319 462 463 10454384
Albernani V Pietrangeli PD Mazza AM The use of grey literature in health sciences: a preliminary survey Bulletin of the Medical Libraries Association 1990 78 358 363
Cook AM Finlay IG Edwards AG Hood K Higginson IJ Goodwin DM Normand CE Douglas HR Efficiency of searching the grey literature in palliative care J Pain Symptom Manage 2001 22 797 801 11532593 10.1016/S0885-3924(01)00315-3
Crawford MJ Rutter D Manley C Weaver T Bhui K Fulop N Tyrer P Systematic review of involving patients in the planning and development of health care BMJ 2002 325 1263 12458240 10.1136/bmj.325.7375.1263
Suarez-Alamor ME Belseck E Homik J Dorgan M Ramos-Remus C Identifying clinical trials in the medical literature with electronic databases: MEDLINE alone is not enough Control Clin Trials 2000 21 476 487 11018564 10.1016/S0197-2456(00)00067-2
McDonald S Taylor L Adams C Searching the right database: A comparison of four databases for psychiatry Health Libr Rev 1999 16 151 156 10620849 10.1046/j.1365-2532.1999.00222.x
Savoie I Helmer D Green CJ Kazanjian A Beyond Medline: reducing bias through extended systematic review search Int J Technol Assess Health Care 2003 19 168 178 12701949 10.1017/S0266462303000163
Von Elm E Constanza MC Walder B Tramer MR More insight into the fate of biomedical meeting abstracts; a systematic review BMC Med Res Methodol 2003 3 12 12854973 10.1186/1471-2288-3-12
Dickersin K Min YL NIH clinical trials and publication bias Online Journal of Current Clinical Trials 1993 Doc No 50 4967
McAuley L Pham B Tugwell P Moher D Does the inclusion of grey literature influence estimates of intervention effectiveness reported in meta-analyses? Lancet 2000 356 1228 1231 11072941 10.1016/S0140-6736(00)02786-0
Egger M Davey-Smith G Misleading meta-analysis; lessons from an 'effective, safe, simple' intervention that wasn't BMJ 1995 310 752 754 7711568
CareSearch Website
Wenger E Knowledge Management is a donut shaping your knowledge strategy through communities of practice Ivey Business Journal 2004
Davis D Evans M Jadad A Perrier L Rath D Ryan D Sibbald G Straus S Rappolt S Wowk M Zwarenstein M The case for knowledge translation: shortening the journey from evidence to effect BMJ 2003 327 33 35 12842955 10.1136/bmj.327.7405.33
Amos A Goodrick D Understanding Communities of practice and how they apply to the health context – A summary RMIT University Circle
Russell J Greenhalgh T Boynton P Rigby M Soft networks for bridging the gap between research and practice: illuminative evaluation of CHAIN BMJ 2004 328 1174 15142924 10.1136/bmj.328.7449.1174
Conn VS Valentine JC Cooper HM Rantz MJ Grey literature in meta-analyses Nurs Res 2003 52 256 261 12867783 10.1097/00006199-200307000-00008
Westbrook JI Gosling AS Coiera E Do clinicians use online evidence to support patient care? A study of 55 000 clinicians J Am Med Inform Assoc 2004 11 113 120 14662801 10.1197/jamia.M1385
Rinck GC van den Bos GA Kleijnen J de Haes HJ Schade E Veenhof CH Methodologic issues in effectiveness research on palliative cancer care: a systematic review J Clin Oncol 1997 15 1697 1707 9193371
|
16083513
|
PMC1190188
|
CC BY
|
2021-01-04 16:30:52
|
no
|
BMC Palliat Care. 2005 Aug 8; 4:4
|
utf-8
|
BMC Palliat Care
| 2,005 |
10.1186/1472-684X-4-4
|
oa_comm
|
==== Front
BMC Pregnancy ChildbirthBMC Pregnancy and Childbirth1471-2393BioMed Central London 1471-2393-5-121604277310.1186/1471-2393-5-12Research ArticleAntenatal screening for Group B Streptococcus: A diagnostic cohort study Hiller Janet E [email protected] Helen M [email protected] Philip [email protected] Caroline A [email protected] Department of Public Health, University of Adelaide, Adelaide, Australia 50052 Emeritus Microbiologist, Microbiology & Infectious Diseases Department, Women's & Children's Hospital, 72 King William Road, North Adelaide, South Australia, Australia 50063 Department of Nursing & Midwifery Research & Practice Development, 2nd floor, Samuel Way Building, Women's & Children's Hospital, 72 King William Road, North Adelaide, South Australia, Australia 50064 Department of Obstetrics and Gynaecology, University of Adelaide, Adelaide, Australia 50052005 22 7 2005 5 12 12 24 1 2005 22 7 2005 Copyright © 2005 Hiller et al; licensee BioMed Central Ltd.2005Hiller et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
A range of strategies have been adopted to prevent early onset Group B Streptococcal (EOGBS) sepsis, as a consequence of Group B Streptococcal (GBS) vertically acquired infection. This study was designed to provide a scientific basis for optimum timing and method of GBS screening in an Australian setting, to determine whether screening for GBS infection at 35–37 weeks gestation has better predictive values for colonisation at birth than screening at 31–33 weeks, to examine the test characteristics of a risk factor strategy and to determine the test characteristics of low vaginal swabs alone compared with a combination of perianal plus low vaginal swabs per colonisation during labour.
Methods
Consented women received vaginal and perianal swabs at 31–33 weeks gestation, 35–38 weeks gestation and during labour. Swabs were cultured on layered horse blood agar and inoculated into selective broth prior to analysis. Test characteristics were calculated with exact confidence intervals for a high risk strategy and for antenatal screening at 31–33 and 35–37 weeks gestation for vaginal cultures alone, perianal cultures alone and combined low vaginal and perianal cultures.
Results
The high risk strategy was not informative in predicting GBS status during labour. There is an unequivocal benefit for the identification of women colonised with GBS during labour associated with delaying screening until 36 weeks however the results for method of screening were less definitive with no clear advantage in using a combined low vaginal and perianal swabbing regimen over the use of a low vaginal swab alone.
Conclusion
This study can contribute to the development of prevention strategies in that it provides clear evidence for optimal timing of swabs. The addition of a perianal swab does not confer clear benefit. The quantification of advantages and disadvantages provided in this study will facilitate communication with clinicians and pregnant women alike.
==== Body
Background
Group B Streptococcus (GBS) infection in infants as a consequence of vertically acquired infection, is an important cause of neonatal mortality and morbidity, presenting as sepsis or pneumonia [1]. The incidence of early onset group B streptococcus sepsis (EOGBS) occurring within the first week of life has fallen in Australia from 2.0 per 1000 live births in 1991–1993 to 0.5 per 1000 live births in 1995–1997 [2]. This figure is similar to the recently reported annual incidence of 0.48 per 1000 from the United Kingdom and Ireland [3].
Vaginal colonisation occurs in 11–30% of all pregnant women [4-6] and 50–75% of their infants become colonised usually during labour or birth. There is clear evidence that intrapartum colonisation is strongly associated with EOGBS sepsis [7] which has a case-fatality of approximately 4%[1]. Serious morbidities include sepsis, pneumonia, meningitis, osteomyelitis or septic arthritis.
The United States' Centers for Disease Control has endorsed a strategy in which screening of pregnant women is to occur at 35–37 weeks gestation using vaginal and rectal swabs and all women delivering before 37 weeks are to be treated if they are of GBS culture positive or of unknown GBS status, a change from their previous policy in which a strategy of intrapartum chemoprophylaxis based on a risk-based approach also was endorsed [8]. This contrasts with the 2003 recommendation from the Royal College of Obstetricians and Gynaecologists which states that "routine screening (either bacteriological or risk based) for antenatal GBS carriage is not recommended" [9]. There is no standard accepted approach to the prevention of EOGBS. Strategies have evolved including screening antenatally to detect colonisation or treatment of women with risk factors including prolonged rupture of membranes, intrapartum fever, preterm labour and history of maternal colonisation during pregnancy reflecting in part, the impact of local data on the burden of GBS.
Within Australia there is considerable variation in clinical practice in both the prevention of GBS sepsis in neonates and in practitioner opinions as to the appropriate approach to screening for and treatment of GBS [10]. Such variation in views amongst obstetricians and neonatologists reflects uncertainty about the application of differing hospital guidelines.
The current strategy at The Women's and Children's Hospital (WCH) in Adelaide for the prevention of GBS infection in the newborn includes the administration of prophylactic antibiotics during labour to women identified as being colonised with GBS, following universal screening with prenatal low vaginal cultures at 32 weeks gestation.
This study was designed to provide a scientific basis for optimum timing and method of GBS screening as specified in guidelines for antenatal care, to determine whether screening for GBS infection at 35–37 weeks gestation has better predictive values for colonisation at birth than screening at 31–33 weeks, to examine the test characteristics of a risk factor strategy and to determine the test characteristics of low vaginal swabs alone compared with a combination of perianal plus low vaginal swabs per colonisation during labour.
Methods
Study population
Women were eligible for inclusion if they had a singleton pregnancy, attended the Women's and Children's Hospital for their antenatal care over a 13-month period from May 1998 to May, 1999 and expected to deliver at that hospital at term. Women with previous GBS disease were included as were women enrolled in a shared care program between general practitioners and the hospital. Ethics committee approval was obtained from the Women's and Children's Hospital.
Recruitment
Information sessions were held for antenatal clinic and labour ward staff prior to the commencement of recruitment and during the recruitment period, to familiarise staff with the study and incorporate their suggestions into the study protocol if appropriate. Women were informed about the study after their 18-week morphological scan and approached for consent at approximately 28 weeks gestation. Women who consented received vaginal and perianal swabs at 31–33 weeks, 35–38 weeks and during labour. A sample of these women participated in focus groups to explore their views about the collection of swabs antenatally and intra-partum, their attitudes to the prophylactic administration of antibiotics during labour and their understanding about GBS infection [11].
Although we had intended at the outset to recruit private patients, we could not develop cost-effective strategies for their involvement.
Patient management
Participants had a low vaginal swab taken at 31–33 weeks for detection of GBS, the current protocol at the WCH. Standard recommendations for taking swabs were given to clinic staff to ensure that low vaginal swabs were taken without a speculum, by inserting the swab 2–3 cm. into the vagina and rotating the swab with a circular motion, leaving it in the vagina for approximately 5 seconds. A separate perianal swab was taken by gently rotating the swab around the anal margin for approximately five seconds. The swabs were placed in Stuart transport medium for transport to the laboratory within two hours.
At 35–37 weeks additional swabs (vaginal and perianal) were taken in the antenatal clinic by study staff following the protocol described above. The results of both screening swabs were made available to the caregivers. Women found on antenatal screening to be GBS carriers were recommended to have intrapartum antibiotics as per the hospital clinical guidelines.
When women were admitted in labour, a further vaginal and perianal swab was taken for culture by the admitting midwife to determine intrapartum colonisation. Note was made of whether the membranes had ruptured prior to the swab. Medical records for study participants contained an eye-catching sticker to remind labour ward staff that both a perianal and a vaginal swab needed to be taken. In addition, participants were given a card identifying them as participants in the "SWABS" study, to be handed to labour ward staff as a prompt for the taking of swabs. Study midwives provided inservice training on the protocol for taking labour swabs to staff in emergency, labour ward and the birthing centre.
Microbiology
Upon receipt in the microbiology laboratory, the swabs were cultured on layered horse blood agar and inoculated into a selective broth (Todd Hewitt broth containing gentamicin 4 mg/L and nalidixic acid 15 mg/L). The agar plate was incubated at 35°C in a carbon dioxide-enriched environment for 18–24 hours, and the broth was incubated at 35°C overnight, subcultured onto layered horse blood agar and incubated as above. Plates were inspected for β-hemolytic colonies and Streptococci were identified according to standard laboratory procedures. Presumptive GBS colonies were confirmed using the Phadebact latex agglutination method. Growth on the plate was semiquantified as described by [12]. Growth from broth only was described as "scanty".
Data collection and management
Data were collected to define the characteristics of the population including age, insurance status, socioeconomic status, parity, weight at booking, smoking status at booking, previous known GBS infection, asymptomatic bacteriuria, GBS screening at booking, previous preterm birth or preterm prelabour rupture of membranes. Women were classified as being at high risk if they had any of the following risk factors; GBS bacteruria at booking, birth at <37 weeks' gestation, prelabour rupture of membranes or temperature during labour of greater than or equal to 38 degrees Centigrade. Data were keyed into the study computer database (Microsoft ACCESS) with range checking, logic checking and verification of key fields.
Analyses
Sensitivities, specificities, positive predictive values and negative predictive values for colonisation at birth, are reported with exact confidence intervals for antenatal screening at 31–33 (referred to as 32 weeks) and 35–37 weeks gestation (referred to as 36 weeks) for vaginal cultures alone (LVS), perianal cultures alone (PAS) and combined low vaginal and perianal cultures (either). Likelihood ratios were calculated to express the odds that a given level of a diagnostic test result would be expected in a woman colonised at term. Diagnostic odds ratios and their exact 95% confidence interval are an indication of the strength of the association between having a positive likelihood ratio and being diseased.
These test characteristics were compared to identify the best method of screening as determined by site of swab, timing and the interaction between site and timing, using a multinominal logit model with the weighted least squares method of estimation. The interaction term was not statistically significant and thus the results of the main effects models alone are reported [13].
Different screening strategies were examined to determine the relative value of screening at 8 weeks prior to expected birth (with the potential for lower predictive values but with the potential also for the identification of women colonised with GBS who may give birth between 32 and 35–37 weeks and screening closer to term or using risk factors to identify likely to be colonised at birth.
Sample Size
The sample size was calculated assuming a weighted least squares analysis [14]. Calculations were based on assumptions concerning estimates of prevalence of GBS infection at the WCH at the time of routine screening (13%) and at birth (10%) (McDonald, personal communication). We assumed that the positivity rate at a 36-week screen would be 13%. Research published prior to the design of this study reported that a late antenatal screen had a sensitivity of approximately 87% [15]. Using a more conservative estimate of 80.5 percent, a sample size of approximately 839 was adequate to detect a sensitivity of 87% or greater. Alpha was set at 0.05 and β was selected to be 0.2. The correlation between estimated sensitivities at the two time periods was estimated to be approximately 0.7.
Results
A total of 865 women, of the 1168 approached, consented to participate in the SWABS study giving a participation rate of 74% (Figure 1). A number of these women (35) withdrew from the study over the period of follow-up, reflecting mobility of the patient population (9), reluctance of the woman or her partner to continue (21), a reaction to a positive test (3) and unknown reasons (2). Swabs were obtained from 93% of participants at 32 weeks, the time of the routine hospital antenatal screen for GBS. At 36 weeks swabs were taken from 94% of women who had neither given birth nor withdrawn while labour swabs were taken for 84% of women who had not withdrawn.
Figure 1 Recruitment and data collection. LVS Low vaginal swab PAS Perianal swab
Study participants reflected the composition of the low-risk antental clinic at the Women's and Children's Hospital (Table 1) in their age, gravidity, parity and model of care. Thirty women (3.5%) were positive for GBS with bacteriuria at booking, although these results were not known for over 13% of participants. Of the 48% of women with a previous pregnancy who reached 20 weeks or more, 55 (13%) had a history of GBS carriage in a previous pregnancy. Only one woman reported a history of having a child with neonatal GBS sepsis.
Table 1 Characteristics of the 865 participants (Means and standard deviations or numbers and (percentages)).
Characteristic Mean (sd) or Number (%)
Age Mean (sd) 28.0 (5.5)
Age group
≤ 20 74 (8.6)
21–34 680 (78.6)
≥ 35 111 (12.8)
Parity
0 447 (51.7)
1–3 402 (46.5)
≥ 4 16 (1.8)
Model of care (booking)
Traditional + midwifery antenatal care 461 (53.3)
Birth Centre 185 (21.4)
GP Shared Care 219 (25.3)
Smoking at booking 185 (21.4%)
GBS bacteriuria at booking
Positive (4% of all with known values) 30 (3.5%)
Unknown 117 (13.5%)
History of GBS among women with parity 1+ (n = 418)
Positive in previous pregnancy (27.3% of known) 55 (13.2)
Unknown 184 (44%)
History of neonatal GBS sepsis n = 418 1 (0.2)
Unknown 26 (36.2)
History of
Preterm birth 32 (7.7)
Preterm prelabour rupture of membranes 12 (2.9)
Sociodemographic characteristics
Race
Caucasian 818 (94.6)
Aboriginal/Torres Strait Islander 8 (0.9)
Asian 35 (4.0)
Other 4 (0.5)
Education
School student 5 (0.6)
Left school aged < 16 55 (6.4)
Left school aged ≥ 16 369 (42.7)
Trade qualification 36 (4.2)
Certificate or Diploma 241 (27.9)
Bachelors Degree or higher 159 (18.4)
Colonisation rates were constant across the gestational ages examined, with approximately 20% of all participants colonised regardless of the gestational age of screening and the swab site. At all times, the colonisation rate was slightly higher with the perianal than the vaginal swab. (Table 2). Colonisation rates would have been reported as being between 14–17% if results for selective broth were not included, approximately 5% lower than those actually identified using the broth.
Table 2 Colonisation rates (%) and 95% confidence intervals by gestational age and swab sitea. Using selective broth
a. Using selective broth
Swab site 32 weeks gestation 36 weeks gestation Labour
Prevalence 95 % CI Prevalence 95% CI Prevalence 95% CI
Low Vaginal Swab (LVS) 19 16–21 20 17–23 21 18–24
Perianal Swab (PAS) 21 18–24 22 19–25 22 19–25
Either LVS or PAS 22 19–25 24 21–26 24 21–27
b. Removing "scanty" levels of colonisation as would occur in the absence of selective broth
Swab site 32 weeks gestation 36 weeks gestation Labour
Prevalence 95 % CI Prevalence 95% CI Prevalence 95% CI
Low Vaginal Swab (LVS) 14 11–16 15 13–18 17 14–20
Perianal Swab (PAS) 15 12–18 15 12–17 16 13–19
Either LVS or PAS 17 15–20 18 16–21 20 17–23
Using the risk factor algorithm a total of 146 women (18% of the 830 women who did not withdraw, 95% CI 15% – 20%) would have been identified as being eligible for intrapartum antibiotic chemoprophylaxis.
Analysis of the test characteristics associated with different screening schedules was undertaken for the subset of 600 (69%) study participants for whom complete data were available from testing at 32 weeks and 36 weeks gestation as well as in labour (Table 3). Analysis of the use of a risk factor protocol was undertaken for the 699 participants for whom a labour swab was taken (Table 3). It was apparent that tests at 36 weeks were more sensitive and had higher negative predictive values and lower Likelihood Ratio negative values than tests at 32 weeks. There were no statistically significant differences in specificity, positive predictive value or Likelihood Ratio positive values associated with the timing of screening. Having a positive perianal or low vaginal swab – the more inclusive definition, was a more sensitive test than low vaginal swab alone but with a trade-off in terms of specificity and positive predictive value. There was no statistical difference in the Likelihood Ratio positive, but Likelihood Ratio negative was lower using the more inclusive definition of positivity (Table 4). Classifying women as being at high risk was not sensitive (Table 3) and was not informative in predicting GBS status during labour (Likelihood ratio tests were not different than 1).
Table 3 Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive and negative likelihood ratios with 95% confidence intervals for screening at 32 weeks or 36 weeks, using low vaginal (LVS) and/or perianal (PAS) swabs for 600 women and for using a risk factor strategy*.
32 Weeks (n = 600) 36 Weeks (n = 600) (n= 699)
LVS PAS Either LVS PAS Either
Estimate + 95% CI Estimate + 95% CI Estimate + 95% CI Estimate + 95% CI Estimate + 95% CI Estimate + 95% CI Estimate + 95% CI
Sensitivity 63 54–71 70 61–77 72 63–79 73 65–81 76 67–83 81 73–87 19 13–26
Specificity 94 92–96 94 91–96 93 90–95 95 93–97 94 91–96 93 90–95 83 79–86
PPV* 77 68–84 76 68–84 75 66–82 82 74–88 78 70–85 77 69–84 25 18–34
NPV 90 87–92 91 89–94 92 89–94 92 90–95 93 90–95 94 92–96 77 73–80
Likelihood ratio (+ve) 11.3 7.6–16.7 11.2 7.7–16.2 10.1 7.2–14.3 15.5 10.2-23.6 12.1 8.4-17.5 11.7 8.3-16.6 1.10 0.76-1.59
Likelihood ratio (-ve) 0.39 0.31–0.49 0.32 0.25–0.42 0.30 0.23–0.40 0.28 0.21–0.37 0.26 0.19–0.35 0.21 0.15–0.29 0.98 0.90–1.06
Diagnostic Odds Ratio 28.7 16.4–50.6 34.5 19.8–60.4 33.4 19.4–57.9 55.4 30.1–103 46.5 26.1–83.1 56.7 31.4–103 1.12 0.69–1.79
* Risk factor strategy included women with GBS bacteruria at booking, birth at <37 weeks' gestation, prelabour rupture of membranes or temperature during labour of greater than or equal to 38 degrees Centrigrade.
Table 4 Analysis of timing and site of swabs.
Timing (36 v 32 weeks) Site (Either Lower Vaginal or Perianal vs Lower Vaginal Swab alone)
Difference (36 – 32) 95% Confidence Interval p-value Difference (Either vs LVS) 95% Confidence Interval p-value
Sensitivity 8.8% (1.2%, 16.3%) 0.023* 8.2% (4.6%, 11.7%) <0.001†
Specificity 0.6% (-1.2%, 2.5%) 0.493 -1.8% (-2.6%, -0.9%) <0.001§/P >
Negative Predictive Value 2.3% (0.3%, 4.3%) 0.026* 2.0% (1.0%, 3.0%) <0.001†
Positive Predictive Value 3.6% (-2.8%, 9.9%) 0.269 -3.1% (-5.8%, -0.5%) 0.019§/P >
Likelihood Ratio + 1.62 (-2.69, 5.92) 0.462 -1.59 (-3.68, 0.49) 0.134
Likelihood Ratio - -0.09 (-0.18, -0.01) 0.021* -0.08 (-0.12, -0.04) <0.001†
Diagnostic Odds Ratio 14.3 (-8.06, 36.8) 0.210 3.63 (-4.99, 12.2) 0.409
* favours swabs taken at 36 weeks over 32 weeks
† favours either site over LVS only
§favours LVS only over either site
Multivariate analysis examined the impact of timing (screening at 36 rather than 32 weeks) and method of screening (low vaginal swabs or both low vaginal and perianal swabs) (Table 4). The interaction between method and timing was not significant. It was clear that there is an unequivocal benefit associated with delaying screening until 36 weeks. The results for method of screening were less definitive. Although sensitivity, negative predictive value and likelihood ratio negative were improved using a combined low vaginal and perianal swabbing regimen, the LVS swab alone was associated with higher specificity and positive predictive value.
Discussion
This study has demonstrated that colonisation during pregnancy with Group B Streptococcus is common amongst a low risk antenatal population. Regardless of the timing of the testing, approximately 20% of women were identified with a positive swab and therefore would have been eligible, using the hospital protocol, for antibiotic use during labour. Although this colonisation rate is a little lower than that reported by Yancy et al. in their investigation of timing of swabs in 826 women (26.5% vs approx 20% in this study) there were slight differences in the study population with the former study excluding women who had received antibiotics within a week prior to birth. The test characteristics reported from the Yancy study (Sensitivity 87%, Specificity 96%, PPV 87%, NPV 96%) were all stronger than in this current study (36 weeks screen: Sensitivity 81%, Specificity 93%, PPV 77%, NPV 94%). Likelihood ratios and diagnostic odds ratios were not reported for the former study.
This study provides clear evidence about screening timing and strategy in order to identify women colonised with GBS in labour, with more equivocal evidence about methods. Screening for GBS infection at 35–37 weeks gestation has better test characteristics and predictive values for colonisation at birth than screening at 31–33 weeks. As the hospital in which this research was undertaken has a policy of routine administration of antibiotics to women at risk of preterm birth, a delay in the timing of screening would not exclude those women at higher risk for GBS infection. In an environment in which this was not policy however, screening at 35–37 weeks may miss a particularly high-risk group.
We have defined 'high risk' as GBS bacteruria at booking, preterm birth <37 weeks' gestation, prelabour rupture of membranes or pyrexia in labour (temperature greater than or equal to 38 degrees Centrigrade). The test characteristics of a screening strategy using these risk factors are relatively poor although the use of a non-independent reference standard (colonisation in labour identified following LVS and PVS) is an issue. This finding reinforces those from a multisite study sponsored by the Centers for Disease Control [16] whose guidelines state that a risk-based strategy is not supported.
Low vaginal swabs, perinanal, or combined?
The use of both low vaginal swabs and perianal swabs identifies a higher proportion of colonised women. Whether in fact this higher antepartum detection rate will contribute to lower rates of neonatal infection and morbidity has yet to be determined. The additional costs of such an approach would need to be examined to determine whether hospital guidelines should be altered.
There is variability in screening practices in clinical practice nationally and worldwide. The companion paper from this study reporting the results of qualitative interviews with participants highlights that pregnant women are keen to do everything possible to ensure that they have a healthy liveborn infant and, that swabbing is not seen as particularly intrusive. Although these women expressed little concern about the potential adverse effects of antibiotic use [11] such concern is an appropriate one for healthcare workers.
Conclusion
This study has many strengths. A very high proportion of eligible patients agreed to participate and follow-up rates were high. The approach used to taking swabs and to their subsequent analysis is in accord with current best practice. A more detailed approach to the statistical analysis of the data is presented than has been reported in prior research. The study had the ability to examine the potential impact of a high risk strategy and a screening strategy among the same women. What this study cannot do is provide much needed direct evidence about the relative effectiveness of different strategies in terms of prevention of Early Onset GBS. Such evidence would require the conduct of very large randomised controlled trials which to date have not been seen to be feasible. For this reason, evidence from studies such as this is essential for institutions developing screening policies.
List of abbreviations used
CI Confidence Interval
EOGBS Early Onset Group B Streptococcal Sepsis
GBS Group B Streptococcus
LVS Low vaginal swabs
NPV Negative predictive value
PAS Perianal swabs
PPV Positive predictive value
WCH Women's and Children's Hospital
Declaration of competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
JH contributed to conception and design, coordination of data collection, analysis and interpretation of data, drafting and revising the article and approval for publication.
HMcD contributed to the design, data collection, analysis and interpretation of microbiological data and participated in drafting and revising the article
PD contributed to the study design, interpretation of results and revision of the article
CC contributed to the conception and design, data collection, interpretation of results and drafting and revision of the article.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We are grateful to Rebecca Smith, Anne Back, Carmel Collins and Julian Flint for their role in recruitment, swab taking and data collection; Elizabeth Griffith, Kristyn Willson, Heather McElroy and Sarah Russell for data management and analysis; and Sally Barr, Ena Ribic, and Amanda Liebelt for microbiological analysis. In addition, the study would not have been possible without the participation of staff from Antenatal Clinics, Labour Ward and Emergency and study participants.
This study was supported by the National Health and Medical Research Council Project Grant Number 980164.
==== Refs
Schuchat A Epidemiology of Group B Streptococcal Disease in the United States: Shifting Paradigms Clin Microbiol Rev 1998 11 497 513 9665980
Isaacs D Royle JA Intrapartum antibiotics and early onset neonatal sepsis caused by group B streptococcus and by other organisms in Australia. Australasian Study Group for Neonatal Infections Pediatr Infect Dis J 1999 18 524 528 10391182 10.1097/00006454-199906000-00009
Heath PT Balfour G Weisner AM Efstratiou A Lamagni TL Tighe H O'Connell LAF Cafferkey M Verlander NO Nicoll A McCartney AC on behalf of the PHLS GBS Working Group Group B streptococcal disease in UK and Irish infants younger than 90 days Lancet 2004 363 292 94 14751704 10.1016/S0140-6736(03)15389-5
Boyer KM Gadzala CA Burd LI Fisher DE Paton JB Gotoff SP Selective intrapartum chemoprophylaxis of neonatal group B streptococcal early-onset disease. I. Epidemiologic rationale J Infect Dis 1983 148 795 801 6355316
McDonald HM O'Loughlin JA Jolley P Vigneswaran R McDonald PJ Prenatal microbiological risk factors associated with preterm birth Br J Obstet Gynaecol 1992 99 190 196 1606115
Gilbert GL Hewitt MC Turner CM Leeder SR Epidemiology and predictive values of risk factors for neonatal group B streptococcal sepsis Aust N Z J Obstet Gynaecol 2002 42 497 503 12495094 10.1111/j.0004-8666.2002.00497.x
Benitz WE Gould JB Druzin ML Risk factors for early-onset Group B Streptococcal sepsis: estimation of odds ratios by critical literature review Pediatrics 1999 103 e77 10353974 10.1542/peds.103.6.e77
Centers for Disease Control and Prevention Prevention of perinatal group B streptococcal disease MMWR 2002 51 1 24
Royal College of Obstetricians and Gynaecologists Prevention of Early Onset Neonatal Group B Streptococcal Disease RCOG Guideline 2003 36
McLaughlin K Crowther C Universal antenatal group B streptococcus screening? The opinions of obstetricians and neonatologists within Australia Aust NZ J Obstet Gynaecol 2000 40 338 340
Darbyshire P Collins C McDonald HM Hiller JE Taking antenatal GBS seriously: women's experiences of screening and perceptions of risk Birth 2003 30 116 123 12752169 10.1046/j.1523-536X.2003.00230.x
Rotheram EB Schick NF Nonclostidial anaerobic bacteria in septic abortion Am J Med 1969 46 80 89 4951428 10.1016/0002-9343(69)90060-6
Grizzle JE Sturmer CF Koch GG Analysis of categorical data by linear models Biometrics 1969 25 489 503 5824401
Rochon J The application of the GSK method to the determination of minimum sample sizes Biometrics 1989 45 193 205 2655729
Yancey MK Schuchat A Brown LK Ventura VL Markenson GR The accuracy of late antenatal screening cultures in predicting genital Group B Streptococcal colonization at delivery Obstet Gynecol 1996 88 811 815 8885919 10.1016/0029-7844(96)00320-1
Schrag SJ Zell ER Lynfield R Roome A Arnold KE Craig AS Harrison LH Reingold A Stefonek K Smith G Gamble M Schuchat A A population-based comparison of strategies to prevent early-onset group B streptococcal disease in neonates N Engl J Med 2002 347 233 239 12140298 10.1056/NEJMoa020205
|
16042773
|
PMC1190189
|
CC BY
|
2021-01-04 16:32:05
|
no
|
BMC Pregnancy Childbirth. 2005 Jul 22; 5:12
|
utf-8
|
BMC Pregnancy Childbirth
| 2,005 |
10.1186/1471-2393-5-12
|
oa_comm
|
==== Front
BMC PediatrBMC Pediatrics1471-2431BioMed Central London 1471-2431-5-211599846410.1186/1471-2431-5-21Research ArticleDeterminants of response to a parent questionnaire about development and behaviour in 3 year olds: European multicentre study of congenital toxoplasmosis Salt A [email protected] K [email protected] A [email protected] N [email protected] W [email protected] G [email protected] D [email protected] HK [email protected] RE [email protected] European Multicentre Study on Congenital Toxoplasmosis (EMSCOT) [email protected] The Neurodisability Service, Great Ormond Street Hospital for Children and Institute of Child Health, London, UK2 Albert Einstein College of Medicine, Department of Epidemiology and Population Health, New York, U.S.A3 Department of Pediatrics, Division of Neonatology and Intensive Care, Medical University of Vienna, Austria4 CHU de NICE, Service Parasitologie – Mycologie, Hopital L'Archet II, BP 3079, 06202 NICE Cedex 35 Perinatal Infection Unit, Dept of Pediatrics, University of Naples Federico II, Naples, Italy6 Karolinska University Hospital, Huddinge, Stockholm, Sweden7 Department of Parasitology, Staten Seruminstitut, Copenhagen, Denmark8 Centre for Paediatric Epidemiology and Biostatistics, Institute of Child Health, London, UK2005 5 7 2005 5 21 21 27 10 2004 5 7 2005 Copyright © 2005 Salt et al; licensee BioMed Central Ltd.2005Salt et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
We aimed to determine how response to a parent-completed postal questionnaire measuring development, behaviour, impairment, and parental concerns and anxiety, varies in different European centres.
Methods
Prospective cohort study of 3 year old children, with and without congenital toxoplasmosis, who were identified by prenatal or neonatal screening for toxoplasmosis in 11 centres in 7 countries. Parents were mailed a questionnaire that comprised all or part of existing validated tools. We determined the effect of characteristics of the centre and child on response, age at questionnaire completion, and response to child drawing tasks.
Results
The questionnaire took 21 minutes to complete on average. 67% (714/1058) of parents responded. Few parents (60/1058) refused to participate. The strongest determinants of response were the score for organisational attributes of the study centre (such as direct involvement in follow up and access to an address register), and infection with congenital toxoplasmosis. Age at completion was associated with study centre, presence of neurological abnormalities in early infancy, and duration of prenatal treatment. Completion rates for individual questions exceeded 92% except for child completed drawings of a man (70%), which were completed more by girls, older children, and in certain centres.
Conclusion
Differences in response across European centres were predominantly related to the organisation of follow up and access to correct addresses. The questionnaire was acceptable in all six countries and offers a low cost tool for assessing development, behaviour, and parental concerns and anxiety, in multinational studies.
==== Body
Background
Measurement of children's development, behaviour, and impairment is essential in studies that seek to determine the impact of early life events on functional abilities. Because professional administered standardised assessments are extremely resource intensive, parent-completed questionnaires are used increasingly, particularly in large studies of populations at low risk of impairment [1-4]. Uncertainties about the validity of parent-reported outcomes have been addressed by several studies showing that, compared with professional assessments, parents correctly report moderate to severe cognitive or speech and language impairment, behavioral problems, and disability [3,5-12]. Much less is known about the reliability and acceptability of parent-completed questionnaires in different countries, languages and cultures, except for tools measuring behaviour or quality of life [13-17]. Such information is particularly relevant for multinational studies.
This report is based on a prospective multicenter cohort study, The European Multicentre Study on Congenital Toxoplasmosis (EMSCOT), which was initiated to determine the effects of congenital toxoplasmosis and prenatal treatment on development, behaviour, specific impairments, and parental anxiety. We were constrained by the need for the assessment tool to be low cost, require minimal input by local investigators, cover all domains of development, and avoid measurement of vocabulary or other language-specific attributes. In addition, we wanted a tool that maximised response, minimised bias among responders, measured the same entity, and was similarly acceptable, in all six countries studied. The aim of the tool was to detect moderate to severe abnormality in the outcomes measured.
In order to assess the potential for bias when using the postal questionnaire, we examined the influence of organizational factors within centers and characteristics of the individual child, on three outcomes: response to the questionnaire, the age at response, and, among responders, completion of the drawing tasks by children. The aim was to determine the applicability and acceptability of this low cost tool in different European settings.
Methods
Study population
The study population comprised children with and without congenital toxoplasmosis born to women exposed to toxoplasmosis in pregnancy (Figure 1). In eight centres (six in France, one in Vienna, and one in Naples), women were identified by prenatal screening. In Stockholm, infected women were identified by retrospective testing of stored prenatal samples, and in two centres (Copenhagen and Poznan), children were identified by neonatal screening for congenital toxoplasmosis. We enrolled all children in the French centres, where the ratio of uninfected to infected was about three to one. However, in Vienna, Stockholm and Naples there were about nine uninfected children to each infected child. We therefore randomly selected four uninfected children for each infected child for inclusion in the survey at three years of age. In Poznan and Copenhagen, only infected children were identified by neonatal screening. Therefore in Poznan we enrolled the next six uninfected children born after each infected child who underwent Guthrie Card screening. No uninfected children were enrolled in Copenhagen. More than 90% of women received prenatal anti-toxoplasma treatment in France, Vienna, and Naples. In the other centers, none of the women were treated. The screening schedule and duration of postnatal treatment of infected children are summarized in Table 1 and reported in more detail elsewhere [18].
Figure 1 Flow diagram to show recruitment into the study. Flow diagram to show recruitment into the study. Criteria for possible referrals reported elsewhere [18]. CT = congenital infection status: - depicts uninfected, and + infected children.
Table 1 Summary of clinical management and follow up protocols, and response rates for each center
Organizational attributes
Study Center Total sent questionnaires Prenatal re-testing interval (months)1 Duration post-natal treatment (month) A B1 B2. C Total
FRANCE
Lyon1 184 1 14 3 0 3 3 9
Paris1 182 1 12 1 1 2 2 6
Grenoble1 34 1 12–24 1 0 1 1 3
Marseille1 91 1 12–24 3 1 3 3 10
Nice1 44 1 24 4 0 3 3 10
Toulouse1 73 1 12 4 0 2 2 8
AUSTRIA
Vienna1 187 3 12 4 1 3 2 10
ITALY
Naples1 53 3 12 4 1 3 3 11
SWEDEN
Stockholm1 16 NS4 12 4 1 2 1 8
POLAND
Poznan2 180 NS 12 4 1 3 2 10
DENMARK
Copenhagen3 14 NS 3 1 1 1 2 5
TOTAL
NS = neonatal screening
Uninfected children: 1 born to infected women; 2 sampled from general population, 3 none 4 Detection of maternal infection based on neonatal testing of neonatal Guthrie card bloodspots and retrospective testing of stored maternal serum
Organizational attributes
A. Degree of local study clinician direct involvement in follow-up (FU) of children and contact with child's local paediatrician
4 = FU >= 75% children and regular contact
3 = FU >= 50% children and regular contact
2 = FU < 50% children and some contact
1 = FU <25% and no regular contact
B. Access to addresses
B1: 1 = National or local address register
0 = No use of population address register
B2: 3 = regular contact with parents/paediatricians to update addresses >50%
2 = initial contact address >50% but additional methods to update addresses
1 = >75% through initial contact address only
C. Direct contact with parents to encourage return of questionnaire
3. Telephone contact and special letter
2. Follow-up letter
1. No special efforts
Postnatal follow up
All children born to toxoplasma infected women had paediatric, ophthalmic and cranial ultrasound examinations in early infancy, and infected children were assessed annually [18]. The exception to this rule was the group of uninfected children in Poznan who were not offered specialist clinical follow up. At 36 months of age, a questionnaire was mailed to parents together with a stamped addressed reply envelope, an information sheet, and crayons for the child. Two reminders were mailed to non-responders at 2 monthly intervals.
After the study, we sent a questionnaire to each centre to measure organisational attributes, such as whether local study investigators were directly involved in provided clinical follow up for the child, whether they had regular contact with the child's own paediatrician, access to a central address register for tracing families, and contact with families to encourage response. These factors were summed to generate a total unweighted score (see Table 1).
Outcomes
The questionnaire consisted of 30 questions measuring motor, speech and language, and cognitive development, behavior, parental concerns about development, parental anxiety about the health of their child now and in the future, referral to a specialist, and specific impairments (including vision, hearing, cerebral palsy, and epilepsy). The full version is available at: . The assessment tools from which the questions were derived are summarized in Table 3. For behavior, we used the entire assessment tool, as validated in a large community sample of children [12]. However, for speech and language, and cognition, we scrutinized the correlation coefficients in unpublished data provided by the developers of these tools and, with their permission, selected those items which were most predictive and independent. We considered that the population of children born to toxoplasma infected women would be similar to the general community populations in which these assessment tools had been validated. Although components of the questionnaire have been validated, the entire questionnaire in the format used in this study has not yet been validated.
Table 3 Source for questions measuring development, behavior, and parental concerns and anxiety
Outcome Question number in questionnaire* Score range Source
Development and behavior
Motor development 6 (a–h) 0–16 Griffiths Mental Development Scales [23], Denver Developmental screening test [24]
Speech Development
Language development 10 (a–c)
10 (d–g) 0–6
0–8 General Language Screen**, parent completed questionnaire for 3 year olds [25]. Validated against four standardised speech and language tests administered by an assessor.
Cognitive ability (non-verbal) 11 (a–g) 0–7 PARCA3** (Parent Report of Children's Abilities) [6] validated for 3 year olds against the MacCarthy Scale
Behavior 13 (a–y) 1–16 'Strengths and difficulties questionnaire' (SDQ), validated in 3 to 16 year olds against clinician assessment of behavior disorder [10,12]. Entire questionnaire, published translations, and scoring algorithm used.
Parental concerns, specialist referral, and parental anxiety
Parental concerns
a) Learning, behavior, development
b) Speech and language 5
8 1–3
1–3 Adapted from PEDS** (Parent Evaluation of Developmental Status). Predicts risk for developmental and behavioral problems and the need for clinical assessment.[26,27]
Impact of behavior on family 14, 14a–d 0–2 SDQ questionnaire [10,12,28]
Parental anxiety 25,26,27 0–15 Adaptation of rating scales for measuring anxiety during pregnancy and postpartum [29] in relation to antenatal screening (numbered six point horizontal scale with verbal anchors at extremes).
Child completed questions
Cognitive and fine motor skills
Copying a line, circle, and cross.
Draw a man 12 (a–c)
30 0–3
1–24 The child's ability to copy a circle, line and cross was assessed using scoring and normative data available from the Beery Buktenica Developmental Test of Visual Motor Integration [30]. The 'draw a man' was scored using a standardised system and normative data from the Goodenough Draw a Man test [31], using raw scores.
Confounding variables
Education level achieved 28 0–3 Educational level achieved based on standard categories defined by the Organisation for Economic Co-operation and Development (OECD) for Europe.[32]
* Full version of questionnaire available on
**A subset of the most predictive and least correlated questions were selected.
Questionnaire development and piloting
Development of the questionnaire involved collaborating pediatricians, obstetricians, parasitologists and psychologists in different countries to ensure that questions would be widely understood and acceptable. Questionnaires were translated into the six languages in the study, back translated to English by someone unaware of the original English version, and compared to the original version to ensure meaning was retained. The questionnaire was piloted in the six countries in general pediatric outpatient clinics, high risk (preterm) follow-up clinics, and day care centers, and parents were asked about difficult or offensive questions, the length of the questionnaire, and how long it took to complete. Research ethics approval was obtained for all participating centres
Analyses
Development of scores
To summarize the responses relating to development and parental anxiety, unweighted scores were derived without knowledge of infection or treatment status. Spearman correlation coefficients were used to identify redundancies among items, and the final scale was based on items with relatively low inter-item correlations. If less than 50% of answers were missing for each outcome, the total score was prorated. For behaviour, and the children's drawings, we used the published scoring systems (see Table 3). All scores were coded so that a high score was abnormal. One third of the children's drawings were scored by a second assessor and discrepancies reviewed.
Analysis of response
We developed multivariate models to identify factors associated with each of the three outcomes: a) whether the questionnaire was completed and returned; b) the child's age at questionnaire completion; and c) whether the child completed the 'draw a man' task. Age at response was analysed as a surrogate marker for 'response or not' that might be susceptible to family as well as center factors. The child-completed task was included to assess its acceptability and the potential bias involved in such assessments that require additional effort from the family.
Initially, we examined the heterogeneity of effects within French centres, and found significant differences in response across centres. Thus, we decided to use a hierarchical generalized linear model for dichotomous outcomes (response to the survey, and response to 'Draw a Man') to account for heterogeneity among centers within France, along with other centres in the model. A generalized estimating equation (SAS Version 9.1 PROC GENMOD with the ASSESS options to assess fit of the model) with centers nested within country was derived to determine characteristics associated with response to questionnaire, and completion of the 'draw a man' task. Goodness of fit was assessed using the Pearson Chi-square result divided by its degrees of freedom. Values close to 1 indicated lack of overdispersion of the model [19]. To predict child's age at survey completion, we used multiple linear regression. Lyon, the largest centre, was used as the reference category.
The models examined the effect of centre, and the score for centre organisational attributes. We also examined the effect of the patient characteristics (see Table 4), including the presence of intracranial lesions, or abnormal neurological findings (microphthalmia, microcephaly, seizures, or abnormal neurological examination requiring referral to a specialist) before 4 months of age. This cut-off was chosen as the number of examinations was similar for infected and uninfected children up until this age.
Table 4 Characteristics associated with response to questionnaire (N = 1058 total)
Characteristic Number responding (%) Odds ratio for response4 (95% confidence interval) Final model: Adjusted odds ratio5 (95% CI)
All centers 714 (67.5)
Center Variables
Lyon (reference) 134 (72.8) reference
Paris 91 (50.0) 0.37 (0.24, 0.57)
Grenoble 8 (23.5) 0.11 (0.05, 0.27)
Marseille 69 (75.8) 1.16 (0.65, 2.06)
Nice 33 (75.0) 1.12 (0.53, 2.38)
Toulouse 44 (60.3) 0.56 (0.31, 0.99)
Copenhagen 9 (64.3) 0.67 (0.21, 2.1)
Vienna 134 (71.7) 0.97 (0.61, 1.54)
Stockholm 8 (50.0) 0.37 (0.13, 1.05)
Naples 50 (94.3) 6.22 (1.85, 20.84)
Poznan 134 (74.4) 1.09 (0.68, 1.73)
Total score for organisational attributes3 9 (6,10) 1.36 (1.27, 1.45) 1.15 (1.09, 1.23)
Infection status1
CT+ 178 (80%) 2.95 (2.01, 4.31) 4.96 (3.58, 6.88)
CT- 527 (64%)
Maternal age1,2,3 (mean years, 95% CI) 582 (67%) 28.4 (27.6, 29.2) 1.03 (0.99, 1.07) 1.02 (0.99, 1.05)
Parity1,2,3 (mean, 95% CI) 503 (67%) 0.9 (0.8, 1.0) 1.05 (0.90, 1.23)
Gestational age at birth1,2,3 (mean weeks, 95% CI) 530 (68%) 39.0 (38.8, 39.1) 1.00 (0.92, 1.09)
Child's gender1,2
Male 379 (68%) Reference
Female 316 (68%) 0.94 (0.71, 1.25)
Prenatal treatment1,2
Any prescribed 532 (67%) 1.85 (0.98, 3.49)
None 62 (66%)
Neurological abnormality and/or intracranial lesions1,2
Yes 25 (83.3) 2.69 (0.56, 13.00)
No 569 (66.3)
1 Excludes 14 infected children from Denmark (9 responded), as no uninfected controls available.
2 Excludes 156 children in Poznan (134 responded), as uninfected control group selected from the general population had no information on these variables.
3 Odds ratio per unit increase in characteristic
4 Adjusted for congenital infection status and center
5 Adjusted for all factors shown. Goodness of fit statistic was close to 1 (1.0063)
As the total score for organizational attributes (given in Table 1) was a proxy for centre, we repeated all analyses, initially adjusting for centre, and then adjusting for the total unweighted score for organizational attributes. Potential covariates were added to a model with congenital infection status and center to determine the magnitude of association with outcome. Variables with associations that resulted in p-values less than .20 were included in the initial multivariate model. A monitored backwards stepwise approach was conducted, and models were assessed for convergence. The final model included only variables (or categories of variables) significant at p < .05. Bivariate associations were assessed using a Chi-square or Exact test for categorical characteristics and Wilcoxon Rank Sum tests for ordinal or non-normally distributed characteristics. The best fitting, and most parsimonious model was included in the results presented. Odds ratios or estimated means are presented along with 95% confidence intervals.
Results
Piloting of questionnaire
115 parents completed pilot questionnaires (70 healthy, 31 seen in pediatric clinics for clinical problems, 14 not specified), in France (32), Italy (40), Sweden (9), Denmark (10), and Poland (24). On average, parents took 21 minutes to complete the questionnaire (range 6 to 40 minutes, SD 7.88). Most parents (88%) thought that the questionnaire was the right length, 5% reported it was short, and 7% too long. Sixteen (14%) had difficulty understanding, or objected to one or more questions, of which the most frequent were: 'maternal age when last in full time education', questions about behavior (three parents felt the questions were not suitable for the age group), and one cognitive question about puzzles which was removed.
Survey of organizational attributes and reasons for non-response
Table 1 shows that Naples, followed by Nice, Marseille, Vienna, Poznan, and Lyon had the highest total scores for the level of direct involvement in follow up by the local study centre, access to addresses, and methods to encourage compliance with the postal questionnaire. These centres also had the highest response rates (see Table 2). Overall 67% (714/1058) of parents responded, but the rate ranged from 24% (8/34) in Grenoble to 94% (50/53) in Naples. Table 2 shows that the main reason given by the local study coordinator for non-response was lack of a correct address (accounting for 44%, 151/344 of non-responders). Few parents refused to participate (n = 60, 17% of non-responders). As no reason was given for most non-responders (37%; 129/344), we could not tell whether these parents had not received a questionnaire, had refused to participate, or had completed but failed to return their questionnaire.
Table 2 Summary of response rates for each center
Reasons for non-response (% non-responders)
Study Center Response rate (%) Total non-responders Address not known Refused to participate No response Other1
FRANCE
Lyon 73% 50 12 (24) 1 (2) 37 (74) 0
Paris 50% 91 46 (50) 0 45 (49) 0
Grenoble 24% 26 1 (4) 0 25 (96) 0
Marseille 76% 22 17 (77) 0 4 (18) 1 (4)
Nice 75% 11 8 (73) 3 (6) 0 0
Toulouse 60% 29 21 (72) 0 8 (28) 0
AUSTRIA
Vienna 72% 53 21 (40) 32 (60) 0 0
ITALY
Naples 94% 3 1(33) 2 (66) 0 0
SWEDEN
Stockholm 50% 8 3 (37) 1 (12) 2 (25) 2 (25)*
POLAND
Poznan 74% 46 20 (43) 21 (46) 4 (9) 1 (2)
DENMARK
Copenhagen 64% 5 1 (20) 0 4 (80) 0
TOTAL 344 151 (44) 60 (17) 129 (37) 4 (1)
1 Other reasons eg. died, mentally retarded parents
Analysis of determinants of response to questionnaire
As shown in Table 4, there were statistically significant differences between centres in the proportion of parents responding to the questionnaire. Response was more common in Naples, and less common in Paris, Grenoble and Toulouse, than in Lyon. A more parsimonious model involved replacement of the centre variable with the centre score for organisational attributes which was significantly associated with increased response. The only other significant factor in this model was congenital infection status. (see Table 4).
Determinants of age at response
On average the questionnaire was completed at 39.7 months of age (95% CI: 39.5, 40.0; range 35.4 to 63.9 months). In the multivariable analysis, factors significantly associated with older age at completion were study centre (delayed in Paris, Vienna and Naples), duration of prenatal treatment and detection of a neurological abnormality and/ or intracranial lesions in the first 4 months of life (mean difference in months at response was 1.67; with standard error = 0.72; R2 = 0.16.).
Determinants of child's response to 'draw a man'
94% of children copied drawings of a line, circle, and cross, but only 70% of children responded to the request to 'draw a man.' In multivariable analyses, completion of 'draw a man' was more common in girls (OR 0.62; 95% CI 0.44, 0.88), age at completion of the questionnaire (Odds ratio 1.13 per additional month of age; 95% CI 1.06–1.21), and centre (children in Poznan were more likely to respond than in Lyon; odds ratio 2.53, 95% CI 1.36, 4.68). The goodness of fit statistic was 1.0092.
Parent completed questions
Most questions (>99%) were completed. Questions with the lowest rate of completion were on hearing loss (94%), vision (92%), and age when mother was last in full time education (92%).
Discussion
The score for organizational attributes varied between study centers and was one of the main determinants of response to the questionnaire. Centers where study clinicians were directly involved in patient follow up, had access to a central address register, and directly contacted parents to encourage return of the questionnaire, had the highest response rates. There was no evidence that organizational attributes were associated with age at response, nor with whether the child drew a man. Congenital infection status was strongly associated with response to the survey, but only weakly associated with age at response, and was not significantly associated with whether the child completed the 'draw a man' task.
The response rate to this parent report survey suggests that a parent-completed postal questionnaire on development, behavior, and parental concerns and anxiety is acceptable across the six European countries studied. The high response rate in this study was achieved by clinicians without dedicated research coordinators in the local centres, although there were dedicated staff centrally. In some centres clinicians were laboratory-based and not directly involved in follow up of the child. The results of this study should therefore be widely applicable.
Our findings concur with those of a systematic review of methods for increasing response rates to postal questionnaires [20]; response was higher among parents for whom the study was of most interest (those with infected children), and response was improved by follow-up contact. Other important elements of survey design highlighted in the review by Edwards et al included keeping the questionnaire short, and mailing a second copy.
Non-response can introduce bias if non-responders differ from responders with respect to prognostic characteristics. In our analyses, we found no evidence that non-responders differed in terms of maternal age, parity, or prenatal treatment or with respect to prognostic factors associated with poor developmental outcome such as gestational age at birth or abnormal clinical manifestations in early infancy. Difficulties tracing the correct address probably favored inclusion of infected children in our study, but we found no evidence for a bias in favor or against inclusion of more severely affected children.
The age at response was largely determined by the centre, and was not significantly associated with the organizational attributes. Although questionnaires were intended to be mailed as soon as they arrived at the local centre, actual practice may have varied. The centre effect may therefore be explained by unmeasured centre characteristics such as availability of staff to mail questionnaires. The weak association between duration of prenatal treatment and increased age at response may be a chance finding. However, return of questionnaires was delayed from children with an intracranial lesion or neurological abnormality. This may reflect difficulties contacting such families. A similar finding was reported in a cohort study of children born preterm; families with severely neurologically affected children were most difficult to contact (a higher proportion of families of children with severe disability repeatedly failed to attend appointments, moved frequently or were adopted or fostered) [21].
Although completion of individual questions was high, certain questions fared less well, most notably when the child was asked to 'draw a man'. Low response may be because the task is quite difficult for three year olds, the lower age limit for this test, as completion did improve with age. As child-completed tasks can provide additional objective information about development, inclusion of more age-appropriate tasks may be valuable [6,7]. Failure of the 6% to 8% of parents to answer the questions on vision and hearing may reflect parental uncertainty or lack of confidence in reporting medical information whereas functional information was well reported. The question on 'maternal age when last in full time education' received most comments during the pilot phase, and was less popular than a related question about highest level of education achieved, which has been used most widely in Europe [22].
One of the main limitations of the study is that reasons for non-response were poorly documented. Consequently, refusals to participate may have been underestimated, and acceptability may have been overestimated. A further limitation is that study centre clinicians were surveyed about the organizational attributes of their centre after the study, when they were aware of how response rates differed among centers. This may have favored overestimation of the importance of organizational attributes for response.
Conclusion
The parent completed questionnaire was acceptable in 11 centres in six European countries. Differences in response appeared to be related to organisation of follow up, and access to correct addresses. The questionnaire offers a low cost tool for assessing development, behaviour, and parental concerns and anxiety, in multinational studies.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
*Members of the European Multicentre Study on Congenital Toxoplasmosis (EMSCOT)
Coordinating Committee: M-H Bessieres, W Buffolano, H Dumon, R Gilbert, E Petersen (chairperson), A Pollak, P Thulliez, M Wallon.
Writing Committee: Freeman K, Salt A, Prusa A, Malm G, Ferret N, Buffolano W, Petersen E, Gilbert RE (coordinator).
Centres contributing data (number of patients contributed to this report): M Paul (134; University Medical Sciences, Poznan), A Prusa, M Hayde, A Pollak (134; University Children's Hospital, Vienna), M Wallon, F Peyron (134; Hôpital de la Croix Rousse, Lyon), S Romand, P Thulliez (91; Institut de Puericulture, Paris), J Franck, H Dumon, P Bastien, E Issert (69; Hôpital de la Timone, Marseille; CHU de Montpellier), W Buffolano (50; Universita di Napoli, Naples), M-H Bessieres (44; Hôpital de Rangueil, Toulouse), N Ferret, P Marty (33; Hôpital de l'Archet, Nice), H Pelloux, H Fricker-Hidalgo, C Bost-Bru (8; Centre Hospitalier Universitaire de Grenoble), G Malm, B Evengard (8; Huddinge Hospital, Stockholm), E Petersen (0; Statenseruminstitut, Copenhagen), C Chemla, (0; Hôpital Maison Blanche, Reims), E Semprini, V Savasi (0; Milan).
Study design and coordination: R Gilbert (Principal Investigator), L Gras, Hooi Kuan Tan, J Rickett, A Salt, L Valenti (Institute of Child Health, London)
Statistical analysis: Freeman K, Gras L (Institute of Child Health, London).
AS developed the questionnaires, participated in analyses and wrote the paper. KF did the statistical analyses and wrote the paper. RG had the idea for the study, designed the study, obtained the funding, developed the questionnaires, coordinated the study, and wrote the paper. All authors contributed to the design of the study and the writing of the paper. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We are grateful to Robert Plomin, Frances Glascoe, and Carol Stott forallowing us to use adapted versions or items from their parentquestionnaires, to Robert Goodman for use of his questionanaire andtranslations, and to Bruno Senta-Loys (Lyon) for contributing ideas to thedesign of the questionnaire. The research was part of the European multicentre study on congenital toxoplasmosis, funded by the European Commission (BIOMED II No. BMH4-CT98-3927 and QLG5-CT-2000-00846). Additional support has been provided by the National Eye Institute, grant code R03 EY015287-01
==== Refs
Johnson S Marlow N Wolke D Davidson L Marston L O'Hare A Peacock J Schulte J Validation of a parent report measure of cognitive development in very preterm infants Dev Med Child Neurol 2004 46 389 397 15174530 10.1017/S0012162204000635
Doig KB Macias MM Saylor CF Craver JR Ingram PE The Child Development Inventory: A developmental outcome measure for follow-up of the high-risk infant J Pediatr 1999 135 358 362 10484803
Eisert DC Spector S Shankaran S Faigenbaum D Szego E Mothers' reports of their low birth weight infants' subsequent development on the Minnesota Child Development Inventory J Pediatr Psychol 1980 5 353 364 7463264
Petrill SA Saudino K Cherny SS Emde RN Fulker DW Hewitt JK Plomin R Exploring the genetic and environmental etiology of high general cognitive ability in fourteen to thirty-six-month old twins. Child Dev 1998 69 68 74 9499557
Heiser A Curcin O Luhr C Grimmer I Metze B Obladen M Parental and professional agreement in developmental assessment of very-low-birthweight and term infants Dev Med Child Neurol 2000 42 21 24 10665971 10.1017/S0012162200000050
Oliver B Dale DS Saudino KJ Petrill SA Pike A Plomin R The validity of a parent-based assessment of cognitive abilities in three-year olds. Early Child Dev Care 2002 172 337 348 10.1080/03004430212713
Saudino K Dale P Oliver B Petrill S Richardson V Rutter M Simonoff E Stevenson J Plomin R The validity of parent-based assessment of the cognitive abilities of 2-year-olds British Journal of Developmental Psychology 1998 16 349 363
Dale PS The validity of a parent report measure of vocabulary and syntax at 24 months J Speech Hear Res 1991 34 565 571 2072681
Dinnebeil L Rule S Congruence between parents' and professionals judgements about the development of young children with disabilities: A review of the literature. Topics in Early Childhood Special Education 1994 14 1 25
Glascoe FP Sandler H Value of parents' estimates of chldren's developmental ages. J Pediatr 1995 127 831 835 7472847
Gottfried AW Guerin D Spencer JE Meyer C Validity of Minnesota Child Development Inventory in screening young children's developmental status J Pediatr Psychol 1984 9 219 230 6470904
Goodman R Ford T Simmons H Gatward R Meltzer H Using the Strengths and Difficulties Questionnaire (SDQ) to screen for child psychiatric disorders in a community sample Br J Psychiatry 2000 177 534 539 11102329 10.1192/bjp.177.6.534
Smedje H Broman JE Hetta J von Knorring AL Psychometric properties of a Swedish version of the "Strengths and Difficulties Questionnaire" Eur Child Adolesc Psychiatry 1999 8 63 70 10435454 10.1007/s007870050086
Klasen H Woerner W Wolke D Meyer R Overmeyer S Kaschnitz W Rothenberger A Goodman R Comparing the German versions of the Strengths and Difficulties Questionnaire (SDQ-Deu) and the Child Behavior Checklist Eur Child Adolesc Psychiatry 2000 9 271 276 11202102 10.1007/s007870070030
Pachter LM Dworkin PH Maternal expectations about normal child development in 4 cultural groups Arch Pediatr Adolesc Med 1997 151 1144 1150 9369877
Anderson RT Aaronson NK Wilkin D Critical review of the international assessments of health-related quality of life Qual Life Res 1993 2 369 395 8161975 10.1007/BF00422215
Perneger TV Leplege A Etter JF Cross-cultural adaptation of a psychometric instrument: two methods compared J Clin Epidemiol 1999 52 1037 1046 10526997 10.1016/S0895-4356(99)00088-8
Gilbert R Gras L Effect of timing and type of treatment on the risk of mother to child transmission of Toxoplasma gondii BJOG 2003 110 112 120 12618153
Zelterman D Zelterman D Basic log-linear models and PROC Genmod procedure Advanced log-linear models using SAS 2002 Cary, North Carolina, SAS Institute Inc 19 54
Edwards P Roberts I Clarke M DiGuiseppi C Pratap S Wentz R Kwan I Increasing response rates to postal questionnaires: systematic review BMJ 2002 324 1183 12016181 10.1136/bmj.324.7347.1183
Tin W Fritz S Wariyar U Hey E Outcome of very preterm birth: children reviewed with ease at 2 years differ from those followed up with difficulty Arch Dis Child Fetal Neonatal Ed 1998 79 F83 F87 9828731
Mackenbach JP Kunst AE Measuring the magnitude of socio-economic inequalities in health: an overview of available measures illustrated with two examples from Europe Soc Sci Med 1997 44 757 771 9080560 10.1016/S0277-9536(96)00073-1
Griffiths R The Griffiths Mental Development Scales from birth to eight years 2004 Oxford, The Test Agency Ltd.
Denver Developmental Screening Test
Stott CM Merricks MJ Bolton PF Goodyer IM Screening for speech and language disorders: the reliability, validity and accuracy of the General Language Screen Int J Lang Commun Disord 2002 37 133 151 12012612 10.1080/13682820110116785
Glascoe FP Dworkin PH The role of parents in the detection of developmental and behavioral problems Pediatrics 1995 95 829 836 7539122
Sonnander K Parental developmental assessment of 18 month old children: reliability and predictive value. Dev Med Child Neurol 1987 29 351 362 3596071
Goodman R Scott S Comparing the Strengths and Difficulties Questionaire and the Child Behaviour Checklist: is small beautiful? J Abnorm Child Psychol 1999 27 17 24 10197403 10.1023/A:1022658222914
Marteau TM Cook R Kidd J Michie S Johnston M Slack J Shaw R The psychological effects of false positive results in prenatal screening for fetal abnormality: a prospective study Prenat Diagn 1992 12 205 214 1375378
Beery KE Norman A The Beery-Buktenica Developmental Test of Visual Motor Integration 1997 New Jersey, Modern Curriculum Press
Goodenough FL Measurement of Intelligence by Drawings 1926 New York, World Book Company
Organisation for Economic Co-operation and Development (OECD). Education in OECD Countries 1987-1988 A compendium of statistical information 1990 Paris
|
15998464
|
PMC1190190
|
CC BY
|
2021-01-04 16:31:06
|
no
|
BMC Pediatr. 2005 Jul 5; 5:21
|
utf-8
|
BMC Pediatr
| 2,005 |
10.1186/1471-2431-5-21
|
oa_comm
|
==== Front
BMC Public HealthBMC Public Health1471-2458BioMed Central London 1471-2458-5-791604280110.1186/1471-2458-5-79Research ArticleSocioeconomic factors and adolescent pregnancy outcomes: distinctions between neonatal and post-neonatal deaths? Markovitz Barry P [email protected] Rebeka [email protected] Louise H [email protected] Terry L [email protected] Saint Louis University School of Public Health, St. Louis, Missouri, USA2 Saint Louis University School of Medicine, St. Louis, Missouri, USA3 Saint Louis University School of Nursing, St. Louis, Missouri, USA4 Washington University School of Medicine, St. Louis, Missouri, USA2005 25 7 2005 5 79 79 16 12 2004 25 7 2005 Copyright © 2005 Markovitz et al; licensee BioMed Central Ltd.2005Markovitz et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Young maternal age has long been associated with higher infant mortality rates, but the role of socioeconomic factors in this association has been controversial. We sought to investigate the relationships between infant mortality (distinguishing neonatal from post-neonatal deaths), socioeconomic status and maternal age in a large, retrospective cohort study.
Methods
We conducted a population-based cohort study using linked birth-death certificate data for Missouri residents during 1997–1999. Infant mortality rates for all singleton births to adolescent women (12–17 years, n = 10,131; 18–19 years, n = 18,954) were compared to those for older women (20–35 years, n = 28,899). Logistic regression was used to estimate adjusted odds ratios (OR) and 95% confidence intervals (CI) for all potential associations.
Results
The risk of infant (OR 1.95, CI 1.54–2.48), neonatal (1.69, 1.24–2.31) and post-neonatal mortality (2.47, 1.70–3.59) were significantly higher for younger adolescent (12–17 years) than older (20–34 years) mothers. After adjusting for race, marital status, age-appropriate education level, parity, smoking status, prenatal care utilization, and poverty status (indicated by participation in WIC, food stamps or Medicaid), the risk of post-neonatal mortality (1.73, 1.14–2.64) but not neonatal mortality (1.43, 0.98–2.08) remained significant for younger adolescent mothers. There were no differences in neonatal or post-neonatal mortality risks for older adolescent (18–19 years) mothers.
Conclusion
Socioeconomic factors may largely explain the increased neonatal mortality risk among younger adolescent mothers but not the increase in post-neonatal mortality risk.
==== Body
Background
Adolescent pregnancy has long been considered to increase the likelihood of adverse infant outcomes [1]. The infant mortality rate in the United States (U.S.) for young women has remained persistently high despite widespread study of this phenomenon. In 1999, the infant mortality rate for women under 20 years of age was 10.3 deaths per 1,000 live births compared to 7.0 per 1,000 live births for all ages [2] Although the U.S. teenage (ages 15–19) birth rate has declined substantially during the last decade, from 62.1 to 48.7 per 1,000 women in 2000 [3], this country still leads the developed countries by wide margins. Simple calculations would suggest that, with an estimated 470,000 births to teen mothers in the year 2000, over 1,500 excess infant deaths occurred among adolescent mothers.
The etiology explaining the higher mortality risk for adolescent mothers has been debated for as long as the phenomenon has been observed. Miller argued that socioeconomic status has a direct effect on infant mortality, confounding the role of young maternal age, in a study of 2,019 U.S. counties from 1971–1975 [4]. Many studies point towards low birth weight (including infants born preterm and those small-for-gestational age) as the predominant proximate marker of infant mortality [5-10]. Fraser, in a large study spanning 20 years of live births from Utah, found maternal age to be an independent predictor for poor birth outcomes, but this study did not specifically address infant mortality or socioeconomic status directly [11]. The defining influence of the biology of adolescent pregnancy, risky adolescent behavior, nutrition, race, socioeconomics, and adequacy of prenatal care – or the interaction among these variables – remains controversial. Certainly poverty, minority status, suboptimal prenatal care, poor education and unmarried status are more common in teenage compared to older mothers, and are known risk factors for low birth weight, preterm birth and higher infant mortality in the United States [11]. Adolescent mothers are disproportionately poor. For example, in Missouri for 1997–1999, 79% of births to women aged 14–19 were covered by Medicaid compared to only 35.7% of births to women 20 and older [12]. Among adolescents living in poor communities, 69% report having had intercourse, compared to 37% of those living in more affluent neighborhoods [13]. Support can be found to defend or refute all or none of these factors as independent risk factors [6,7,11,14-16].
We examined linked birth-death certificate data for Missouri residents to assess the role of poverty status and other potential risk factors on infant mortality among teenage mothers. We hypothesized that teen motherhood is highly confounded by poverty, and that there is no independent effect of adolescent age on infant mortality. Our analysis distinguished neonatal and post-neonatal mortality among younger and older adolescent mothers.
Methods
Study design
This was a population-based cohort study using Missouri's linked birth-death certificate files for 1997–1999. This period was chosen since it provided a sufficiently large study population with the most current birth-related data. To test the primary hypothesis, we compared infant mortality rates of adolescent and older mothers while controlling for socioeconomic status (defined by participation in poverty assistance programs) and other known and purported risk factors.
Cohort description
The study population included all singleton live births for mothers who were Missouri residents and less than 20 years of age during 1997–1999. The study population was stratified by maternal age to identify younger (12–17 years) and older (18–19 years) adolescent mothers. A comparison population included a random sample of singleton live births for mothers who were 20–34 years old during 1997–1999, since the infant mortality rates were relatively stable for this age group but higher for older mothers. Women other than non-Hispanic white or African American were excluded, because they comprised less than 3% of all live births in Missouri during the study period.
Covariates
The following covariates were extracted from the birth certificate: race/ethnicity (non-Hispanic black or non-Hispanic white), maternal education level (recorded as age appropriate based on the number of years of formal education completed by the mother), parity, marital status, tobacco or alcohol use (as dichotomous and categorical covariates based on the number of cigarettes per day or drinks per week), maternal weight gained during pregnancy, adequacy of prenatal care utilization (based on the Kotelchuck index)[17], and socioeconomic status. The poorest women were defined as those receiving food stamps, since this program includes only those with personal incomes less than 130% of the federal poverty level. Poor women were those defined as receiving Medicaid or WIC. In Missouri, pregnant women at or below 185% of the federal poverty level are eligible for Medicaid and WIC. While personal incomes can be used to identify women participating in poverty programs, we cannot assume that all non-participants are above 185% of the federal poverty level. All other women included those who did not participate in any of the three poverty assistance programs. Other covariates were clinical estimate of gestational age (20–28, 29–32, 33–36, ≥37 weeks) and birth weight (<2500, ≥ 2500 grams). Infant mortality included deaths that occurred during the neonatal (0–27 days) and post-neonatal (28–364 days) period. The underlying causes of death were divided into five categories: 1) accidental, 2) respiratory arrest or sudden infant death syndrome, 3) infectious, 4) perinatal or congenital, 5) other deaths.
Analysis
Chi-square tests and student's unpaired t-tests were used to determine if categorical and continuous covariates, respectively, were significantly different for younger versus older mothers. Overall infant mortality, as well as neonatal mortality and post-neonatal mortality, was assessed as a function of the three age cohorts and stratified by race. Logistic regression was used to assess how each covariate affected the role of maternal age on infant mortality risk. Covariates that altered the odds ratio (OR) for the primary relationship by more than 10% were considered significant and included in the full model. 95% confidence intervals (CI) were computed to estimate the precision of the OR. First-order interaction terms between each covariate and maternal age were tested in a secondary analysis. The significance of adding each interaction was tested with a p-value <0.01. The final model was examined for outliers, influential data and goodness of fit (Omnibus Test and Hosmer and Lemeshow test, p <0.05). The role of each covariate in the full model was also assessed using neonatal and post-neonatal mortality as the primary outcome in separate logistic regression analyses. SPSS (Chicago, IL, version 10.0) was used for all statistical analyses.
This research, reviewed by the Saint Louis University institutional review board, was classified as exempt from the U.S. Department of Health and Human Services regulations for the protection of human subjects. The exemption 45 CFR 46.101(b)(4) permits epidemiologic research that uses existing publicly available data that are maintained in such a manner that subjects cannot be identified directly or through identifiers linked to the subjects.
Results
The distribution of covariates and outcomes for the three maternal age cohorts is reported in Table 1. Comparing all adolescent mothers (12 to 19 years) to older mothers, newborns of older mothers were, on average, 167 grams heavier than newborns of younger mothers (p < 0.001). Although the younger mothers gained 1.36 kg (3 pounds) on average more during pregnancy than older mothers (p < 0.001), this is not likely to be clinically significant. Compared to older mothers, younger mothers were more likely to be nulliparous (72.6% vs. 29.1%), smoke (27.2% vs. 18.2%), of black race (26.7% vs. 13.8%), and on Medicaid (78% vs. 34.8%), WIC (75.1% vs. 36%), or food stamps (25.5% vs. 16.4%). Not surprisingly, fewer younger mothers were married (20.2% vs. 72.3%) and more had inadequate prenatal care (16.8% vs. 7%). Younger mothers were less likely to have an age-appropriate educational level (72.6% vs. 86.8%), and more likely to have a preterm infant (9.8% vs. 7.6%).
Table 1 Characteristics of Missouri women with singleton live births, 1997–1999, by maternal age.
12 to 17 (N = 10,131) 18 to 19 (N = 18,954) 20–35 (N = 28,899)
Characteristic N % N % N %
Infant mortality 121 1.2 158 0.8 173 0.6
Neonatal mortality 69 0.7 92 0.5 111 0.4
Post-neonatal mortality 52 0.5 65 0.3 60 0.2
Clinical Estimate of Gestational Age <24 Weeks 40 0.4 50 0.3 59 0.2
24–28 weeks 102 1.0 110 0.6 139 0.5
29–32 weeks 161 1.6 252 1.3 260 0.9
33–36 weeks 808 8.0 1,318 7.0 1,739 6.0
>36 weeks 8,961 89.0 17,109 90.8 26,571 92.4
Birth Weight Category Normal (> 2500 gm) 9,114 90.0 17,358 91.6 27,207 94.1
Low (≤ 2500 gm) 1,017 10.0 1,596 8.4 1,692 5.9
Maternal Race Missing/other 137 1.4 256 1.4 635 2.2
Non-Hispanic white 6,688 66.0 14,234 75.1 24,288 84.0
Non-Hispanic black 3,306 32.6 4,464 23.6 3,976 13.8
Age appropriate education level 8,592 86.1 12,516 66.0 25,086 86.8
Medicaid 7,698 77.1 14,983 80.4 10,070 35.4
W.I.C. 7,761 77.8 14,085 75.6 10,412 36.7
Food Stamp Program 2,060 20.7 5,351 28.7 4,729 16.7
Parity nulliparous 8,660 85.5 12,375 65.3 8,382 29.0
Unmarried 9,140 90.3 14,064 74.3 7,986 27.7
Tobacco use During Pregnancy 2,486 24.7 5,425 28.7 5,262 18.3
Alcohol use during pregnancy 61 0.6 90 0.5 268 0.9
Kotelchuck index Unknown 487 4.8 747 3.9 805 2.8
Inadequate 2,101 20.7 2,789 14.7 2,037 7.0
Adequate/Intermediate 5,273 52.0 11,121 58.7 18,832 65.2
Adequate Plus 2,270 22.4 4,297 22.7 7,225 25.0
p ≤ 0.001 for all comparisons
Overall, infants born of the youngest mothers (12 to 17 years) were 1.69 (CI 1.24, 2.31) times more likely to die during the neonatal period and 2.47 (CI 1.70, 3.59) times more likely during the post-neonatal period compared to older mothers Table 2 illustrates the risk of neonatal mortality and post-neonatal mortality for the adolescent age cohorts compared to older mothers, stratified by race. There is an increased risk of infant (9.7 deaths/1000 live births; OR 1.95, CI 1.42, 2.67), neonatal (5.3 deaths/1000 live births; 1.63, CI 1.02, 2.47) and post-neonatal mortality (4.4 deaths/1000 live births; OR 2.58, CI 1.56, 4.24) for the 12–17 year old non-Hispanic white mothers compared to older mothers (5.0, 3.2 and 1.7 infant, neonatal and post-neonatal deaths/1000 live births respectively). Among non-Hispanic white mothers 18–19 years of age, there is an increased risk of infant (6.7 deaths/1000 live births; OR 1.35, CI 1.02, 1.79) and post-neonatal (3.0 deaths/1000 live births; OR 1.75, CI 1.11, 2.75) mortality, but not neonatal mortality (3.7 deaths/1000 live births; OR 1.14, CI 0.79, 1.64). No association between age and infant, neonatal or post-neonatal mortality was seen for non-Hispanic black mothers. However, the crude rates of infant, neonatal and post-neonatal mortality for non-Hispanic black mothers were approximately twice those for non-Hispanic white mothers across all age groups.
Table 2 Infant, neonatal and post-neonatal mortality risk by age group, stratified by race
Race Age group Neonatal mortality Post-neonatal mortality
Rate (per 1000 births) cOR (95% CI) Rate (per 1000 births) cOR (95% CI)
Non-Hispanic White 12–17 years 5.3 1.63 (1.02, 2.47) 4.4 2.58 (1.56, 4.24)
18–19 years 3.7 1.14 (0.79, 1.64) 3.0 1.75 (1.11, 2.75)
20–35 years 3.2 reference 1.7 reference
Non-Hispanic Black 12–17 years 10.4 1.41 (0.84, 2.39) 7.0 1.63 (0.84,3.19)
18–19 years 9.0 1.23 (0.74, 2.04) 5.0 1.15 (0.59, 2.26)
20–35 years 7.3 reference 4.3 reference
cOR = crude odds ratio, CI = confidence intervals
Table 3 includes the results of the logistic regression model. Race, appropriate education level, marital status, parity, smoking during pregnancy, level of prenatal care and participation in Medicaid, WIC, or food stamp programs significantly affected the primary effect of maternal age on infant mortality risk and were included in the final model. Controlling for these variables, the risk for neonatal and post-neonatal mortality was 1.43 (CI 0.98, 2.08) and 1.73 (CI 1.14, 2.64), respectively, for the youngest adolescent mothers. There was no increased risk of neonatal or post-neonatal mortality for older adolescent mothers. Furthermore, adjusting for gestational age did not appreciably change any of the adjusted OR for the primary effect of age on infant mortality. First-order interactions between covariates and age, considered as a block, did not reach significance and were dropped from subsequent analyses.
Table 3 Logistic regression models: effect of maternal age on infant mortality.
Neonatal Mortality Post-Neonatal Mortality
Maternal Age (years) aOR 95% CI aOR 95% CI
12–17 1.43 0.98, 2.08 1.73 1.14, 2.64
18–19 1.15 0.83, 1.60 1.04 0.71, 1.53
20–34 1.00 reference 1.00 reference
Non-Hispanic black 2.08 1.56, 2.78 1.45 1.02, 2.04
Age-inappropriate educational Level 1.39 1.01, 1.90 1.73 1.23, 2.44
Unmarried 1.50 1.06, 2.14 1.25 0.83, 1.88
Nulliparous 1.00 0.99, 1.02 1.01 0.99, 1.03
Tobacco Use 1.49 1.10, 2.01 1.26 0.89, 1.79
Kotelchuck Index
Unknown 4.04 2.53, 6.45 0.88 0.35, 2.21
Inadequate 1.98 1.36, 2.88 1.58 1.05, 2.39
Adequate 1.00 reference 1.00 reference
Adequate Plus 2.89 2.17, 3.85 1.89 1.35, 2.66
Poverty Status
Poorest * 0.36 0.25, 0.53 2.29 1.36, 3.86
Poor ** 0.45 0.33, 0.62 1.60 0.98, 2.61
Other 1.00 reference 1.00 reference
cOR = crude odds ratio, aOR = adjusted odds ratio, CI = confidence intervals
* defined as participation in food stamps
** defined as participation in WIC or Medicaid but not food stamps
Of particular interest was the differential effect of the poverty construct in the model on neonatal and post-neonatal mortality risks. Poverty was defined by participation in Medicaid, WIC and/or food stamp programs, and the effect of such programs appeared to be strongly protective against neonatal mortality when controlling for maternal age, but a significant risk factor for post-neonatal mortality. Mothers participating in the food stamps program (our "poorest" category) had a reduced risk of neonatal mortality (OR 0.36, CI 0.25, 0.53) compared to mothers who did not participate in any of the poverty assistance programs. However, their risk of post-neonatal mortality (OR 2.29, CI 1.36, 3.86) was significantly higher. Stratifying the post-neonatal deaths by cause of death appears to show a higher percentage of deaths due to accidents or infections for the youngest mothers compared to older mothers in our study population (Table 4).
Table 4 Underlying causes of post-neonatal death by maternal age (n = 180).
Cause of death (%)
Maternal age (years) Accidental n (%) SIDS/respiratory arrest n (%) Perinatal/congenital n (%) Infectious n (%) All other n (%) Total n (%)
12–17 8(15) 16(31) 12 (23) 4(8) 12 (23) 52 (100)
18–19 11 (17) 34 (52) 9(14) 1(2) 11 (17) 66 (100)
20–35 4(6) 22 (35) 16 (26) 2(3) 18 (29) 62 (100)
chi-square = 14.3, p = 0.08
Discussion
We conclude that neonatal mortality rates do not differ by maternal age, but post-neonatal rates may be higher for mothers aged 12–17 years after adjusting for socioeconomic and other known risk factors. Our results are most similar to Reichman and Pagnini, who studied all New Jersey births in 1989 and 1990 [18]. They found that controlling for medical and socioeconomic factors accounted for some, but not all the effect of young maternal age on infant mortality. The unadjusted odds ratio of infant mortality among white 15–17 year olds in their study was 2.5; the adjusted value was 1.6. Their study was strengthened by access to linked uniform billing hospital discharge records in addition to state vital statistics. They were therefore able to assess socioeconomic status much more directly in their model, comparing Medicaid recipients and "self-pay" to the privately insured.
In our study, we were limited to inferring poverty status by the dichotomous values of participation in WIC, Medicaid, or food stamps programs as noted on the birth certificate. Recognizing the eligibility criteria for these programs (130% of federal poverty level for food stamps and less than 185% of poverty level for WIC and Medicaid) allowed us to predict confidently income levels of program participants, but we cannot conclude that those who are not participants are of higher income levels. Particularly in the aftermath of the Personal Responsibility and Work Opportunity Reconciliation Act of 1996, many individuals who were eligible may not have enrolled in such programs. Nevertheless, 75–80% of the two younger maternal cohorts were enrolled in Medicaid and/or WIC, compared to only 36% of the older mothers. Such dramatic differences are likely to be truly related to socioeconomic distinctions and not bureaucratic or regulatory obstacles unique to older women. This study is also limited by the validity of the original data entered on birth and death certificates. Of particular concern would be the possibility of differential bias in reporting of information between older and younger mothers since there is evidence that reporting infant mortality may suffer from a racial bias and the race distributions are not identical in our age cohorts [19].
Our finding that participation in WIC and Medicaid appears to be protective against neonatal mortality is consistent with a study by Moss and Carver (1998) [20]. Using data from the 1988 National Maternal and Infant Health Survey, among women whose household income was below 185% of federal poverty level, they found an adjusted odds ratio of infant mortality for WIC participation of 0.64 (CI 0.52, 0.77). In their full model, there was no significant effect of Medicaid, but this compares to an adjusted odds ratio of 1.39 (CI 1.04, 1.86) for "self-pay" individuals versus privately insured as the reference population. Recognizing that WIC and Medicaid, as well as the food stamps program, are not simply markers of low socioeconomic status but are also intervention programs, designed to improve the economic and nutritional condition of the disadvantaged – in particular pregnant women – may explain the apparent protective effect. Our study provides additional evidence that this is a biologically plausible observation. If we accept the premise that low socioeconomic status increases infant mortality by increasing the risk of preterm births, thus resulting in higher neonatal mortality, then adding the infant's gestational age into the logistic regression model should not significantly affect the primary relationship between maternal age and neonatal mortality, and indeed this is the case in our study.
Infant mortality is not a "black box;" the reasons newborns may die are very different from why 11 month olds die. As we can see, distinguishing neonatal from post-neonatal mortality is critical to appreciate fully the relationship between young maternal age and infant mortality. Certainly this increased post-neonatal mortality rate cannot be ascribed to "biologic" factors of younger mothers, where the debate over the mechanism of increased infant mortality with young maternal age has so often raged. The increased risk of post-neonatal mortality among the 12–17 year old mothers remains despite adjustment for covariates in our model suggesting social factors not explained by poverty (at least as measured on our model). It is interesting that the factors protective against neonatal mortality – participation in Medicaid, WIC and/or food stamp programs – are apparent risk factors for post-neonatal mortality (aOR 1.60 for poor and aOR 2.29 for poorest women in the 12–17 year olds). We speculate that whatever protective effect these programs have on neonatal mortality has dissipated after the newborn period, and then they are only markers of lower socioeconomic status. It is intriguing to consider what the causes of death might be in the post-neonatal period and how they differ in younger mothers. Despite our cohort of almost 60,000 infants and an apparent doubling of the accidental and infectious death rates in the youngest mother cohort, these differences did not achieve statistical significance.
Conclusion
In summary, this population-based cohort study confirms the known association of young maternal age and infant mortality, but adds considerably to our understanding by the distinction of younger from older teens and neonatal from post-neonatal mortality. Socioeconomic factors likely account for most of the young mothers' increased risk of neonatal mortality, but despite adjustment, a substantially increased risk of post-neonatal mortality exists for the youngest mothers. Our analysis points towards an increased risk of accidental and infectious deaths in these infants, raising questions of maternal maturity and ability to supervise adequately these developing infants.
Competing interests
The author(s) declare they have no competing interests.
Authors' contributions
The study hypothesis was proposed by LF. All authors shared in the design of this study. RC, BPM and TLL primarily carried out the analysis. BPM and RC drafted the manuscript. All authors read, edited and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We thank Garland Land, Joseph Stockbauer, and Janice Bakewell from the Missouri Department of Health and Senior Services for their helpful comments and support while conducting this study.
==== Refs
Stickle GM Ma P Pregnancy in adolescents: scope of the problem Contemporary OB/GYN 1975 5 85 95
National Center for Health Statistics Infant Mortality Statistics from the 1999 Period Linked Birth/Infant Death Data Set National Vital Statistics Report 2002 50 2001 1120 (PHS)
National Center for Health Statistics Births to Teenagers in the United States, 1940–2000 National Vital Statistics Report 2001 49 1 24 (PHS) 2001-1120
Miller MK Stokes CS Teenage fertility, socioeconomic statue and infant mortality Journal of Biosocial Science 1985 17 147 155 3997910
Dollfus C Patetta M Siegel E Cross AW Infant mortality: a practical approach to the analysis of the leading causes of death and risk factors Pediatrics 1990 86 176 183 2371093
Geronimus AT The effects of race, residence, and prenatal care on the relationship of maternal age to neonatal mortality American Journal of Public Health 1986 76 1416 1421 3777288
Geronimus AT Korenman S Maternal youth or family background? On the health disadvantages of infants with teenage Mothers American Journal of Epidemiology 1993 137 213 225 8452126
Zuckerman B Alpert JJ Dooling E Neonatal outcomes: is adolescent pregnancy a risk factor Pediatrics 1983 71 489 493 6835732
Hutchins FL Norman K Rubino J Experience with teenage pregnancy Journal of the American College of Obstetricians and Gynecologists 1979 54 1 5
Lee K Corpuz M Teenage pregnancy: trend and impact on rates of low birth weight and fetal, maternal, and neonatal mortality in the United States Clinics in Perinatology 1988 15 929 942 3061708
Fraser AM Brockert JE Ward RH Association of young maternal age with adverse reproductive outcomes The New England Journal of Medicine 1995 332 1113 1117 7700283 10.1056/NEJM199504273321701
Missouri Department of Health and Senior Services Accessed April 25, 2005
Vital and Health Statistics Trends in pregnancies and pregnancy rates by outcome: Estimates for the United States, 1976–1996 National Vital Statistics System 2000 21
Cowden AJ Funkhouser E Adolescent pregnancy, infant mortality, and source of payment for birth: Alabama residential live births, 1991–1994 Journal of Adolescent Health 2001 29 37 45 11429304 10.1016/S1054-139X(01)00217-8
DuPlessis HM Bell R Richards T Adolescent pregnancy: understanding the impact of age and race on outcomes Journal of Adolescent Health 1997 20 187 197 9069019 10.1016/S1054-139X(96)00174-7
Lawlor DA Shaw M Too much too young? Teenage pregnancy is not a public health problem International Journal of Epidemiology 2002 31 552 554 12055151 10.1093/ije/31.3.552
Kotelchuck M The adequacy of prenatal care utilization index: Its U.S. distribution and association with low birthweight American Journal of Public Health 1994 84 1486 1489 8092377
Reichman NE Pagnini DL Maternal birth outcomes: data from New Jersey Family Planning Perspectives 1997 29 268 272 9429872
Farley DO Richards T Bell RM Effects of reporting methods on infant mortality rate estimates for racial and ethnic subgroups Journal of Health Care for the Poor and Undeserved 1995 6 60 75
Moss NE Carver K The effect of WIC and Medicaid on infant mortality in the United States American Journal of Public Health 1998 88 1354 1361 9736876
|
16042801
|
PMC1190191
|
CC BY
|
2021-01-04 16:28:57
|
no
|
BMC Public Health. 2005 Jul 25; 5:79
|
utf-8
|
BMC Public Health
| 2,005 |
10.1186/1471-2458-5-79
|
oa_comm
|
==== Front
BMC Plant BiolBMC Plant Biology1471-2229BioMed Central London 1471-2229-5-131608079510.1186/1471-2229-5-13Research ArticleNew insights into the tonoplast architecture of plant vacuoles and vacuolar dynamics during osmotic stress Reisen Daniel [email protected] Francis [email protected] Nathalie [email protected] UMR PME INRA/CNRS/Université de Bourgogne BP 47870, boulevard Gabriel, 21078 Dijon Cedex, France2 Department of Molecular Biology and Genetics, 321 Biotechnology Building, Cornell University, Ithaca, NY 14853, USA2005 4 8 2005 5 13 13 8 4 2005 4 8 2005 Copyright © 2005 Reisen et al; licensee BioMed Central Ltd.2005Reisen et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The vegetative plant vacuole occupies >90% of the volume in mature plant cells. Vacuoles play fundamental roles in adjusting cellular homeostasis and allowing cell growth. The composition of the vacuole and the regulation of its volume depend on the coordinated activities of the transporters and channels localized in the membrane (named tonoplast) surrounding the vacuole. While the tonoplast protein complexes are well studied, the tonoplast itself is less well described. To extend our knowledge of how the vacuole folds inside the plant cell, we present three-dimensional reconstructions of vacuoles from tobacco suspension cells expressing the tonoplast aquaporin fusion gene BobTIP26-1::gfp.
Results
3-D reconstruction of the cell vacuole made possible an accurate analysis of large spanning folds of the vacuolar membrane under both normal and stressed conditions, and suggested interactions between surrounding plastids. Dynamic, high resolution 3-D pictures of the vacuole in tobacco suspension cells monitored under different growth conditions provide additional details about vacuolar architecture. The GFP-decorated vacuole is a single continuous compartment transected by tubular-like transvacuolar strands and large membrane surfaces. Cell culture under osmotic stress led to a complex vacuolar network with an increased tonoplast surface area. In-depth 3-D realistic inspections showed that the unity of the vacuole is maintained during acclimation to osmotic stress. Vacuolar unity exhibited during stress adaptation, coupled with the intimate associations of vacuoles with other organelles, suggests a physiological role for the vacuole in metabolism, and communication between the vacuole and organelles, respectively, in plant cells. Desiccation stress ensuing from PEG treatment generates "double" membrane structures closely linked to the tonoplast within the vacuole. These membrane structures may serve as membrane reservoirs for membrane reversion when cells are reintroduced to normal growth conditions.
Conclusion
3-D processing of a GFP-labeled tonoplast provides compelling visual constructions of the plant cell vacuole and elaborates on the nature of tonoplast folding and architecture. Furthermore, these methods allow real-time determination of membrane rearrangements during stresses.
==== Body
Background
Space-filling, turgor-driving vacuoles must have originated at a very early stage of biological evolution and have subsequently evolved to undertake various functions, well-known in algae, fungi (including yeast), and plants [1]. Many of the advanced, complex functions operate on or are closely associated with the vacuolar sap-bounding membrane, i.e. the tonoplast. A detailed structural study of dynamic events mediated by the tonoplast should extend our knowledge of its cellular functions. In the past, in vivo observations of the vacuolar membrane were restricted by the resolution of light microscopy. Like other cell components of a size below the limit of resolution of the light microscope, the tonoplast has been studied mostly by electron microscopy. However, this permits only post-mortem observations of thin sections from rapidly frozen freeze-substituted or chemically fixed cells [2], or replicas of fracture faces from fast-frozen cells [3,4]. Although these studies have provided invaluable insights into the organization and biogenesis of the tonoplast, analysis of thin specimens provides only limited information about its spatial architecture. High voltage electron microscopes have the ability to retrieve large amounts of information from thick sections through fixed cells, but depend on the use of a restricted number of selective (non-vital) "staining" techniques to overcome the decrease in image contrast typically seen at high accelerating voltages [4-6]. As such, the use of these different techniques makes it difficult to study and understand the dynamic changes of the vacuolar membrane in living cells.
The confocal fluorescence microscope can eliminate out-of-focus blur. Three-dimensional (3-D) data from intact biological specimens can therefore be obtained by non-invasive optical sectioning [7]. A new revolution in microscopy came over a decade ago with the use of "green fluorescent protein" (GFP) from jellyfish in vivo [8]. GFP and its variants are now frequently used to generate a fluorescent organelle, e.g. mitochondria [9], chloroplasts [10] or components of the secretory system [11], allowing one to study their dynamics in living cells. Furthermore, several studies present data of GFP-labeled tonoplasts [12-21], of which only a few followed tonoplast dynamics [20,21].
Since the best resolution of living cells is obtained with laser scanning confocal microscopy images, plant-compatible GFP cDNA [22] was fused, in frame, to the 3' end of the cauliflower BobTIP26-1 tonoplast-specific aquaporin cDNA cloned in our laboratory [23]. The resultant chimeric gene was expressed in tobacco cells (var. Wisconsin 38) grown in suspension, which display a GFP-decorated tonoplast [17]. This method of in vivo labeling was used to 3-dimensionally reconstruct the tonoplast of cells at different stages of growth and under various osmotic stresses. Our data show that the GFP-decorated vacuole in tobacco suspension cells is a single continuous compartment. Furthermore, osmotic stress conditions yield a greater tonoplast surface area while maintaining vacuole unity. Moreover, PEG treatment generates spherical structures that are associated with the inner side of the vacuolar membrane. We propose that these structures serve as membrane reservoirs necessary for membrane reorganization after cells are returned to normal growth conditions.
Results
Three-dimensional reconstruction of the vacuole
We have recently expressed a tonoplast aquaporin fusion gene BobTIP26-1::gfp under the CaMV35S promoter in tobacco cells [17]. The proper targeting of the resultant fusion protein to the tonoplast was confirmed by laser scanning confocal microscopy (Additional file 1). Tobacco cells expressing the chimeric protein were used for 3-D reconstructions of the cell vacuoles.
In 7 day-old turgid cells, vacuoles occupy almost the entire cell volume. In a projection view, where 40 optical sections of such cells were merged (z-step = 1 μm; Fig. 1a), a 3-D relief of the vacuole was barely observable, and structural details of the tonoplast were not easily discerned. Realistic 3-D pictures of the tonoplast (Fig. 1b, 1c) were obtained after isosurface extraction, a procedure in which volume images were converted into geometric surfaces [24] by using the 3-D visualization software Imaris 2.7 (Bitplane AG, Switzerland). The intravacuolar surface of the tonoplast was readily scrutinized when only "half vacuoles" were reconstructed. As expected, red autofluorescent chloroplasts were seen at the outer surface of the GFP-labeled membranes, confirming their cytoplasmic localization (Fig. 1b). Chloroplasts were also clustered around the nucleus, where the tonoplast forms a cavity (Fig. 1, arrow). Labeled transvacuolar cytoplasmic strands were seen radiating throughout the vacuole, extending from the nuclear region to the cell periphery (Fig. 1c, arrowheads). Furrows were also occasionally observed on the cytoplasmic side of the tonoplast (Fig. 2a, arrows). When the isosurface mode was used to visualize simultaneously chloroplasts and tonoplasts within large vacuolated cells, numerous chloroplasts were seen lying within these furrows (Fig. 2b).
Figure 1 Three-dimensional vacuole reconstruction of a vacuolated BobTIP26-1::gfp expressing cell 7 days after subculture. (a) Projection view of 40 confocal serial pictures corresponding to the half depth of the cell (i.e. 40 μm). Bar = 25 μm. (b) 3-D view after isosurface extraction showing the protoplasmic side of the vacuole. (c) Interior view of the vacuole after isosurface extraction. Green and red correspond to the tonoplast and the chloroplasts, respectively. Arrow: nuclear pouch; arrowheads: transvacuolar strands. The missing domains of the tonoplast surface in (b) and (c) result from an under-sampling of confocal images. The rendering of a completely smooth 3-D view would have required use of additional intermediate sections.
Figure 2 Outside shape of a vacuole. (a) Furrows on the outer part of the tonoplast (arrows) revealed by 3-D surface rendering. (b) Chloroplasts – in red – fill up the furrows on the tonoplast surface. Isosurfaces of both channels are displayed. Bar = 10 μm.
We observed cells of different sizes and shapes in the non-synchronized tobacco cell suspension. While confocal images suggest the existence of several vacuoles inside each cell (Fig. 3a, 3c, 3e), 3-D representations of each cells' tonoplast, and the visualization of openings within the vacuolar lumens clearly support a "one cell, one vacuole" model, i.e. vacuolar continuity exists between GFP-labeled vacuolar compartments within a cell (Fig. 3b, 3d, 3f). Indeed, large surface areas of tonoplast may transect the vacuole, but gaps exist, allowing continuity of the vacuolar interior. The three joined cells (Fig. 3c) each possesses such transvacuolar layers, but 3-D representations show that they do not define discrete vacuoles in each cell (Fig. 3d). An animated sequence of these 3-D reconstructed vacuoles is shown in the Additional File 2. What seemed to be individual vacuolar cavities in the cells observed by confocal microscopy was seen as a single vacuolar compartment by 3-D reconstruction. Careful analysis through a tomogram of a protoplast prepared from BobTIP26-1::gfp expressing tobacco cells is also in accord with this feature (Additional File 3).
Figure 3 Vacuole continuity through gaps inside tonoplast sheets. (a, c, e) Single confocal images. The focal plane chosen corresponds to the one through the vacuoles' center. Arrowheads (>) in (c) delimit joined cells. n, nuclear region. Bars = 50 μm. (b, d, f) 3-D representation of the vacuole halves after isosurface extraction. Notice the openings in the vacuolar layers (b and d) and the many transvacuolar strands emanating from the nuclear region (f).
Cellular architecture details that are evident as a result of 3-D reconstructions include the demonstration of a vacuole with numerous transvacuolar strands, outwardly radiating from the nuclear region to the cell periphery (Fig. 3e, 3f). 3-D reconstructions also showed that chloroplasts do not merely surround the nucleus; their distribution also parallels that of the transvacuolar strands.
Tonoplast behavior during plasmolysis
We further analyzed the behavior of the labeled tonoplast during plasmolysis. Tobacco suspension cells were bathed in culture media supplemented with different osmotica (170 mM sodium chloride, 0.6 M mannitol or 0.5 M sorbitol) to induce plasmolysis-deplasmolysis cycles. Plasmolysis occurred 30 to 60 sec after the osmoticum entered the perfusion chamber. The phenomenon preferentially started at the cell corner, with complete plasmolysis occurring after 3 to 5 min. Two forms of plasmolysis were observed in tobacco cells, as defined by Oparka [25]. The convex form of plasmolysis was most frequently observed, which results from an even separation of the protoplast from the walls, thereby forming a symmetrical and roughly spherical protoplast (Fig. 4a). Occasionally, the concave form of plasmolysis was observed (Fig. 4b). In this form, concave pockets are formed as the plasma membrane separates from the wall. During both form of plasmolysis, the protoplast was tightly connected to the cell wall by Hechtian strands (Fig. 4c, arrow).
Figure 4 Cell perfusion with MS supplemented with 0.6 M mannitol. (a) A confocal fluorescent image merged with a Nomarski interference contrast image of convex plasmolysed cells. Arrowhead: transvacuolar strand. Bar = 10 μm. (b) A confocal fluorescent image merged with a Nomarski interference contrast image of a concave plasmolysed cell. n, nucleus. Bar = 10 μm. (c) A Nomarski interference contrast image of a plasmolysed cell. Hechtian strands (arrow) attach the protoplast tightly to the cell wall. n, nucleus. Bar = 5 μm.
During the plasmolysis step, a peculiar folding of the GFP-labeled tonoplast characterized by complex curling surrounding the nucleus developed within the vacuole (Fig. 4a, 4b). Closer examination of the curled structures revealed that they were slightly more fluorescent than the tonoplast to which they were tethered, an observation that might be due to the joining of two adjacent labeled membranes. Furthermore, the vacuole and the tonoplast remained intact and no vesicle formation was detected. A cellular tomogram through a plasmolysed cell clearly demonstrates the uninterrupted integrity of the tonoplast, its folds and curves being readily visible (Additional File 4). The absence of tonoplast inside the Hechtian strands was confirmed by the absence of labeling (Additional File 4). However, transvacuolar cytoplasmic strands could still be observed inside the vacuoles of plasmolysed (Fig. 4a, arrowhead; Additional File 5) and deplasmolysed cells (Additional file 5). These intravacuolar structures appear to be less flexible than the peripheral tonoplast. Cells were kept plasmolysed for 10 to 15 min before they were bathed again in normal MS culture medium. Plasmolysed cells returned to their normal shape after about 10 to 20 min (Additional File 5), and the curled structures unwound. The position of the nucleus within the cell was nearly constant throughout the plasmolysis-deplasmolysis process.
Effects of osmotic stress on tonoplast architecture
Consistent with the results obtained with 3-D reconstructions, we tried to ascertain the tonoplast architecture during osmotic stress culture. Indeed, during NaCl culture conditions of tobacco suspension cells, a vacuolization phenomenon, characterized by a fragmentation of the central vacuole into multiple smaller ones, has been described [26]. To prove the existence of this fragmentation in tobacco cells exhibiting a fluorescent tonoplast, BobTIP26-1::gfp expressing cells were subcultured in hyper-osmotic MS medium containing 170 mM NaCl, 0.6 M mannitol or 0.5 M sorbitol, with final water potentials of -1.4 MPa; -1.6 MPa and -1.2 MPa, respectively. Compared with cells subcultured in normal MS medium (Fig. 5a, 5b), cells acclimated in hyperosmotic medium were smaller (Fig. 5c, 5d). [Hereafter, cells which have been grown in media containing osmoticum are referred to as "acclimated cells"]. The extrapolated volumes of cells grown in normal MS and in hyperosmotic medium are on average 175·103 μm3 and 53·103 μm3, respectively. After acclimation, cell volume decreases ~3.3 times. If the vacuole represents 90% of the cell volume in cells grown in normal MS medium, the original vacuole volume would be 157·103 μm3.
Figure 5 Normal cells versus osmotic stressed cells. (a, b) Cells under normal growth conditions. Bar = 25 μm. (c, d) Cells grown in MS supplemented with 0.6 M mannitol (i.e. acclimated cells). Arrows correspond to dead cells (i.e. non-acclimated cells). Bar = 25 μm. (e) 3-D reconstruction of the vacuole after isosurface extraction from acclimated cells cultured in MS supplemented with 0.5 M sorbitol.
The vacuolization phenomenon could be monitored under Nomarski contrast interference optics (Fig. 5c). Furthermore, this phenomenon was more easily observed with fluorescent labeling of the tonoplast (Fig. 5d). In these cultures, only ~25% of the cells were completely plasmolysed (Fig. 5c arrows), and did not show any fluorescence, indicating that they are non-acclimated, dead cells (Fig. 5d arrows). Additionally, vital staining with neutral red showed that only acclimated cells contained the dye in their vacuoles (data not shown).
To determine if the apparently numerous, independent, small vacuoles of acclimated cells are part of a continuous vacuolar compartment, 3-D reconstructions of the entire vacuole were achieved. The images indicated the existence of a complex vacuolar network. A depth inspection with stereo-viewers showed interconnections between the vacuolar cavities in completely reconstructed vacuoles. Halves of such reconstructed vacuoles are presented in Figure 5e. The nuclear pouch can be seen at the cell center, surrounded by small, interconnected vacuolar cavities. Despite strong evidence for the existence of a solitary vacuolar space, we cannot rule out the possibility that acclimated cells contain small, discrete cavities. However, no small structures of this nature could be discerned using the complicated reconstructions. Since vacuolar cavities are interconnected, we suggest that the tonoplast surface area increases significantly while the continuity of the vacuolar lumen remains unaltered. The software (Imaris 2.7) used for the 3-D reconstructions did not allow quantification of the tonoplast surface area, nor comparison of the tonoplast surfaces of acclimated cells to those of normal cultured cells.
Tonoplast architecture modifications during dehydrative stress
Dehydrative stress was mimicked using the macromolecule polyethylene glycol (PEG8000), which passes with either difficulty or not at all through the cell wall [27], thereby reducing the extracellular free water concentration [28]. While an overall increase in membrane surface area occurred in hyperosmotically stressed suspension cells, cells cultured in MS medium, supplemented with 10% PEG8000, displayed numerous spherical fluorescent structures of 5–10 μm in diameter, but neither complex membrane rearrangement nor folding. These inner vacuolar structures exhibited a brighter membrane fluorescence compared to that of the peripheral tonoplast (Fig. 6a, 6b, arrowheads). Indeed, the fluorescence intensity of the membrane surrounding these structures was at least twice that of the peripheral tonoplast (Fig. 6c). These spherical structures were also dynamic, moving inside the vacuolar lumen, but seemingly attached to the peripheral vacuolar membrane (Fig. 6d, arrowhead). They were easily distinguishable from the transvacuolar strands, which maintained their morphology (Fig. 6e, arrow). Indeed, the transvacuolar strands transect the lumen of the vacuole from one part of the vacuole to the other, while the spherical structures are tethered to the vacuole perimeter.
Figure 6 PEG-stressed cells. (a, b) Localization of BobTIP26-1::GFP in the membrane of spherical structures (arrowheads) within vacuoles. Bar = 10 μm. (c) A histogram of fluorescent intensity values collected from both the tonoplast of the cell periphery and vesicles. (d) A 3-D slice reconstruction through the vacuole at the position of the nucleus (n). Arrowhead = spherical structure. (e) 3-D reconstruction through a part of a vacuole. Transvacuolar strands are still present in these cells (arrow); spherical structures are primarily fixed onto the tonoplast (arrowheads).
Discussion
3-D reconstruction of the vacuole under native conditions
The plant vacuole is a multi-functional organelle [1] which serves as a true milieu intérieur [29], playing key roles during cell growth [30], and possibly in the osmoregulation of water during osmotic stress of the cytoplasm [31]. For the purpose of redefining the vacuole architecture, we present the use of 3-D reconstructions of the vacuolar apparatus from cells containing a GFP-labeled tonoplast under "native" conditions. The tonoplast marker system [17] was developed to gain new insights of the tobacco cell vacuole morphology during both normal and osmotic stress growth conditions. A 3-D representation of the vacuole using confocal serial pictures offers the most accurate illustration of how a vacuole is folded within a plant cell. Indeed, previous examination of mitotic BY-2 cells expressing a GFP-labeled syntaxin yielded comprehensive images of vacuolar architecture by reconstructing 3-D surfaces obtained from sequential confocal sections [32]. A side effect of the morphology of the tonoplast due to the over-expression of a membrane protein such as BobTIP26-1::GFP can, a priori, not be excluded, but results obtained for both 35S-GFP-AtVam3p and the lipophilic probe FM4-64 in BY-2 cells revealed that vacuolar morphology was not artificially affected in transgenic BY-2 cells [32]. The fluorescent tonoplast in BobTIP26-1::gfp expressing cells appears to be similar to the FM4-64 stained one in BY-2 cells, assuming that vacuolar morphology in BobTIP26-1::gfp expressing cells is not altered.
The software used for the representation of the laser confocal microscope data offered two forms of volume visualization: isosurface rendering and direct volume rendering [33]. We chose the isosurface rendering method, in that it allowed us to represent a shaded surface as an easily interpretable 3-D object. However, the two techniques are complementary and the conclusions offered by the final 3-D pictures are equivalent.
A 3-D vacuole structure of BY-2 tobacco cells expressing TIP::gfp was first established by Mitsuhashi and co-workers [13], but their representation was mainly a projection view of sequential confocal images that revealed the presence of several large vacuoles folded within the cell. However, as demonstrated in the present study, a projection view may not be as informative as 3-D reconstruction (Fig. 1, Fig. 3). Whereas projection views showed multiple vacuoles folded within a single cell, 3-D reconstructions accurately portray a single vacuolar continuum within cells cultivated under normal growth conditions (Fig. 3). Such continuity of the vacuolar lumen has been described by Palevitz and co-workers [34], who analyzed the vacuole during cell differentiation of Allium stomata cells. Although thin sections of these fixed cells observed by electron microscopy revealed individual small vacuoles, a 3-D reconstruction using 0.25 to 0.50 μm thick serial sections, viewed at 100 kV, clearly showed the sections originated from a continuous reticulate network.
The concept of a continuous vacuole is important for the plant cell because the content of the vacuolar sap can thereby flow between all regions of the cell when homeostatic measures are undertaken by stressed cells.
To obtain enhanced resolution of the vacuole, reconstructions based on electron tomograms should be used as exemplified in the data obtained for both Arabidopsis thaliana mitotic cell plate formation [35-37] and the Golgi apparatus of animal [38] and yeast [39] cells.
Tonoplast surface architecture
With 3-D reconstructions of the vacuole, we were able to analyze its surface architecture from a unique perspective. Earlier 3-D renderings of the vacuole-cytoplasm interface of tobacco cell vacuoles showed ripples on the acridine orange labeled vacuole surface [40], which differ from surface furrows we have described. Indeed the vacuole ripples dip into the cytoplasm, but not into the vacuole [40]. We observed the furrows just after isosurface reconstruction and they seem to result from the pressure exerted by some organelles onto the turgid vacuole. The organelles were thus squeezed between the plasma membrane and the tonoplast creating hence the furrows on the cytoplasmic interface with the vacuole. We show here that the 3-D rendering is also useful to visualize interactions between organelles.
The tonoplast surface is enlarged during osmotic stress
Although membranes are targets of stress-induced cellular damage, and vacuoles are thought to have a central role during stress, few studies of the tonoplast during stress conditions have been reported. We reasoned that scrutinizing the pattern of the tonoplast during different osmotic stresses would build upon our current knowledge of how the cell organizes this membrane during environmental changes. Glycophytic cells, such as tobacco cells (Nicotiana tabacum var. Wisconsin 38), were previously analyzed for their ability to acclimate to NaCl [41] and determined to have an increased vacuolization as well as an extensive network of transvacuolar membrane strands [26]. Transvacuolar strands are easily visualized with Normarski optics under a light microscope; however, their composition is more difficult to deduce. Our use of the tonoplast intrinsic GFP-tagged protein that we developed [17] offers in situ evidence that transvacuolar strands transecting the vacuole lumen also contain tonoplast associated molecules (Fig. 1, Fig. 3f).
Tobacco cells growing in the presence of diverse osmotica developed an extensive membrane network that appears to transect the vacuole, thereby creating multiple compartments. Observations of similar phenomena were previously described for cells grown during NaCl acclimation [26]. With a 3-D representation, we found that growth in media supplemented with salt, mannitol or sorbitol resulted in transvacuolar strands completely changing their organizational pattern, shifting from a thin shape during normal conditions to larger surface areas. Concomitant with strand pattern alterations was the appearance of several small cavities. These phenomena were observed regardless of the osmoticum used for acclimation. The 3-D picture obtained from acclimated cells clearly demonstrates that the number of vacuoles remains constant while the total tonoplast surface area increases, thereby creating a more complex vacuolar pattern (Fig. 5e). Thus, the ratio of tonoplast surface area to volume of cytoplasm is optimized for exchanges (transport of water and ions) between the cytosol and the vacuole. As the total surface area of tonoplast membrane increases, it follows that the total mass of membrane compounds also increases. It is notable that similar vacuolization changes were observed to occur after hyper-osmotic treatment of wild type tobacco cells; therefore, the increased membrane surface is linked neither to the presence of the aquaporin nor the greater cell size of BobTIP26-1::gfp expressing cells [17]. Measurements of the surface areas of the tonoplasts in acclimated cells are complicated by the absence of vacuole shape uniformity. In contrast, the prolate spheroid shape of Vicia faba guard cells allows for the determination of both their surface areas and volumes after 3-D reconstruction efforts [42]. In our studies such standardization of conditions was not possible.
Folding of the plasmolytic vacuole
Notwithstanding the massive changes in vacuolar surface area which accompany plasmolysis, only a few observations on the fate of the vacuole during osmotic contraction have been described in the relevant literature. Plasmolysis-deplasmolysis cycles were realized in a perfusion chamber, allowing observation of fluorescent tonoplast labeled with GFP. During plasmolysis, it was observed that the tonoplast undergoes folding and that the vacuole does not vesiculate into discrete multilamellar vesicles severed from the tonoplast as "sac-like, rod-like or doughnut-shaped structures" as previously described [43]. Furthermore, complete tomography of a plasmolysed cell (Additional File 4) demonstrated that the tonoplast folds in a peculiar way inside the protoplast, rather than being broken into small vesicles, as reported earlier [25]. Such a process seems better suited than vesicle formation for a faster reestablishment of the vacuole during deplasmolysis. No membrane fusion is necessary.
Transvacuolar strands, thin tubular structures that traverse the vacuole, were observed not to break down both during plasmolysis and after deplasmolysis (Additional File 5). Earlier studies showed that transvacuolar strands were stabilized by actin filaments [44,45] and rearranged by myosin motors through their interactions with actin filaments [46]. Despite these observation of strand stability, they have also been described as dynamic and delicate [26]. Our data support the notion that transvacuolar strands exhibit both "strength" and stability, in that their positions remaining static inside the vacuole during abrupt environmental changes (Additional File 5).
PEG treatment results in spherical structures composed of tonoplast
Surprisingly, spherical structures with twice the fluorescent intensity as the tonoplast were observed inside the vacuole lumen of tobacco suspension cells cultured in MS medium supplemented with 10% PEG8000. These structures were similar to those frequently observed in either tonoplast GFP-labeled tobacco leaves [47,48] or in Arabidopsis cotyledons [49]. Additionally, such structures have been described in cotyledons, hypocotyls and roots of the Arabidopsis vacuolar biogenesis bub (bubble-bath) mutants [50], as well as in transiently transformed Nicotiana benthamiana plants expressing a GFP fusion protein homologous to a high-affinity tonoplast phosphate transporter [51]. The structures we observed during PEG treatment were independent of the water regulation or over-expression of an aquaporin gene because we, as well as other authors, found similar bulbs in Arabidopsis transformed cells with native TIP1;1 promoter (Bouhidel K., personal communication) and in cells expressing other tonoplast proteins under the control of the 35S promoter [51]. The structures observed in Arabidopsis cotyledons [49] were called "bulbs", a term which reflects a spherical structure that is not entirely closed. The open section of these bulbs may be the result of incomplete vesicle formation [49]. The bulbs appear to be attached to the actin cytoskeleton, as actin inhibitor treatment allowed their immobilization and 3-D reconstructions [51]. The 3-D images of the spherical structures we observed in tobacco suspension cells cultured under dehydration stress are similar to the 3-D reconstructed cylindrical structures observed in GFP-AtVam3p expressing root protoplasts of A. thaliana [52]. Recently, fluorescent circular structures were seen in germinating pollen tubes of δ -TIP::GFP expressing Arabidopsis plants [20]. These mobile cytoplasmic invaginations may be a widespread characteristic of actively growing tissues.
Why do vacuoles of PEG acclimated suspension cells contain "bulbs"? PEG mimics dehydration when applied to the culture medium, resulting in a decrease of turgor. Similarly, cotyledons undergo a desiccation process. In Spirodela intermedia upper mesophyll cells, a breakdown of the tonoplast into small vesicles was observed after PEG inclusion [53]. Correlation of observed membrane fluorescence intensity with protein quantity suggests that the membrane of the spherical vacuolar structures is likely to contain twice the amount of aquaporins per surface area as the peripheral tonoplast, suggesting a higher level of water exchange. The spherical structures could reflect the presence of lipid domains in the tonoplast where aquaporins are concentrated and vesicle formation occurs. Indeed, some studies showed a high tonoplast fluidity [54], as well as a specific feature of fatty acid composition that may be responsible for the tonoplasts' unique fluidity and high elasticity [55] required for osmotic processes in the cell. Our results support the invagination model, postulated by Uemura and co-workers [52], where parts of the tonoplast form a double-layered membrane structure inside the vacuolar lumen. The spherical structures could serve as a reservoir, not only for membrane expansion, but also to allow for quicker homeostasis adjustments. A higher tonoplast surface area to cell volume ratio would greatly enhance the cells' capacity to maintain large ion pools during growth [56].
Conclusion
The data presented in this study demonstrates the utility of the aquaporin BobTIP26-1::GFP as a powerful tool for visualizing 3-D membrane rearrangements within stressed tobacco cells. The technique used here provides highly resolved pictures and support of the notion that tobacco suspension cells contain a single major vacuole with a lytic function. Further investigations are required to establish the exact origin and function of the membrane enclosed, intra-vacuolar circular structures exhibited by tobacco cells under PEG mediated stress.
Methods
Plant material
Tobacco (Nicotiana tabacum L. var. Wisconsin 38) suspension cells expressing BobTIP26-1::gfp [17] were grown in MS medium [57], with regular shaking, at 24°C under constant photosynthetic illumination (200 μE·m-2·sec-1).
Three-dimensional reconstruction
Three to seven day-old cells were analyzed under a Leica TCS 4D laser confocal microscope (Leica Microsystems, Wetzlar, Germany) equipped with an argon-krypton laser (488/515 BP-FITC). The laser was focused on individual cells through a 40x NA1 oil-immersion objective. For each cell, a stack of between 100 and 200 images was collected (resolution 256 × 256 with 0.50 to 0.65 μm of z-step). Merged individual confocal images (red and green channels) were composed using Corel Photo-Paint 7 software (Corel, Ottawa, Canada). To obtain 3-D reconstructions, confocal image stacks were imported into the three-dimensional visualization software Imaris 2.7 (Bitplane AG, Switzerland) running on a Silicon Graphics® Octane2™ workstation (SGI, Paris, France). After baseline subtraction, a subregion was defined. Then, the isosurface module of Imaris was used to reconstitute the 3-D pictures. An adequate isovalue was defined for each channel prior to viewing of the computed surface using IvView. Isosurface rendered pictures were then stored as tiff files using the MediaRecorder media tool. Realistic on-screen representations of 3-D reconstructions were analyzed in depth with stereoscopic viewing devices (CrystalEyes®, StereoGraphics®).
Cell perfusion
Cell perfusion was performed in a homemade perfusion chamber linked to a peristaltic pump (flux = 20 mL·h-1) (IBMP, Strasbourg, France). BobTIP26-1::gfp expressing cells were first adhered to poly-lysinized glass cover slips. The following solutions were used for cell perfusions: MS supplemented with either 0.17 M NaCl or 0.6 M mannitol or 0.5 M sorbitol. Plasmolysis and deplasmolysis were monitored using an inverted laser confocal microscope LSM510 (Zeiss Axiovert 100 M, Jena, Germany) fitted with a Zeiss 63X water-immersion objective. Confocal time lapse series were then collected. Osmolarities were measured with a Vapro® vapor pressure osmometer (Model 5520; Wescor, Logan, UT, USA).
Acclimation stress conditions
Five mL of BobTIP26-1::gfp expressing cells at exponential growth phase, were subcultured in MS medium supplemented with any of the following osmotica: 170 mM NaCl, 0.6 M mannitol, 0.5 M sorbitol or 10% PEG8000. The stressed and control (MS without an osmoticum) cells were analyzed over 7 days and 3-D reconstructed as described above. Fluorescent intensities were measured using IPLab software (Scanalytics, Fairfax, VA, USA).
Authors' contributions
DR drafted the manuscript, carried out the experiments, including the microscopic observations and computer generated reconstructions. FM and NLC conceived of the study, and participated in its design and coordination. NLC also helped to draft the manuscript. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Cellular tomogram through two overlapping BobTIP26-1::gfp expressing cells.
Click here for file
Additional File 2
Animation of 3-D reconstructed vacuoles. The reconstructed vacuoles from Figure 3d were rotated. Notice chloroplasts (red) on the outer sides of the vacuoles, within transvacuolar strands and the cells' peripheries.
Click here for file
Additional File 3
Cellular tomogram through a protoplast with a GFP-labeled tonoplast. The protoplast was prepared from BobTIP26-1::gfp expressing cells as previously described [17].
Click here for file
Additional File 4
Cellular tomogram through a plasmolysed cell demonstrating the intact tonoplast structure and its folding. The tonoplast fluoresces in green. Hechtian strands are not labeled. Bar = 5 μm.
Click here for file
Additional File 5
Plasmolysis-deplasmolysis cycle of cells bathed with 0.5 M sorbitol. Time is marked at the bottom left; changes of media are given at the top left (10 min: + 0.5 M sorbitol; 23.5 min: + normal MS medium). Bar = 20 μm.
Click here for file
Acknowledgements
The authors would like to thank J.C. Robbins for English corrections and meaningful discussions. D.R. was supported by the Ministère de l'Education Nationale de la Recherche et de la Technologie (N° 99-5-12166). The Octane2™ workstation and the Imaris software were used at the CMAB-SERCOBIO (Université de Bourgogne). We thank C. Ritzenthaler at The Inter-Institute Confocal Microscopy Plate-Form in the Institut de Biologie Moléculaire des Plantes (Strasbourg) for his assistance with the perfusion experiments. The Conseil Régional de Bourgogne financed this work.
==== Refs
Marty F Plant Vacuoles. Plant Cell 1999 11 587 600 10213780 10.1105/tpc.11.4.587
Buvat R Electron microscopy of plant protoplasm. Int Rev Cytol 1963 14 41 55 14283576
Fineran BA Association between endoplasmic reticulum and vacuoles in frozen-etched root tips. J Ultrastruct Res 1973 43 75 87 4349960
Marty F High voltage electron microscopy of membrane interactions in wheat. J Histochem Cytochem 1980 28 1129 1132 7419901
Cosslet VE The "ultrastructure" of tissues. Revelations by electron-microscope. Radiography 1962 28 203 204 13881557
Favard P Cassano N The preparation and observation of thick biological sections in the high voltage electron microscope. J Microsc 1973 97 59 81 4122662
Amos WB White JG How the confocal laser scanning microscope entered biological research. Biol Cell 2003 95 335 342 14519550 10.1016/S0248-4900(03)00078-9
Tsien RY The green fluorescent protein. Annu Rev Biochem 1998 67 509 544 9759496 10.1146/annurev.biochem.67.1.509
Köhler RH Zipfel WR Webb WW Hanson MR The green fluorescent protein as a marker to visualize plant mitochondria in vivo. Plant J 1997 11 613 621 9107047 10.1046/j.1365-313X.1997.11030613.x
Köhler RH Cao J Zipfel WR Webb WW Hanson MR Exchange of protein molecules through connections between higher plant plastids. Science 1997 276 2039 2042 9197266 10.1126/science.276.5321.2039
Brandizzi F Irons SL Johansen J Kotzer A Neumann U GFP is the way to glow: bioimaging of the plant endomembrane system J Microsc 2004 214 138 158 15102062 10.1111/j.0022-2720.2004.01334.x
Ueoka-Nakanishi H Tsuchiya T Sasaki M Nakanishi Y Cunningham KW Maeshima M Functional expression of mung bean Ca2+/H+ antiporter in yeast and its intracellular localization in the hypocotyl and tobacco cells. Eur J Biochem 2000 267 3090 3098 10806410
Mitsuhashi N Shimada T Mano S Nishimura M Hara-Nishimura I Characterization of organelles in the vacuolar-sorting pathway by visualization with GFP in tobacco BY-2 cells. Plant Cell Physiol 2000 41 993 1001 11100771 10.1093/pcp/pcd040
Czempinski K Frachisse JM Maurel C Barbier-Brygoo H Mueller-Roeber B Vacuolar membrane localization of the Arabidopsis 'two-pore' K+ channel KCO1. Plant J 2002 29 809 820 12148538 10.1046/j.1365-313X.2002.01260.x
Kutsuna N Hasezawa S Dynamic organization of vacuolar and microtubule structures during cell cycle progression in synchronized tobacco BY-2 cells. Plant Cell Physiol 2002 43 965 973 12354913 10.1093/pcp/pcf138
Yamaguchi H Nishizawa NK Nakanishi H Mori S IDI7, a new iron-regulated ABC transporter from barley roots, localizes to the tonoplast. J Exp Bot 2002 53 727 735 11886893 10.1093/jexbot/53.369.727
Reisen D Leborgne-Castel N Özalp C Chaumont F Marty F Expression of a cauliflower tonoplast aquaporin tagged with GFP in tobacco suspension cells correlates with an increase in cell size. Plant Mol Biol 2003 52 387 400 12856944 10.1023/A:1023961332391
Yano K Matsui S Tsuchiya T Maeshima M Kutsuna N Hasezawa S Moriyasu Y Contribution of the plasma membrane and central vacuole in the formation of autolysosomes in cultured tobacco cells. Plant Cell Physiol 2004 45 951 957 15295079 10.1093/pcp/pch105
Kataoka T Watanabe-Takahashi A Hayashi N Ohnishi M Mimura T Buchner P Hawkesford MJ Yamaya T Takahashi H Vacuolar sulfate transporters are essential determinants controlling internal distribution of sulfate in Arabidopsis. Plant Cell 2004 16 2693 2704 15367713 10.1105/tpc.104.023960
Hicks GR Rojo E Hong S Carter DG Raikhel NV Geminating pollen has tubular vacuoles, displays highly dynamic vacuole biogenesis, and requires VACUOLESS1 for proper function. Plant Physiol 2004 134 1227 1239 14988481 10.1104/pp.103.037382
Saito C Morita MT Kato T Tasaka M Amyloplasts and vacuolar membrane dynamics in the living graviperceptive cell of the Arabidopsis inflorescence stem. Plant Cell 2005 17 548 558 15689424 10.1105/tpc.104.026138
Haseloff J Siemering KR Prasher DC Hodge S Removal of a cryptic intron and subcellular localization of green fluorescent protein are required to mark transgenic Arabidopsis plants brightly. Proc Natl Acad Sci USA 1997 94 2122 2127 9122158 10.1073/pnas.94.6.2122
Barrieu F Thomas D Marty-Mazars D Charbonnier M Marty F Tonoplast intrinsic proteins from cauliflower (Brassica oleracea L. var. botrytis): immunological analysis, cDNA cloning and evidence for expression in meristematic tissues. Planta 1998 204 335 344 9530877 10.1007/s004250050264
Morgan F Barbarese E Carson JH Visualizing cells in three dimensions using confocal microscopy, image reconstruction and isosurface rendering: application to glial cells in mouse central nervous system. Scanning Microsc 1992 6 345 357 1281337
Oparka KJ Plasmolysis: new insights into an old process. New Phytol 1994 126 571 591
Chang PFL Damsz B Kononowicz Reuveni M Chen Z Xu Y Hedges K Tseng CC Singh NK Binzel ML Narasimhan ML Hasegawa PM Bressan RA Alterations in cell membrane structure and expression of a membrane-associated protein after adaptation to osmotic stress. Physiol Plant 1996 98 505 516 10.1034/j.1399-3054.1996.980310.x
Baron-Epel O Hernandez D Jiang LW Meiners S Schindler M Dynamic continuity of cytoplasmic and membrane compartments between plant cells. J Cell Biol 1988 106 715 721 3346323 10.1083/jcb.106.3.715
Rains DW Jones HG, Flowers TJ and Jones MB Plant tissue and protoplast culture: applications to stress physiology and biochemistry. Plants under stress 1989 Cambridge, Cambridge University Press 181 169
Marty F Leigh A and Sanders D The biogenesis of vacuoles: insights from microscopy. The plant vacuole Advances in Botanical Research 1997 25 San Diego, Academic Press 1 42
Raven JA The role of vacuoles. New Phytol 1987 106 357 422
Hasegawa PM Bressan RA Zhu JK Bohnert HJ Plant cellular and molecular responses to high salinity. Annu Rev Plant Physiol Plant Mol Biol 2000 51 463 499 15012199 10.1146/annurev.arplant.51.1.463
Kutsuna N Kumagai F Sato MH Hasezawa S Three-dimensional reconstruction of tubular structure of vacuolar membrane throughout mitosis in living tobacco cells. Plant Cell Physiol 2003 44 1045 1054 14581629 10.1093/pcp/pcg124
Razdan A Patel K Farin GE Capco DG Volume visualization of multicolor laser confocal microscope data. Comp Graph 2001 25 371 382 10.1016/S0097-8493(01)00061-9
Palevitz BA O'Kane DJ Kobres RE Raikhel NV The vacuole system in stomatal cells of Allium vacuole movements and changes in morphology in differentiating cells as revealed by epifluorescence, video and electron microscopy. Protoplasma 1981 109 23 55 10.1007/BF01287629
Otegui MS Mastronarde DN Kang BH Bednarek SY Staehelin LA Three-dimensional analysis of syncytial-type cell plates during endosperm cellularization visualized by high resolution electron tomography. Plant Cell 2001 13 2033 2051 11549762 10.1105/tpc.13.9.2033
Otegui MS Staehelin LA Electron tomographic analysis of post-meiotic cytokinesis during pollen development in Arabidopsis thaliana. Planta 2004 218 501 515 14610676 10.1007/s00425-003-1125-1
Segui-Simarro JM Austin JR White EA Staehelin LA Electron tomographic analysis of somatic cell plate formation in meristematic cells of Arabidopsis preserved by high-pressure freezing. Plant Cell 2004 16 836 856 15020749 10.1105/tpc.017749
Marsh BJ Mastronarde DN Buttle KF Howell KE McIntosh JR Organellar relationships in the Golgi region of the pancreatic beta cell line, HIT-T15, visualized by high resolution electron tomography. PNAS 2001 98 2399 2406 11226251 10.1073/pnas.051631998
Mogelsvang S Gomez-Ospina N Soderholm J Glick BS Staehelin LA Tomographic evidence for continuous turnover of Golgi cisternae in Pichia pastoris. Mol Biol Cell 2003 14 2277 2291 12808029 10.1091/mbc.E02-10-0697
Verbelen JP Tao W Mobile arrays of vacuole ripples are common in plant cells. Plant Cell Rep 1998 17 917 920 10.1007/s002990050509
Binzel ML Hasegawa PM Handa AK Bressan RA Adaptation of tobacco cells to NaCl. Plant Physiol 1985 79 118 125 16664356
Shope JC DeWald DB Mott KA Changes in surface area of intact guard cells are correlated with membrane internalization. Plant Physiol 2003 133 1314 1321 14551331 10.1104/pp.103.027698
Johnson-Flanagan AM Singh J Membrane deletion during plasmolysis in hardened and non-hardened plant cells. Plant Cell Environ 1986 9 299 306
Staiger CJ Yuan M Valenta R Shaw PJ Warn RM Lloyed CW Microinjected profilin affects cytoplasmic streaming in plant cells by rapidly depolymerizing actin microfilaments. Curr Biol 1994 4 215 219 7922326 10.1016/S0960-9822(00)00050-6
Kost B Spielhofer P Chua NH A GFP-mouse talin fusion protein labels plant actin filaments in vivo and visualizes the actin cytoskeleton in growing pollen tubes. Plant J 1998 16 393 401 9881160 10.1046/j.1365-313x.1998.00304.x
Hoffmann A Nebenführ A Dynamic rearrangements of transvacuolar strands in BY-2 cells imply a role of myosin in remodeling the plant actin cytoskeleton. Protoplasma 2004 224 201 210 15614481 10.1007/s00709-004-0068-0
Reisen D Etude de l'expression d'un gène codant une aquaporine tonoplastique de chou-fleur chez le tabac. Analyse de la dynamique vacuolaire et de l'effet de stress osmotiques. PhD Thesis 2003 University of Burgundy, Dijon, France
Kotzer AM Brandizzi F Neumann U Paris N Moore I Hawes C AtRabF2b (Ara7) acts on the vacuolar trafficking pathway in tobacco leaf epidermal cells. J Cell Sci 2004 117 6377 6389 15561767 10.1242/jcs.01564
Saito C Ueda T Abe H Wada Y Kuroiwa T Hisada A Furuya M Nakano A A complex and mobile structure forms a distinct subregion within the continuous vacuolar membrane in young cotyledons of Arabidopsis. Plant J 2002 29 245 255 11844103 10.1046/j.0960-7412.2001.01189.x
Avila EL Zouhar J Agee AE Carter DG Chary SN Raikhel NV Tools to study plant organelle biogenesis. Point mutation lines with disrupted vacuoles and high-speed confocal screening of green fluorescent protein-tagged organelles. Plant Physiol 2003 133 1673 1676 14681530 10.1104/pp.103.033092
Escobar NM Haupt S Thow G Boevink P Chapman S Oparka K High-throughput viral expression of cDNA-green fluorescent protein fusions reveals novel subcellular addresses and identifies unique proteins that interact with plasmodesmata. Plant Cell 2003 15 1507 1523 12837943 10.1105/tpc.013284
Uemura T Yoshimura SH Takeyasu K Sato MH Vacuolar membrane dynamics revealed by GFP-AtVam3 fusion protein. Genes Cells 2002 7 743 753 12081650 10.1046/j.1365-2443.2002.00550.x
Klich MG Didone NG Fernandez OA Mujica MB Ultrastructural changes in Spirodela intermedia in response to osmotically-induced water shortage. Biocell 2000 24 85 88 10979606
Leborgne N Dupou-Cezanne L Teulières C Canut H Tocanne JF Boudet AM Lateral and rotational mobilities of lipids in specific cellular membranes of Eucalyptus gunnii cultivars exhibiting different freezing tolerance. Plant Physiol 1992 100 705 715
Makarenko SP Konenkina TA Salyaev RK Characteristics of fatty acid composition of lipids in higher plant vacuolar membranes. Membr Cell Biol 2000 13 687 695 10987391
Reuveni M Bennett AB Bressan RA Hasegawa PM Enhanced H+ transport capacity and ATP hydrolysis activity of the tonoplast H+-ATPase after NaCl adaptation. Plant Physiol 1990 94 524 530 16667744
Murashige T Skoog F A revised medium for rapid growth and bioassays with tobacco tissue cultures. Physiol Plant 1962 15 473 497
|
16080795
|
PMC1190192
|
CC BY
|
2021-01-04 16:37:14
|
no
|
BMC Plant Biol. 2005 Aug 4; 5:13
|
utf-8
|
BMC Plant Biol
| 2,005 |
10.1186/1471-2229-5-13
|
oa_comm
|
==== Front
BMC PsychiatryBMC Psychiatry1471-244XBioMed Central London 1471-244X-5-271593264910.1186/1471-244X-5-27Research ArticleAlterations of prolyl endopeptidase activity in the plasma of children with autistic spectrum disorders Momeni Naghi [email protected]öm Berit M [email protected] Vibeke [email protected] Hassan [email protected] Bengt V [email protected] Department of Health Sciences, Autism Research, Lund University, Lund, Sweden2 Department of Neurology, Golestan University of Medical Science, Gorgan, Iran2005 2 6 2005 5 27 27 27 8 2004 2 6 2005 Copyright © 2005 Momeni et al; licensee BioMed Central Ltd.2005Momeni et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Prolyl Endopeptidase (PEP, EC 3.4.21.26), a cytosolic endopeptidase, hydrolyses peptide bonds on the carboxyl side of proline residue in proteins with a relatively small molecular weight. It has been shown that altered PEP activity is associated with various psychological diseases such as schizophrenia, mania and depression. Autistic Spectrum Disorders (ASD) are neuropsychiatric and behavioural syndromes affecting social behaviours and communication development. They are classified as developmental disorders. The aim of this study was to examine the hypothesis that PEP activity is also associated with ASDs.
Methods
Fluorometric assay was used to measure PEP activity in EDTA plasma in children with ASD (n = 18) aged 4–12 years (mean ± SD: 7.9 ± 2.5). These results were then compared to PEP activity in a control group of non-ASD children (n = 15) aged 2–10 years (mean ± SD: 6.4 ± 2.2).
Results
An alteration in PEP activity was found in the children with ASD compared to the control group. There was much greater variation of PEP activity in the group of ASD children when compared to the controls (SD= 39.9 and SD 9.6, respectively). This variation was significant (p < 0.0005), although the mean level of PEP activity in the group of ASD children was slightly higher than in the control group (124.4 and 134.1, respectively).
Conclusion
Our preliminary finding suggests a role for PEP enzyme in the pathophysiology of autism but further research should be conducted to establish its role in the aetiology of psychiatric and neurological disorders, including autism and related spectrum disorders.
==== Body
Background
Prolyl Endopeptidase (PEP, EC 3.4.21.26) is a cytosolic endopeptidase. PEP cleaves peptide bonds on the carboxyl side of proline residues in low molecular weight proteins containing the recognition sequence X-Pro-Y, where X is a peptide or protected amino acid and Y is an amide, a peptide, an amino acid, an aromatic amine or an alcohol [1]. PEP can only hydrolyse small peptides and is thought to be involved in the metabolism of hormones/neuropeptides. However, PEP also degrades many active hormones/neuropeptides, e.g. oxytocin, arginine vasopressin (AVP), substance P, neurotensin, luteinizing hormone-releasing hormone (LH-RH) and thyrotropin-releasing hormone (TRH) [2]. These low molecular weight proteins, particularly oxytocin, AVP, TRH, neurotensin, and substance P, profoundly affect social behaviour, emotions, stress level, responsivity, reward-seeking and positive reinforcement behaviour [3]. Altered PEP activity has been observed in psychiatric disorders such as depression, mania and schizophrenia [4].
Autistic Spectrum Disorders (ASD) are neuropsychiatric and behavioural syndromes affecting social and communicative development. They were classified as developmental disorders in DSM-IV (American Psychiatric Association 1994) [5]. Severe communication deficits and social and behavioural abnormalities often appear during the first three years of life but the diagnosis is often made later, due to a lack of resources. The aetiology of ASD is not yet known. Symptoms of ASDs are related to the abnormal functioning of certain centres within the brain: in particular the cerebellum, brain stem and limbic region [6]. ASDs are also associated with several specific dysfunctions including fragile X syndrome [7], a cascade of complex gene-environment interactions [8,9], hyperserotoninemia, [10,11], increased levels of opioid [12,13] and high levels of arginine-vasopressin (AVP) [our unpublished observation]. Low plasma levels of the neuropeptide hormone oxytocin have also been found in a group of children with ASD when compared to the normal age-matched controls [14].
Altered levels of the neuropeptide hormones oxytocin, arginine vasopressin and other related hormones/peptides may be a result of proteolytic enzyme activity such as PEP, which is involved in the formation and degradation of various neuropeptides. Our hypothesis is that altered activity of proteolytic enzymes, such as PEP, in children with ASD (children <12 years) might lead to the degradation of some specific neuropeptide hormones, affecting social behaviour and communication.
Methods
Materials
N-Benzyloxycarbonyl-glycyl-prolyl-4-methylcoumarinyl-7-amide (Z-Gly-Pro-4-methylcoumarinyl-7-amide) was obtained from Bachem in Bubendorf, Switzerland. Dithiothereitol, ethylenediamineteraacetic acid disodium salt dihydrate (EDTA), benzamidinium chloride, p-chloromercuribenzoate (PCMB) and pepstatin, sodium azide, dithiothereitol (DTT) were obtained from Sigma in St. Louis, Missouri USA. Acetic acid and 1,4-dioxan were obtained from Merk in Darmstadt, Germany. A Perkin Elmer Fluorimeter LS 50B was used to determine the release of 7-amino-4-methylcoumarine at excitation and emission wavelengths of 370 nM and 440 nM.
Subjects
Eighteen ASD and a control group of 15 non-ASD children participated in this study. The children with ASD were selected from children attending rehabilitation centres in Sweden. The original diagnosis of ASD was made jointly by a psychiatrist and psychologist who made the diagnosis in accordance with DSM IV (APA, 2000) and the International Classification of Diseases (ICD) (WHO, 1993). Their diagnosis was then independently confirmed by the specialist in autism spectrum disorders at Lund university hospital. As a routine measure, these children undergo health checks including a dental examination. This requires the children to be anaesthetised, since many children with ASD are unable to understand what is required of them and incapable of cooperating when a dental examination is carried out. During our research period 18 ASD children underwent this examination. Wechsler Intelligence Scale for Children (WISC) was used to estimate the children's functional abilities. The ASD group consisted of 14 boys and four girls ranging from 4 to 12 years (mean 7.9 years; SD, 2.5). Information about additional dysfunction and medication was not available. Paediatricians at the children's hospital selected the children in the control group when they came to the hospital to be treated for various physical conditions. None of the children in the control group had any mental disabilities. The control group (non-ASD) consisted of nine boys and seven girls ranging from 2–10 years (mean 6.4 years; SD, 2.2), (Table 1).
Table 1 PEP activity (fluorescence intensity unit) of EDTA plasma and age in control (n = 15) and ASD groups (n= 18).
Study groups Percentiles % Fluorescence intensity Child's age
intesity
Control group
Mean 124.4 6.4
Std. Deviation 9.6 2.2
Minimum 105.4 2
Maximum 144.0 10
Percentiles 25 115.8 5
50 125.6 7
75 130.7 8
ASD group
Mean 134.1 7.9
Std. Deviation 39.9 2.5
Minimum 48.1 4
Maximum 201 12
Percentiles 25 109.3 6
50 147.6 8
75 160.2 10
The Ethics Committee at the Faculty of Medicine, Lund University, approved this study (LU-70-00).
Sample collection
Venous blood from ASD children was collected under general anaesthesia when they were undergoing another medical treatment. This was done in the presence of a child psychiatrist with special training in the field of childhood psychosis. Venous blood from a control group of non-ASD children was collected in evacuated 4 mL EDTA tubes, containing 0.084 ml of 0.34 M K3-EDTA solution. These tubes (Vacutainer System) were obtained from Becton-Dickinson Inc., Plymouth, UK. Plasma from EDTA-containing blood was produced immediately after collection by centrifugation at 1300 g for 10 min at 4°C. 30 μL of cocktail inhibitors per 1 mL plasma was then added to the produced plasma sample. The inhibitor cocktail stock solution used was Tris 2.0 M, Na-EDTA 0.9 M, Benzamidin 0.2 M, E-64,10 μM and Pepstatin 48 μM. The PEP activity of the samples was analysed immediately after the production of plasma. The remaining samples were stored for further investigation at -70°C.
Assay procedure
The method used to assay the PEP using the hydrolysis of the fluorogenic substrate (Z-Gly-Pro-4-methylcoumarinyl-7-amide) has previously been described by Momeni et. al. [15]. This study showed that different factors such as temperature, freeze-thawing cycles, substrate concentration, the organic solvent used to dissolve the substrate and the time of incubation of enzyme-substrate mixture influenced the final fluorescence intensity. 20 μL of plasma was incubated with 200 μL of buffer (100 mM phosphate buffer, pH 7.5, with 1 mM EDTA, 1 mM DTT and 1 mM sodium azide) for 10 min at 37°C to reach thermal equilibrium. 5 μL of the substrate solution containing 18.4 mM Z-Gly-Pro-4-methylcoumarinyl-7-amide was then added and the mixture incubated at 37°C for 120 min. The reaction was then terminated by the addition of 1000 μL of 1.5 M acetic acid and the release of 7-amino-4-methylcoumarin measured in a fluorimeter (λex: 370 nm; λem: 440 nm; slit width: 2.5). The substrate solution was prepared by dissolving Z-Gly-Pro-4-methylcoumarinyl-7-amide in 100% 1,4-dioxane and then diluting to 50% (v/v) with incubation buffer. All measurements were carried out in triplicate.
The flourometric assay originally used by Goossens et al [16] was incapable of detecting low PEP activity in CSF. By further developing the procedure of previous work [15] it was possible to achieve a 400% improvement in assay sensitivity and detect PEP in CSF [15]. Triple assays of each sample were carried out. Any variation in results was insignificant, which confirmed the reliability of the procedure. The average for each of the three results was used for the calculations.
Statistical analysis
The figure for the fluorescence intensity and the children's age (Mean +/- statistical deviation) in the two groups were calculated and plotted on a graph. Mann-Whitney U-test was used to test the difference of mean for PEP activity. Levene's test for equality of variances was used for the comparison between the ASD and non-ASD groups of children. SPSS version 11.2 (Norusis, M.J./SPPS Inc., 2004) was used.
Results
Basal plasma PEP activities in the control group (n = 15) were between 105.4 and 144.0 fluorescence intensity units (mean 124.4, median 125.6). The activity of PEP in the 18 ASD children ranged from 48.1 to 201.9 fluorescence intensity units (mean 134.1, median 147.6) (Table 1).
The mean level of PEP activity in children with ASD was only slightly higher than that in the controls but the variation of PEP activity was much larger in ASD children than in the controls (SD = 39.9 and 9.6, respectively). The difference was significant (Levene's test for equality of variances yielded F (17,14) = 16.4, p < 0.0005) (Fig. 1). The ASD children were 4–12 years old (mean ± SD: 7.9 ± 2.5), the control group were 2–10 years old (mean ± SD: 6.4 ± 2.2). The variation of enzyme activity is shown in Fig. 2.
Figure 1 Variation of PEP activity in EDTA plasma in the control group (n = 14), interquartile range (115.8–130.7), and in the ASD group (n = 18), interquartile range (109.3–160.2).
Figure 2 Variation of plasma PEP activity in the children with ASD (n = 18) and in a non-ASD control group (n = 14) associated to age.
Discussion
Proline endopeptidase, a cytosolic enzyme isolated from human tissues, cleaves different low-molecular-weight neuropeptide hormones such as oxytocin, AVP, TRH, neurotensin, bradykinin and substance P. The neuropeptide hormones, which contain a proline in the carboxyl side of their sequences, act as a substrate for PEP. It has been reported that PEP activity is altered in individuals with depression, mania and schizophrenia [4]. High PEP serum activity has also been reported in patients with PTSD (post-traumatic stress disorder) [17]. The result of this study showed a significantly higher variance of the PEP activity in the group of ASD children. There may be various explanations for this finding, including the heterogeneity of individuals in the ASD group and the effects of pharmaceuticals [4]. The general anaesthesia (GA) may also have an impact on PEP activity. The precise effect of GA on enzyme activity is currently unknown. This question will be addressed in the next study.
This alteration of PEP activity may support our hypothesis that PEP might be involved in the aetiology of ASD. However, our working hypothesis is that ASD can be caused or influenced by external events in early childhood, possibly as result of a genetic predisposition. As a result there may be an inappropriate release of the cytosolic proteolytic enzyme PEP into the circulating blood stream and the cerebrospinal fluid (CSF). PEP cleaves different neuropeptides or their precursor molecules leading to an alteration of the concentration of neuropeptides and this may have a negative effect upon proper brain function.
The main deficits of children with ASD include early difficulties with social contact, such as eye contact and social smile [18,19], attention [20], affects [21,22], reciprocity [18,23], turn-taking, timing and answering parents' signals [21]. Many different problems can arise with respect to co-ordination and motor planning [18,24], body tonus deficits [19], and problems with mobility. Children with ASD have a tendency to ignore other people or may even prefer to be alone [19]. They also have difficulties in signalling for attention and they communicate without meeting the gaze of another person [18]. This could be explained by the effects PEP might have on the neuropeptide hormones when cleaving them and interfering with their proper functions in the processes of early brain development.
In three children with ASD, the PEP activity was lower than the mean activity of the control group. Twelve ASD children had higher PEP activity than average and the remaining three had PEP activity equal to that of the control group. This variation may be related to different psychiatric disorder from which the patients were suffering, such as depression or mania. There was no significant gender difference in enzyme activity in the control group, neither was there any great variation but in the ASD group there was a significant variation, randomly distributed between the sexes.
We did not measure PEP activity in the cerebrospinal fluid (CSF) of the ASD children. It might differ from plasma PEP activity. In a previously published article [15] PEP activities was measured in both CSF and EDTA plasma in patients suffering from another neurological disease in order to investigate whether PEP activity differs in these two substances from the same patient. The result showed a variation in PEP activity in plasma compared to CSF. This variation might be caused by a dysfunction in the blood brain barrier (BBB) that might allow substances such as PEP to pass through the barrier and continue its activity in the circulating blood. BBB dysfunction might also be implicated in the variation of the EDTA plasma PEP activity in the children with ASD.
Due to the pioneering nature of this research it is difficult to relate the finding of this study to other biomedical research on ASD. Research is also limited regarding PEP activity associated with other psychiatric disorders. Altered prolyl endopeptidase activity in plasma has been associated with major depressed patients (low levels) and with manic and schizophrenic patients (high levels) [4]. This preliminary finding may indicate an association between altered PEP activity and neuro-psychiatric disorders such as ASDs.
Conclusion
Our preliminary finding suggests a role for PEP enzyme in the pathophysiology of autism but further research should be conducted to establish its role in the aetiology of psychiatric and neurological disorders, including autism and related spectrum disorders.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
NM planned and performed all experiments presented in this study. HA participated in the design of the study. BN and VH analysed data and participated in the preparation of the manuscript. BS, the corresponding author, is the academic supervisor of NM. BS supervised all aspect of the statistical analysis and the writing of the manuscript. BSs particular interest is the relationship between the biomedical aspects and the autistic spectrum disorders.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors wish to thank Dr. Vivianne Nordin, Lund University, for collecting samples. We also wish to thank the Department of Clinical Chemistry, Lund University, for the use of their laboratory resources.
==== Refs
Walter R Simmons WH Yoshimoto T Proline specific endo-and exopeptidases Mol Cell Biochem 1980 30 111 127 6991912 10.1007/BF00227927
Welches WR Brosnihan KB Ferrario CM A comparison of the properties and enzymatic activities of three angiotensin processing enzymes: angiotensin converting enzyme, prolyl endopeptidase and neutral endopeptidase 24.11 Life Science 1993 52 1461 1480 10.1016/0024-3205(93)90108-F
Insel TR Oxytocin : a neuropeptide for affiliation: evidence from behavioral, receptor autoradiographic and comparative studies Psychoneuroendocrin 1992 17 33 35
Maes M Goossens F Scharpé S Calabrese J Desnyder R Meltzer HY Alteration in plasma prolyl endopeptidase activity in depression, mania, and schizophrenia: effect of antidepressants, mood stabilizers, and antipsychotic drugs Psychiatry Res 1995 58 217 225 8570777 10.1016/0165-1781(95)02698-V
American Psychiatric Association DSM-IV: Diagnostic and statistical Manual of Mental Disorders 1994 4 Washington, D.C.: American Psychiatric Press
Bauman MN Microscopic neuroanatomic abnormalities in autism Pediat 1991 87 791 796
Gillberg C Steffenburg S Wahlström J Gillberg IC Sjöstedt A Martinsson T Liedgren S Olofsson OE Autism associated with marker chromosome J Am Acad Child Adolesc Psychiatry 1991 30 489 494 2055888
London EA The environment as an etiologic factor in autism: a new direction for research Environ Health Perspect 2000 401 404 10852835
Robinson PD Schutz CK Macciardi F White BN Holden JJ Genetically determined low maternal serum dopamine beta-hydroxylase levels and the etiology of autism spectrum disorders Am J Med Genet 2001 100 30 36 11337745 10.1002/ajmg.1187
Singh VK Singh EA Warren RP Hyperserotoninemia and serotonin receptor antibodies in children with autism but not mental retardation Biol Psychiatry 1997 41 753 755 9067002 10.1016/S0006-3223(96)00522-7
Cook EH Leventhal BL The serotonin system in autism Curr Opin Pediat 1996 8 348 354
Scifo R Cioni M Nicolosi A Batticane N Tirolo N Quattropani MC Morale MC Gallo F Marchetti B Opioid-immune interactions in autism: behavioural and immunological assessment during a double-blind treatment with naltrexone Ann Ist Super Sanita 1996 32 351 359 9028057
Gillberg C Terenius L Lonnerholm G Endorphin activity in childhood psychosis Arch Gen Psychiatry 1985 42 780 783 4015322
Modahl C Green AL Fein D Morris M Waterhouse L Feinstein C Levin H Plasma Oxytocin Levels in Autistic Children Biol Psychiatry 1998 43 270 277 9513736 10.1016/S0006-3223(97)00439-3
Momeni N Yashimoto T Grubb A Factors influencing analysis of prolyl endopeptidase in human blood and cerebrospinal fluid: increase in assay sensitivity Scand J Clin Lab Invest 2003 63 387 395 14594319 10.1080/00365510310001951
Goossens F Meester ID Vanhoof G Scharpé S A sensitive method for the assay of serum prolyl endopeptidase Eur J Clin Chem Clin Biochem 1992 30 235 238 1525255
Maes M Lin AH Bonaccoroso S Goossens F Gastel AV Pioli R Delmeire L Scharpé S Higher serum prolylendopeptidase activity in patients with post-traumatic stress disorder J Affect Disord 1999 53 27 34 10363664 10.1016/S0165-0327(98)00086-X
Trewarthen C Aitken K Papoudi D Robarths J Children with autism. Diagnosis and interventions to meet their needs 1998 2 Jessica Kingsley Publishers. London
Adrien JL Lenoir P Martineau J Perrot A Hameury L Larmande C Savage D Blind rating of Early Symptoms of Autism Based upon family Home Movies J Am Acad Child and Adolesc Psychiatry 1993 32 617 626 7684363
Maestro S Muratori F Cavallo MC Pei F Stern D Golse B Palacio-Espasa F Attentional skills during the first 6 months of age in Autism Spectrum Disorders J Am Acad Child Adolesc Psychiatry 2002 41 1239 1245 12364846 10.1097/00004583-200210000-00014
Werner E Dawson G Osterling J Dinno N Recognition of autism spectrum disorders before one year of age: A retrospective study based on home videotapes J Autism Dev Disord 2000 30 157 162 10832780 10.1023/A:1005463707029
Baranek GT Autism during infancy: A retrospective video analysis of senso-motor and social behaviors at 9–12 months of age J Autism Dev Disord 1999 29 213 224 10425584 10.1023/A:1023080005650
Mars AE Mauk JE Dowrick PW Symptoms of pervasive developmental disorders as observed in toddlers in prediagnostic home videos of infants and toddlers J Pediat 1998 132 500 504 9544908
Teitelbaum P Teitelbaum O Nye J Fryman J Maurer RG Movement analysis in infants may be useful for early diagnosis of autism Proc Natl Acad Sci U S A 1998 95 13982 13987 9811912 10.1073/pnas.95.23.13982
|
15932649
|
PMC1190193
|
CC BY
|
2021-01-04 16:33:03
|
no
|
BMC Psychiatry. 2005 Jun 2; 5:27
|
utf-8
|
BMC Psychiatry
| 2,005 |
10.1186/1471-244X-5-27
|
oa_comm
|
==== Front
BMC Struct BiolBMC Structural Biology1472-6807BioMed Central London 1472-6807-5-111598515410.1186/1472-6807-5-11Research ArticleStructural organization and interactions of transmembrane domains in tetraspanin proteins Kovalenko Oleg V [email protected] Douglas G [email protected] William F [email protected] Martin E [email protected] Department of Cancer Immunology and AIDS, Dana-Farber Cancer Institute and Department of Pathology, Harvard Medical School, Boston, USA2 Department of Biochemistry and Biophysics, School of Medicine, University of Pennsylvania, Philadelphia, USA3 Dana-Farber Cancer Institute, D-1430, 44 Binney Street, Boston, MA 02115, USA2005 28 6 2005 5 11 11 29 3 2005 28 6 2005 Copyright © 2005 Kovalenko et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Proteins of the tetraspanin family contain four transmembrane domains (TM1-4) linked by two extracellular loops and a short intracellular loop, and have short intracellular N- and C-termini. While structure and function analysis of the larger extracellular loop has been performed, the organization and role of transmembrane domains have not been systematically assessed.
Results
Among 28 human tetraspanin proteins, the TM1-3 sequences display a distinct heptad repeat motif (abcdefg)n. In TM1, position a is occupied by structurally conserved bulky residues and position d contains highly conserved Asn and Gly residues. In TM2, position a is occupied by conserved small residues (Gly/Ala/Thr), and position d has a conserved Gly and two bulky aliphatic residues. In TM3, three a positions of the heptad repeat are filled by two leucines and a glutamate/glutamine residue, and two d positions are occupied by either Phe/Tyr or Val/Ile/Leu residues. No heptad motif is apparent in TM4 sequences. Mutations of conserved glycines in human CD9 (Gly25 and Gly32 in TM1; Gly67 and Gly74 in TM2) caused aggregation of mutant proteins inside the cell. Modeling of the TM1-TM2 interface in CD9, using a novel algorithm, predicts tight packing of conserved bulky residues against conserved Gly residues along the two helices. The homodimeric interface of CD9 was mapped, by disulfide cross-linking of single-cysteine mutants, to the vicinity of residues Leu14 and Phe17 in TM1 (positions g and c) and Gly77, Gly80 and Ala81 in TM2 (positions d, g and a, respectively). Mutations of a and d residues in both TM1 and TM2 (Gly25, Gly32, Gly67 and Gly74), involved in intramolecular TM1-TM2 interaction, also strongly diminished intermolecular interaction, as assessed by cross-linking of Cys80.
Conclusion
Our results suggest that tetraspanin intra- and intermolecular interactions are mediated by conserved residues in adjacent, but distinct regions of TM1 and TM2. A key structural element that defines TM1-TM2 interaction in tetraspanins is the specific packing of bulky residues against small residues.
==== Body
Background
Tetraspanins constitute a large family of integral membrane proteins, characteristically containing 4, 6 or 8 conserved cysteine residues in the large extracellular loop (including the CCG and PxxCC motifs), which form disulfide bonds, and several conserved polar residues in the intracellular loop and transmembrane regions [1,2]. There are 32 putative tetraspanin family members in mammals, 37 in Drosophila melanogaster and 20 in Caenorhabditis elegans. Tetraspanins play diverse roles in cell adhesion, migration and fusion processes, cellular activation and signaling (reviewed in refs. [2-4]). Mammalian tetraspanins such as CD9, CD63, CD81, CD82, CD151, rds/peripherin, and uroplakins Ia and Ib have been most extensively studied, with mouse knock-out models available for CD9 [5-7], CD81 [8,9], CD151 [10] and a few others. However, the majority of tetraspanins are characterized very little, if at all, at genetic, biochemical or structural levels.
The large extracellular loop (LEL) of tetraspanins has received most attention, since it contains functionally important sites. Sequence QRD (194–196) in CD151 is important for association with integrins, which has functional consequences for integrin-dependent cell spreading and multicellular cable formation [11]. A site in the LEL of CD9, SFQ (residues 173–175), is essential for CD9 function in sperm-egg fusion [12]. The crystal structure of tetraspanin CD81 LEL revealed five α-helixes, A-E [13]. Helices A, B and E form a relatively conserved region in tetraspanins, whereas the region between helices B and E is the most variable [14]. Interestingly, the variable region contains most of the functionally important sites involved in tetraspanin protein-protein interactions.
A remarkable biochemical property of tetraspanin molecules is their ability to associate with a large number of other transmembrane proteins, including integrins, membrane-associated growth factors and receptors, MHC class II molecules, Ig superfamily proteins, and each other [2,3,15]. Several of these lateral associations of tetraspanins are detected in "mild" detergents (Brij series, CHAPS), but are disrupted by "strong" detergents such as Triton X-100 or SDS. Multiprotein complexes of tetraspanins and associated molecules, also called the "tetraspanin web" [16], may represent a distinct tetraspanin-enriched membrane microdomain [17,18]. The formation of this microdomain is influenced by palmitoylation of several conserved juxtamembrane cysteine residues in tetraspanins [19-21].
The transmembrane domains, encompassing nearly half of a tetraspanin protein, are the most conserved part of the molecule (Stipp et al. [1] and this study). However, very little functional information is available on these domains. The differential detergent sensitivity of tetraspanin-tetraspanin associations suggests that hydrophobic interactions between TM helices may play a role. Indeed, when the large extracellular loop (LEL) of CD151 is deleted, the molecule is still able to associate with other tetraspanins [22]. Thus, TM domains are strong candidates for mediating tetraspanin-tetraspanin interactions.
The importance of TM domain interactions in intramolecular organization was demonstrated in a study showing that CD82 fragment TM2-4, lacking TM1, was retained in the endoplasmic reticulum, but was transported to the cell surface upon co-expression of TM1 [23]. This in vivo reconstitution experiment demonstrated a strong interaction between TM1 and the rest of the molecule. Expression of a truncated CD9 molecule (TM3-LEL-TM4) results in intracellular accumulation of the protein and significant misfolding of the LEL, as judged by inappropriate disulfide formation and diminished antibody reactivity (our unpublished data). Similarly, a CD9 epitope in the LEL is lost in molecules lacking either TM2+TM3 or just TM4 [24]. Thus, TM domain interactions and packing are crucial for proper folding, stability and transport of tetraspanin molecules.
In a previous study, we showed that covalent cross-linking of membrane-proximal cysteine residues can be used as a tool for detection of tetraspanin-tetraspanin associations [25]. Inhibition of cysteine palmitoylation by 2-bromopalmitate (2-BP) made cysteines available for cross-linking and enabled demonstration of specific tetraspanin homodimerization and low levels of heterodimerization. We concluded that tetraspanin homodimers, formed in the Golgi, may be a fundamental structural unit within tetraspanin microdomains.
In this study, we carried out detailed sequence analysis of human tetraspanin TM domains. We show that a heptad repeat containing conserved glycine, asparagine and large hydrophobic residues occurs in TM1 and TM2 domains, and predict tight intramolecular association of these two domains by packing of the large residues against the small residues. Moreover, by using cysteine cross-linking we map a dimerization interface in the human CD9 protein, and show that conserved heptad motif glycine residues are also important for intermolecular CD9 associations.
Results
Sequence analysis of tetraspanin transmembrane domains: presence of the heptad repeat motif
We focused our attention on 28 human tetraspanins identified from the SWISS-PROT and GenBank databases. All tetraspanins have in common four hydrophobic stretches (TM domains) of 20–25 residues, and contain highly conserved residues in the second extracellular loop, in particular the Cys-Cys-Gly (CCG) motif. Detailed analysis of the large extracellular loop sequences [14], and dendrograms based on full-length alignment can be found in earlier studies [26,27]. The length of each transmembrane domain was established based on previous sequence analysis of tetraspanin sequences [27,28], and on annotations to the database entries. Manual adjustments based on sequence homology and hydrophobicity profiles were done to fully delineate the TM domains. The resulting lengths of TM domains were: TM1 – 23 residues; TM2 – 21 residues; TM3 – 25 residues; TM4 – 25 residues. Two more residues could be added onto the N-terminal part of TM2; however, relatively small sequence conservation of these residues among tetraspanins and occurrence of polar/charged side chains in some tetraspanins precluded us from doing so for the global alignment.
Figures 1 and 2 show a multiple sequence alignment of four TM domains of 28 human tetraspanins. For each position within the domains, consensus residues were determined and classified (with individual color code) in 4 categories: 1) large hydrophobic residues (including Val, Met, Leu, Ile, Phe, Tyr, Trp), 2) small residues (Gly, Ala, Ser and Thr), 3) Cys, and 4) Asn. When more than two types of residues occupied a given position in a TM, a dual-color pattern that reflected the prevalence of the particular residue type was used (Figure 1). Cysteine residues were shown separately due to their importance as palmitoylation target sites. The highly conserved asparagine residue in TM1 was considered separately. No proline residues are found in TM domains 1–3 of human tetraspanins.
An inspection of the multiple sequence alignment reveals a repeating heptad amino acid pattern, (abcdefg)n, in TM1, 2 and 3 (Figure 1, 2). Heptad repeats promote helical coiled coil interactions in multiple soluble and membrane-spanning proteins [29-31]. In the heptad repeat, hydrophobic residues in positions a and d are of special importance, as they directly mediate interhelical contacts by creating a tight knobs-into-holes packing in the coiled coil structure [32]. For instance, in the leucine zipper of the yeast transcription factor GCN4, positions a and d contain Val and Leu residues, respectively, with an Asn residue in a single a position forming a hydrogen bond across the GCN4 dimer interface [33].
In TM1 of tetraspanins, highly conserved Asn, Gly and Gly residues (numbers 18, 25 and 32 in the CD9 sequence) appear at d positions of the heptad repeats, and residues 14, 21 and 28 are at a positions (Figure 1). In TM2, residues 67, 74 and 81 (consensus Gly, Gly and Ala, respectively) occupy a positions, whereas residues 63, 70 and 77 are at d positions. Another highly conserved glycine, Gly80, occupies the 3rd g position in TM2. In TM3, the conserved pattern consists of two leucine residues (Leu89 and Leu96) and a glutamate/glutamine residue (Glu/Gln103) in a positions (Figure 2). Two d positions are also conserved – Phe/Tyr92 and Ile/Val/Leu99. TM4 lacks a conserved heptad pattern and has only a single conserved position, Glu/Gln209 (with four exceptions). These features of TM1-4 of tetraspanins are displayed on helical wheel diagrams (Figure 3).
Analysis of TM1 sequences
The conserved Asn-Gly-Gly motif, occupying designated d positions of the heptad repeat, is the most prominent structural feature of TM1. We also compared sequences of CD9 orthologs from 10 different organisms (the most available for any tetraspanin) to gain further insight into conservation and variability of the TM1 sequence. As shown in Figure 4, positions a, d and g in TM1 are among the most conserved (0, 1 and 1 substitution, respectively), while interspecies variability tends to occur in other positions: b (5 substitutions), c (4 substitutions), e (4 substitutions) and f (4 substitutions). Thus, the positions typically involved in coiled coil interactions (a and d) are the most conserved.
When residues of TM1 are plotted as a helical wheel, additional structural features are revealed (Figure 3). There are highly conserved aliphatic and aromatic residues in the first three a positions of the heptad motif (Phe15, Trp22 and Leu29 in CD9), as well as in g positions (Leu14, Phe21, Val28 in CD9). The "ridges" formed by these bulky residues are flanking the "groove"-forming Gly residues of the Asn-Gly-Gly position d motif. In contrast, b, c, e and f positions show an overall higher variability among tetraspanins, as also seen in the comparison of CD9 orthologs described above.
Analysis of TM2 sequences
A landmark feature of TM2 in tetraspanins is the presence of highly conserved glycine residues (Gly67, 74, 77 and 80 in CD9, Figure 1). Other substitutions at these positions are almost exclusively small residues, such as Ala or Ser. In addition, Ala, Ser or Thr occupy position 81. This residue, together with Gly67 and Gly74, forms face a of the helix. Residue Gly77 (position d) is preceded by conserved, chiefly large hydrophobic residues on the same helical face (Leu63 and Met70 in CD9). Extremely conserved Gly80 falls into heptad position g (Figure 3). Among CD9 orthologs, heptad positions a and d are absolutely conserved, whereas other positions have the following number of substitutions: b – 3; c – 2; e – 1; f – 3; g – 1 (Figure 4). Two of the f position residues in TM2 (65 and 79) also show higher variability among different tetraspanins (Figures 1, 3). Cysteine residues are frequently found near the cytoplasmic end of TM2 helix at positions 78 and 79; these cysteines are likely to be palmitoylated.
Analysis of TM3 and TM4 sequences
The TM3 domain provides another example of the heptad repeat pattern. Position a is occupied by two highly conserved leucine and a glutamate/glutamine residue (Leu89, Leu96 and Glu/Gln103 in CD9). Furthermore, two d positions are conserved – Phe/Tyr92 (aromatic residue) and Ile/Val99 (β-branched aliphatic residue; Figures 2, 3). In addition, residue 100 in position e is generally Phe or Leu. Among CD9 orthologs, position a has 1 substitution, positions b, c and f each have 6, positions d and e each have 2, and g has 4. Thus, as for TM1 and TM2, positions a and d are among the most conserved, but overall TM3 has more variable positions than TM1 or TM2 (Figure 3). Less than half of TM3 sequences contain cysteine residues, and those tend to occur at the internal positions of the helix (Figure 2).
TM4 shows less conservation among various tetraspanin family members than the other TM domains (Figures 2, 3). The only highly conserved feature is the glutamate/glutamine residue in position 209. In addition, one or two cysteine residues can be found at the C-terminal end of TM4 in some tetraspanins (e.g. CD9, CD81, CD151), and many sequences contain additional polar residues (Arg, Lys, His, Asn, Gln). No conserved heptad motif was identified in TM4, as also confirmed by analysis of substitutions in CD9 orthologs (data not shown).
Mutational analysis of conserved glycine residues in TM1 and TM2
The conserved nature of the Asn and Gly residues in TM1 and TM2 prompted an analysis of their functional role. To this end, we have probed whether mutations of these residues destabilize the protein molecule. We expressed a construct of the first and second TMs of CD9, connected by the small extracellular loop, and tagged with a C-terminal green fluorescent protein (TM(1+2)-GFP molecule). In human rhabdomyosarcoma RD cells, the wild-type fusion protein localized mostly in a reticular, intracellular pattern, without forming any large aggregates (Figure 5, panel A). Remarkably, when double mutants Gly25Leu + Gly32Leu and Gly67Leu + Gly74Leu were expressed, the protein formed distinct large aggregates in a high proportion of cells (Figure 5, panels C and E). In contrast, double mutant Gly77Leu + Gly80Leu did not form such aggregates (Figure 5, panel G). Results with respective single mutants were similar to that with double mutants, with the aggregation being somewhat more pronounced for Leu substitutions of Gly67 and Gly74 compared to Gly25 and Gly32 mutations. No aggregation was observed for Asn18Ser and Asn18Tyr mutants (data not shown). Also, nearly identical results were obtained with human HT1080 cells (data not shown).
We interpret these results as an indication that aggregating mutants are destabilized or misfolded while non-aggregating mutants retain the wild-type conformation. Intriguingly, mutations to the conserved GG7 motifs caused protein aggregation while the mutation of other glycines had no detectable effect. These results also suggest that wild-type GFP, which has weak tendency to self-associate, could enhance non-specific interactions of destabilized mutant TM(1+2) CD9 moieties, leading to their aggregation. Consistent with this hypothesis, the aggregation of Gly25Leu + Gly32Leu and Gly67Leu + Gly74Leu double mutants was suppressed when monomeric GFP molecule, Leu221Lys [34] was used (Figure 5, panels D and F). The use of monomeric GFP did not affect intercellular localization of wild-type CD9 TM(1+2) (Figure 5, panel B), or a Gly77Leu + Gly80Leu double mutant (Figure 5, panel H).
In summary, Leu substitutions of Gly residues that are part of the Asn-Gly-Gly (NGG7) motif in TM1, or Gly-Gly-Ala (GGA7) motif in TM2, resulted in destabilization and aggregation of GFP-fused TM(1+2) proteins, whereas substitutions of Gly77 or Gly80, which are not part of these motifs (Figure 3), failed to show such aggregation.
Prediction and modelling of interaction between TM1 and TM2
Consecutive helices in polytopic membrane proteins frequently interact [35]. Sequence analysis of TM1 and TM2 helices of tetraspanins reveals a remarkable complementarity in the distribution of large and small residues at heptad positions a and d along the helical axis (Figure 3), suggesting that these residues may interact. To further elucidate the potential for TM1-TM2 interaction, the putative interface was modeled using a novel algorithm that considers mutational data during each step of a Monte Carlo simulated annealing cycle (see Methods for details). Specifically, Gly25Leu, Gly32Leu, Gly67Leu and Gly74Leu were scored as disruptive mutations, while Asn18Ser, Gly77Leu and Gly80Leu were scored as silent mutations, based on their effects on protein stability (Figure 5 and data not shown).
The resulting model predicts left-handed crossing of TM1 and TM2 helices at an angle of +28°. The key element of the structure is the apposition of bulky and small heptad position a and d residues, as follows: Gly32-Leu63; Gly67-Leu29; Gly25-Met70; Gly74-Trp22; Asn18-Gly77; Ala81-Phe15 (Figure 6). Our model predicts that each of these residue pairs are in van der Waals contact. Additionally, two potential H-bonds are predicted in this model, indicating close packing: Gly67 Cα to Gly25 carbonyl oxygen, and Trp22 Cα to Met70 carbonyl oxygen. The packing is tighter in the ectodomain-proximal portion of the helices (Figure 6, panel B), as determined by Cα-Cα distances between interacting residue pairs.
The key elements of the model are corroborated by the presence of apparently complementary substitutions in TM1 and TM2 sequences of different tetraspanins (Figure 1, boxed residues). For example, Gly74 is predicted to interact with Trp22. In 8 of the 10 tetraspanins that contain a substitution for Gly74, a compensatory substitution occurs at the Trp22 position (Figure 1). Thus, a larger non-glycine side chain at position 74 may necessitate a less bulky non-Trp side chain in position 22. Likewise, the presence of a Cβ at position 25, typically occupied by glycine, necessitates a non-β-branched amino acid at position 70, which is occupied by a β-branched residue in nearly half of all cases. Indeed, we find that in each of 5 cases in which position 25 contains a Cβ, a leucine residue occurs in position 70. This analysis is consistent with our molecular model that suggests Leu70 will pack most favorably against a Cβ at position 25 than a β-branched residue or a methionine.
Role of TM1 and TM2 heptad motif residues in CD9 dimerization
To probe CD9 dimerization, we used a cysteine-mediated cross-linking approach. We established previously a simple and efficient method for cysteine-mediated cross-linking [25]. After cells are pre-treated with 2-BP for 16–24 hours to expose normally palmitoylated cysteines, the cysteines can be cross-linked using any of the following methods: a) Spontaneous oxidation in Brij97 lysates (a condition that preserves tetraspanin-tetraspanin associations), b) In situ cross-linking, by pre-lysis oxidation of cells with Cu2+-phenanthroline (CuP) to promote disulfide bond formation. c) In situ cross-linking with thiol-reactive cross-linking agents of defined length (e.g. DTME, BMB). The first two approaches produce in essence "zero-length" disulfides, indicative of close proximity of target cysteines and presumably high specificity of interaction. In contrast, chemical cross-linkers with 6–20 Å spacer arm may cross-link with higher efficiency, but not necessarily higher specificity. However, they provide advantages such as variable membrane permeability, and potential linkage cleavability. For tetraspanins such as CD9, membrane-permeable cross-linker DTME (13.3 Å-long, reducible) provides highly specific and efficient cross-linking [25]. Here we have used a cysteine cross-linking strategy, in combination with cysteine-scanning mutagenesis, to map the residues from TM1 and TM2 contributing to the CD9 dimerization interface.
For subsequent cross-linking experiments using CD9 TM(1+2)-GFP protein, the non-dimerizing form of GFP was used. This avoids potential GFP-dependent dimerization and aggregation that can be observed with wild-type GFP, especially when fusions with transmembrane proteins are studied [36]. Importantly, the Leu221Lys mutation in GFP prevented aggregation of mutant forms of CD9 TM(1+2), which was observed with wild-type GFP fusion (Figure 5). The TM(1+2) fragment of CD9 contains three native cysteines – Cys9, Cys78 and Cys79. Single-cysteine mutants of TM(1+2) were constructed, in which a cysteine was placed at various faces of TM1 or TM2 while all of the wild-type cysteines were simultaneously replaced by serines. The mutant proteins were transiently expressed in RD cells (having little endogenous CD9), which were then treated for 16–18 hours with 2-BP. To achieve maximal specificity in cross-linking we used a "zero-length" agent, CuP.
First, single-cysteine replacements were constructed for residues Leu14, Phe15, Gly16, Phe17 and Asn18, covering just over one complete helical turn at the beginning of TM1. While residue Asn18 is highly conserved, positions 14, 15 and 17 are occupied by bulky hydrophobic residues in most tetraspanins, whereas position 16 shows less conservation (Figures 1, 4). All of the single-cysteine mutants showed diffused pattern of protein localization, without any signs of aggregation. As shown in Figure 7A, the highest level of intermolecular cross-linking was observed for Leu14Cys and Phe17Cys mutants, a lower level for Phe15Cys and Gly16Cys mutants, and very little cross-linking for Asn18Cys substitution. These results indicate that: a) the first two transmembrane domains of CD9 alone can mediate its dimerization, and b) the g and c residues of TM1 (e.g. Leu14 and Phe17, Figure 3) are likely to be part of the intermolecular interface.
Similarly, single-cysteine substitutions were made for residues Gly77, Gly80 and Ala81 in TM2; in addition, proteins carrying a single wild-type cysteine, Cys9, Cys78 or Cys79, were tested. No protein aggregation was observed for any of these single-cysteine mutants. As shown in Figure 7B, the relatively low level of intermolecular cross-linking of wild-type CD9 TM(1+2)-GFP protein was enhanced dramatically in single-cysteine TM2 mutants Gly80Cys and Ala81Cys. The Gly77Cys mutant also had an elevated level of cross-linking. In contrast, any of the three native cysteines (9, 78 and 79) produced level of cross-linking not much greater than the wild-type TM(1+2) protein. Similar results were obtained with cysteine-reactive cross-linker BMB (data not shown). Likewise, comparable results were obtained with single-cysteine mutants of untagged, full-length CD9, using CuP (Figure 7C) as well as DTME cross-linker (data not shown).
These cross-linking results for TM1 and TM2 are consistent with our model that places residues Leu14, Phe17 and Gly80 on the same side of the TM1-TM2 pair (Figure 6, panel C). The strong cross-linking with Leu14Cys, Phe17Cys and Gly80Cys places the intermolecular interface toward the c and g phases of the TM1 helix, and the g phase of the TM2 helix, away from its e and f faces containing wild-type cysteines 78 and 79.
Critical residues at the TM1-TM2 interface also affect dimerization indirectly. To assess specific CD9 dimerization, we used a Gly80Cys substitution at the intermolecular interface for cross-linking. As shown in Figure 8A, single replacements of conserved heptad residues in positions 18, 25, 32, 67 and 74 (Asn18Ser, Gly25/32/67/74→Leu) strongly decreased the cross-linking mediated by Cys80. The effect was most pronounced for mutations of residues, Gly32 and Gly67, located in the tightly packed extracellular end of TM helices (Figure 6). In contrast, mutations of residues closer to the cytoplasmic end of TM2 (Gly74 and especially Ala81) had only modest to very little effect on cross-linking.
Relatively low efficiency of intermolecular cross-linking via native residues Cys9, 78, and 79 (Figures 7B,C) correlates well with the predicted location of Cys78 and 79 away from the dimeric interface (Figure 3), and suggests that the extramembrane N-terminal part of CD9 (residues 1–13) does not self-associate. We next examined whether mutations of conserved Asn and Gly residues in TM1 and TM2 decreased low-level background cross-linking via native cysteines. As expected, these mutations had virtually no effect on dimer formation of CD9 TM(1+2)-GFP (Figure 8B). The level of covalent dimer formed was not diminished for triple Asn18Ser + Gly25Leu + Gly32Leu and double Gly67Leu + Gly74Leu mutants, compared to wild-type TM(1+2) CD9 molecule. Similarly, the same triple and double mutations in the context of full-length CD9-GFP protein (with six cysteines) produced wild-type levels of cross-linking (Figure 8C). We interpret these findings as evidence for at least two types of associations between CD9 molecules: primary, involving residues 14, 17 and 80, and dependent on integrity of conserved heptad residues in TM1 and TM2, and less efficient secondary interactions, probably representing random collision events, and independent of the heptad residues (see Discussion for more details).
TM3 and TM4 cysteine residues in CD9 dimerization
After identifying the roles of conserved TM1 and TM2 residues in CD9 dimerization, we next probed whether residues proximal to TM domains 3 and 4 are also involved. To this end, disulfide cross-linking of full-length CD9 molecules containing 3 C-terminal cysteines (87, just before TM3; 218 and 219 in TM4) or 3 N-terminal cysteines (9, 78 and 79) was compared (Figure 9). We found that the C-terminal cysteines were only slightly better than N-terminal cysteines with respect to detection of CD9 dimers. However, markedly more trimers and tetramers were detected using C-terminal cysteines. Thus, residues 87, 218 and 219 at TM3 and TM4 in CD9 can together form contacts across the dimeric interface and also additional contacts with other neighboring CD9 molecules.
Discussion
Here we provide the first detailed analysis of tetraspanin protein transmembrane domains. First, we show 1) the presence of a heptad repeat motif in TM1 and TM2, containing highly conserved Asn and Gly residues, 2) a leucine and glutamate/glutamine-containing heptad motif in TM3, and 3) high variability and absence of heptad repeats in TM4 sequences. Second, we provide evidence for a specific, intramolecular interaction between TM1 and TM2 domains, in which bulky hydrophobic residues pack against GG7 motif, and present a molecular model for this interaction. Third, experimental mapping of the CD9 dimerization interface firmly establishes an additional role for conserved TM1 and TM2 residues in dimeric intermolecular interactions. Fourth, preliminary evidence is provided to suggest that TM3 and TM4 domains contribute to expansion of CD9 dimers into higher order multimers.
Conserved residues in TM1 and TM2 of tetraspanins: role in intramolecular packing
We hypothesized that the first two transmembrane domains of tetraspanins might interact with each other because: a) consecutive TM domains frequently associate in known protein 3D structures [35], and b) they both contain a series of highly conserved amino acids – several Gly residues and an Asn residue (Figure 1). Conserved Gly residues are a frequent theme in the organization of interacting transmembrane domains. Analysis of 3D helix packing in polytopic membrane proteins reveals that Gly residues tend to localize in buried positions, especially at the helix-helix interfaces and helix crossing points [37,38]. Due to the absence of a side chain, Gly provides a flat surface for packing of a side chain from another residue, without loss of side-chain entropy upon interaction. The most common Gly-containing motif is GxxxG [39,40]. In glycophorin A (GpA), the major glycoprotein in erythrocyte cell membranes, Gly79 and Gly83 are part of the LIxxGVxxGVxxT sequence that promotes homodimerization of parallel transmembrane α-helixes [41,42]. In the GpA dimerization motif, Gly residues allow for tight packing in the right-handed helical crossing [43]. There are also examples of left-handed helical crossing in the context of a GxxxG motif [44]. Other membrane proteins that use the GxxxG motif for homo- or heterodimerization include bacteriophage M13 coat proteins [45], yeast alpha factor receptor [46], integrin α IIb subunit [47], and ErbB1 receptor tyrosine kinase [48]. Other small residues, such as Ala and Ser, can substitute for Gly in this motif [49].
A protein motif in which Gly residues are separated by 6 other residues (GG7) is also common in transmembrane helices, especially in transporter/channel-like membrane proteins [50]. However, the structural features associated with this motif are not well known. In particular, it is unclear whether left-handed GG7 heptad repeat motif (as opposed to the "classic" right-handed GxxxG motif) can drive membrane helix association. In a recent work addressing this issue, Lear et al. [51] showed that a synthetic peptide containing Gly at heptad positions a and d could self-associate in vitro, likely in an antiparallel orientation. Heptad repeats containing conserved Gly residues occur in TM domains of α and β chains of MHC class II proteins, and mutations of the Gly residues disrupt the αβ heterodimer [52]. These examples demonstrate that Gly-based heptad motifs may be used for both intra- and intermolecular associations.
In this work, we identified a highly conserved GG7 motif in the first two tetraspanin TM domains. The GG7 sequence in tetraspanins is a part of a larger motif that also includes a conserved Asn residue in TM1 (NGG7) and an Ala/Ser/Thr residue in TM2 (GGA7). The seven-residue periodicity of these motifs strongly suggests their involvement in left-handed coiled coil packing reminiscent of the leucine zipper, rather than right-handed packing of the GpA-like GxxxG motif. For antiparallel helices, the left-handed crossing is in fact predominant over the right-handed in known TM domain structures [44].
In our model, heptad Gly residues in NGG7 and GGA7 sequences provide specific packing between antiparallel tetraspanin TM1 and TM2 helices by allowing tight van der Waals interactions with large hydrophobic residues (Figure 6). Highly efficient packing of bulky side chains against glycine residues is observed in known transmembrane protein 3D structures [38,53,54]. An example includes packing of helices M1 and M2 in potassium channel KcsA, where Val91 in M2 is paired with Gly43 in M1, and Leu36 in M1 contacts Ala98 and Gly99 in helix M2 [54,55]. In addition to facilitating helix-helix packing, Gly residues frequently provide additional CαH...O hydrogen bonds between two helices [44]. In our model, two Cα-backbone carbonyl H-bonds are predicted – between residues Gly27-Gly67, and Trp22-Met70.
Although polar and charged amino acid residues (such as Asn in the TM1 heptad motif) are infrequent in transmembrane domains, they are functionally important. Polar residues such as glutamine, glutamic acid, aspartic acid and asparagine can promote strong oligomerization of model membrane-associated helices [56-58]. Ruan et al. [59] used asparagine scanning mutagenesis to probe the interface of self-associating polyleucine helices by detecting their enhanced self-interaction in vitro and in the E. coli-based ToxR assay. Thus, a hydrogen bond in an apolar environment can result in strong, though not necessarily specific, association of transmembrane helices. In fact, mutations to polar residues in transmembrane proteins are commonly associated with disease [60]. Because of this potential for non-specific interactions, polar residues tend to localize at buried positions in TM domains.
In our case, the conserved Asn18 residue in CD9 is predicted to be a part of the TM1-TM2 interface, though our model does not predict any electrostatic interaction between Asn18 and TM2 (Figure 6). Consistently, substitution such as Asn18Tyr (and Gly77Leu) in TM(1+2)-GFP protein was not destabilizing as analyzed by protein aggregation. Curiously, the full-length Asn18Ser CD9 migrated slightly slower on SDS-PAGE gel (data not shown), suggesting that Asn18 does play a role in maintaining conformation of the molecule. The Asn18Cys single-cysteine mutant shows very little intermolecular cross-linking (Figure 7A), supporting the proposed location of this residue at the intramolecular interface. It is tempting to speculate that the "pocket" between TM1 and TM2 lined by Asn18 and Gly77 might be important for accommodating palmitate moieties that target Cys78 and Cys79 residues, and/or important for access by the putative palmitoyl transferase to those residues. Understanding the exact role of these highly conserved Asn18 and Gly77 residues in tetraspanins awaits further investigation.
In summary, we identified conserved glycine residues of TM1 and TM2 of tetraspanins as key elements required for intramolecular packing. Mutations of these key residues (Gly25, Gly32, Gly67 and Gly74 in CD9) resulted in protein destabilization and aggregation. There is ample evidence in the literature for mutations in transmembrane proteins that lead to protein destabilization, misassembly and pathologic conditions [61]. Thus, we have identified conserved heptad Gly residues in TM1 and TM2 of tetraspanins as plausible targets of destabilizing mutations with potential functional consequences.
Intermolecular interactions in tetraspanins
Tetraspanin CD9 forms mostly homodimers, but also a low level of heterodimers with CD81 and CD151 [25]. Thus, mapping the dimerization interface is an important next step in structure-function analysis of tetraspanins. Disulfide-mediated cross-linking, often in combination with cysteine-scanning mutagenesis, is a common strategy to probe oligomerization or intersubunit interactions of transmembrane proteins such as histidine kinase EnvZ [62], M(3) muscarinic acetylcholine receptor [63], E. coli lactose permease [64], synaptobrevin [65], integrins [66] and many others. In tetraspanins such as CD9, membrane-proximal cysteine residues are especially useful targets for disulfide trapping, as their linkage can be enhanced by pre-treating cells with 2-BP. While the ability of wild-type cysteines in CD9 to be cross-linked may indicate that they are close to the dimerization interface, more precise mapping was achieved here using cysteine-scanning mutagenesis.
Our data clearly identify regions, near the cytoplasmic face of TM1 and TM2, important for dimerization. Intermolecular zero-length cross-linking was highest when single cysteines were placed in positions 14, 17, 77, 80 and 81 in TM(1+2)-GFP molecule, or at positions 77, 80 or 81 in the full-length CD9 protein. Positions 14, 17 and 80 are distinct from the intramolecular interface and are on the same side of the TM1-TM2 pair (Figure 6). Thus, they are well-positioned to participate in an interaction with another molecule. At the same time, the model predicts that wild-type cysteines (Cys78, 79), which do not yield very efficient zero-length cross-linking, are on the other side of TM1-TM2 pair.
While using the cysteine at position 80 as the dimeric interface probe, mutations of conserved residues in TM1 and TM2 (especially Gly32 and Gly67 to Leu) clearly reduced intermolecular cross-linking. We do not suggest that those residues are directly involved in intermolecular interaction. Rather, we propose that destabilization of the intramolecular TM1-TM2 interaction by Gly to Leu substitutions (discussed above) causes an overall conformational change that reduces dimer formation.
An Ala81Leu mutation did not reduce cross-linking via Cys80, even though single-cysteine Ala81Cys molecules themselves produced a high level of cross-linking. These results, together with data on Gly32Leu and Gly67Leu mutations, are consistent with our model predicting that helices 1 and 2 interact more tightly near the extracellular end and less at the cytoplasmic end. This would give more flexibility to a cysteine at position 81 and also limit the effect of an Ala81Leu mutation. Location of this residue at the membrane/cytoplasmic border could also make it more accessible to CuP reagent as compared to residues buried in TM domain, thus elevating the efficiency of disulfide formation of the Ala81Cys mutant.
Multiple interfaces in tetraspanin molecules
In the full-length CD9 molecules, the 3 C-terminal cysteines (Cys87, 218 and 219) located at or in TM3 and TM4 promoted efficient dimer and even more efficient oligomer formation compared to the 3 N-terminal cysteines (Figure 9). Cys87 alone can be used to capture CD9 dimers [25]. These results suggest the existence of two dimeric interfaces in CD9 molecule – the TM 1-2/1-2 interface and the TM 3-4/3-4 interface (Figure 10). In a TM(1+2) molecule, the destabilization of 1–2 interaction, e.g. by Gly→Leu mutations, would affect the 1-2/1-2 interface, as discussed above. However, these mutations would not interfere with the 3-4/3-4 interface in a full-length molecule, which includes Cys87, 218 and 219. Thus, cross-linking of full-length molecules, containing all 6 cysteine residues, would be unaffected, as seen in Figure 8C. Furthermore, wild-type Cys9, Cys78 and Cys79 are apparently not at the primary 1-2/1-2 interface. Their relative inefficiency in cross-linking CD9 TM(1+2) protein likely reflects weak secondary contacts between the molecules, or possibly random collision events. Such events should be independent of mutations in the conserved Gly residues in TM1 and TM2, as was demonstrated in Figure 8B. The potential existence of two interfaces in tetraspanin molecules, 1-2/1-2 and 3-4/3-4, should provide enhanced flexibility for forming additional intermolecular contacts. Current understanding of tetraspanin microdomains assumes a few strong, primary homotypic and heterotypic tetraspanin complexes (e.g. CD9-CD9, CD9-CD81, CD151-α3 integrin, CD81-EWI2) that help bring together various other proteins, forming secondary-type associations. Such properties of tetraspanins may bring signaling molecules such as protein kinase C or phosphatidylinositol 4-kinase to the vicinity of integrins [67,68].
The organization of the TM3 domain points to a potential role in protein-protein interactions. A motif of Leu-Leu-Glu(Gln) spaced 7 residues apart (heptad positions a), with highly conserved residues in two consecutive positions d, poses as a likely interaction module. If responsible for heterologous protein-protein interactions, it would form another distinct interface of tetraspanin molecule. Our preliminary data indicate that replacing the Leu and Glu residues in TM3 of CD9 with Ala has no effect on cell surface expression of the protein and its dimerization (data not shown). It remains to be tested if interactions with other proteins will be affected. Similarly, the TM4 domain may provide additional contributions to lateral tetraspanin associations. Much higher sequence variability, and the lack of a distinct heptad pattern suggests that TM4 is a major contributor to diversity among tetraspanin complexes. Structure-function analysis of TM3 and TM4 domains in tetraspanins is the subject of ongoing investigation.
Conclusion
We have defined the TM1-TM2 intramolecular interface in tetraspanin CD9, providing evidence for glycines (Gly25 and Gly32 in TM1, Gly67 and Gly74 in TM2) packing against apposing bulky aliphatic residues. Second, we mapped an intermolecular CD9 interface (involved in CD9 homodimer formation) to the vicinity of residues Leu14 and Phe17 in TM1 and Gly77, Gly80 and Ala81 in TM2. Finally, we provide preliminary evidence that TM3 and TM4 in CD9 may contribute to a second intermolecular interface. Key CD9 residues involved in intra- and intermolecular interactions are highly conserved throughout the tetraspanin family, thus suggesting that our findings will apply to most tetraspanins.
Methods
Materials
Cell culture reagents were from Invitrogen (Carlsbad, CA). 2-bromopalmitate was from Fluka (Milwaukee, WI), N-ethylmaleimide (NEM) and 1,10-phenanthroline were from Sigma-Aldrich (St. Louis, MO), and chemical cross-linkers dithio-bis-maleimidoethane (DTME) and 1,4-Bis-maleimidobutane (BMB) were purchased from Pierce Endogen (Rockford, IL). Triton X-100, protease inhibitor cocktail and FuGENE 6 transfection reagent were obtained from Roche (Indianapolis, IN). Restriction endonucleases and Pfu DNA polymerase were obtained New England Biolabs (Beverly, MA) and Stratagene (Carlsbad, CA), respectively. All other chemicals were purchased from Sigma-Aldrich or Fisher Scientific (Pittsburg, PA).
Sequence analysis
Tetraspanin sequences were obtained from SWISS-PROT and GenBank databases. Locus designations, accession numbers and the most commonly used protein names are summarized in Tables 1 and 2. TM segments were delineated by inspection of hydrophobicity profiles, using database annotations and previous analyses of TM sequences as a guide ([28], M. Hemler, unpublished), and aligned manually. Residue numbers in human CD9 sequence are used a reference point throughout the study.
DNA cloning and mutagenesis
Sequence encoding CD9 protein was cloned into vector pcDNA3 (Invitrogen, Carlsbad, CA) and pEGFP-N1 (Clontech, Palo Alto, CA), for expression of untagged and C-terminally GFP-tagged CD9, respectively. pEGFP-N1 encoding CD9 TM(1+2) -GFP fusion protein was constructed by subcloning DNA for residues 1–83 of CD9 into HindIII and PstI sites of the vector; to introduce the PstI site, codon GTG for Val82 was changed to CTG (coding for Ala). In the resulting fusion protein, there is a 13-amino acid linker (with no cysteines) between CD9 and GFP. To minimize the low inherent ability of GFP to homodimerize, which could potentially influence the results of CD9 cross-linking, we used a monomeric GFP mutant, Leu221→Lys [34], for cross-linking experiments.
Mutations were introduced in full-length and TM(1+2) CD9 proteins by a PCR-based strategy using mutagenic primers and Pfu DNA polymerase. All mutations were confirmed by DNA sequencing.
Protein expression, microscopy, cysteine disulfide cross-linking and Western blotting
DNA constructs encoding TM(1+2)-GFP or full-length CD9 proteins were transfected into human rhabdomyosarcoma RD cells using the FuGENE 6 reagent. Cells expressing GFP fusion proteins were analysed by fluorescence microscopy 18–28 hours post-transfection. Images were captured using Spot 1.4.0 camera (Diagnostic Instruments, Sterling Heights, MI) attached to Nikon Eclipse TE300 microscope.
For experiments involving cysteine-mediated cross-linking, cells were treated with 50 μM 2-BP starting 24–26 hours post-transfection and continuing for 16–18 hours. Cross-linking was carried out by incubating cells in HBSM buffer (25 mM Hepes-NaOH, pH 7.2, 150 mM NaCl, 2 mM MgCl2) containing either a) 0.6 mM CuSO4 and 1.8 mM 1,10-phenanthroline (CuP complex) or b) 0.2 mg/ml homobifunctional cysteine-reactive cross-linker (e.g. DTME), diluted from fresh 10 mg/ml solution in DMSO. After incubation for 10–15 minutes (with CuP) or 30–45 minutes (with cross-linker), cells were washed twice for 10 minutes with HBSM containing 10 mM NEM to block residual free cysteines. Cells were lysed in HBSM containing 1% Triton X-100, 0.1% SDS and a cocktail of protease inhibitors with 1 mM EDTA at 4°C for 45–60 minutes. Cell lysate was clarified by centrifugation at 14,000 × g for 15 minutes, an aliquot was removed, and proteins from it were precipitated by addition of trichloroacetic acid to 10% on ice followed by centrifugation at 14,000 × g for 10 minutes. After two washes with ice-cold acetone, protein pellet was solubilized in SDS-PAGE sample buffer without a reducing agent (50 mM Tris-HCl, pH 6.8, 1% SDS, 8% glycerol).
In some experiments, CD9 protein was immunoprecipitated using monoclonal antibody Alb6 (Immunotech, Marseille, France). Proteins were separated by SDS-PAGE and analyzed by Western blotting using monoclonal antibody JL-8 (Clontech) for GFP or Alb6 for CD9. Bands from X-ray films were quantitated using GeneTools™ software from Syngene (Frederick, MD).
Modeling of TM1-TM2 interaction
An atomic model of the CD9 TM1-TM2 dimer was constructed with a Monte Carlo-simulated annealing (MCSA) algorithm [69]. Two idealized α-helices corresponding to TM1 residues Tyr12 through Leu35 and TM2 residues Gly59 through Val82 were docked with six orthogonal parameters: three rigid body translations and three rotations. During each step of a MCSA cycle, there was an equal probability of changing either one parameter or all six parameters to random values. A conformation's energy was calculated in vacuo with the AMBER united-atom force field for van der Waals interactions [70]. The van der Waals term was modified as described by Kuhlman and Baker [71]. If a structure had favorable dimerization energy, the energies of select mutants were calculated. Structures were selected with a novel scoring function that maximizes the Boltzmann probability of dimerization for silent mutations while minimizing the probability for disruptive mutations. Asn18Ser, Gly77Leu, and Gly80Leu were scored as silent mutations while Gly25Leu, Gly32Leu, Gly67Leu, and Gly74Leu were considered to be disruptive. Each MCSA cycle consisted of 50,000 steps with an exponential temperature decay from 10,000 to 10 K.
Ten MCSA cycles through global sample space were used to restrict the search area. Parameters were restricted to ± 2 standard deviations from their mean value for structures within 10 kcal of the best structure. MCSA cycles were repeated as described above with additional optimization of χ values: rotamers at the protein-protein interface were optimized with Dead End Elimination [72], and χ values were further optimized with Monte Carlo. All MCSA cycles converged upon structures that were within a root mean squared deviation (RMSD) of 1.5 Å with the best structure, and structures that scored within 5 kcal of the best score had an RMSD of less than 0.5 Å with the best structure.
List of abbreviations
2-BP, 2-bromopalmitate; CuP, Cu2+-phenanthroline; DTME, dithio-bis-maleimidoethane; LEL, large extracellular loop; NEM, N-ethylmaleimide; TM, transmembrane (domain).
Authors' contributions
OVK carried out sequence comparisons, mutational analysis and cross-linking experiments, and drafted the manuscript. DGM built the TM1-TM2 interaction model and contributed to the manuscript. WFD supervised DGM's work. MEH coordinated the whole study and prepared the final manuscript.
Acknowledgements
We thank the rest of the Hemler laboratory for helpful discussions. This work was supported by the NIH grant GM38903 to MEH.
Figures and Tables
Figure 1 Sequence alignment of the transmembrane domains 1 and 2 of 28 human tetraspanins. Residues from select positions of the heptad motifs in TM1 and 2 are highlighted (see text for details). Also highlighted are polar residues and cysteines. Consensus residue types are shown by the color scheme indicated. Boxed residues reflect correlated substitutions for position pairs 22–74 and 25–70 (details are in the text). The numbers refer to CD9 sequence.
Figure 2 Sequence alignment of the transmembrane domains 3 and 4 of 28 human tetraspanins. Residues from heptad positions a and d in TM3 are highlighted. Also highlighted are the conserved Glu/Gln residue in TM4, other polar residues and cysteines. The color scheme is as in Figure 1.
Figure 3 Helical wheel diagrams of transmembrane domains TM1-4 reflecting the consensus residue types. The color scheme is as in Figure 1. The numbers refer to CD9 sequence. Heptad positions a through g are indicated for TM1-3. A predicted interaction between positions a and e in TM1 and a and d in TM2 is shown by dotted lines (see Figure 6 and text for details). Arrows reflect the efficiency of intermolecular cross-linking via single cysteines placed in these positions (see Figure 7 and text for details).
Figure 4 Sequence alignment of TM1-3 for ten vertebrate orthologs of CD9. Heptad positions a and d in TM1, TM2 and TM3 are highlighted in green. Residues that differ between orthologs are shown in yellow.
Figure 5 Expression of wild-type and mutant CD9 TM(1+2)-GFP proteins in human cells. Human rhabdomyosarcoma RD cells were transfected with constructs encoding CD9 TM(1+2)-GFP fusion proteins that carried mutations indicated. Either wild-type or monomeric (L221K) GFP was used. Images were captures 18–28 hours post-transfection.
Figure 6 Structural model of TM1-TM2 interaction in CD9. Shown are space-filling models of CD9 TM1 (panel A), TM2 (panel B), and the two helices together (panel C). Small and large residues of heptad positions a and d, which form the crucial contacts between the helices, are shown in green and red, respectively. Asn18 is shown in blue, Leu14 and Phe17 in yellow, and Gly80 in light green.
Figure 7 Cross-linking of single-cysteine mutants of CD9 protein. GFP fusions of CD9-TM(1+2) protein (panels A and B) or untagged full-length CD9 (panel C), and containing wild-type cysteines (Cys9,78,79), a single cysteine at positions indicated, or no cysteines (all replaced by Ser) were transiently expressed in human RD cells and cross-linked using CuP. Cell lysates were analyzed by Western blotting for GFP (panels A and B) or CD9 (panel C). % dimer is the fraction of dimer in total protein material (monomer + dimer), based on densitometry measurements of respective bands.
Figure 8 Effect of mutations in conserved TM1 and TM2 residues on CD9 cross-linking. RD cells were transfected with constructs encoding the following fusion proteins: panel A, CD9 TM(1+2)-GFP, either wild-type or single-cysteine G80C mutant, which also carries TM1 and TM2 substitutions indicated; panel B, CD9 TM(1+2)-GFP with no cysteines or wild-type cysteines (Cys9,78,79) plus TM1 and TM2 mutations; panel C, full-length CD9-GFP with TM1 and TM2 mutations indicated. The proteins were cross-linked and analyzed by GFP Western as in Figure 7. % dimer was calculated as for Figure 7.
Figure 9 CD9 cross-linking through N- or C-terminal cysteine residues. RD cells were transfected with constructs encoding wild-type CD9 protein, CD9 containing only the three N-terminal cysteines (Cys9, 78 and 79), or the three C-terminal cysteines (Cys87, 218 and 219). Cells were treated with 2-BP for 16 hours and lysed in buffer containing 1% Brij97 for spontaneous disulfide cross-linking. Cell lysates were analyzed by CD9 Western. Relative intensity of bands corresponding to CD9 monomers, dimers, trimers and tetramers was calculated as a percent fraction of total protein material (monomers through tetramers).
Figure 10 A model of the organization of a tetraspanin dimer. Two interfaces, the intramolecular TM1/TM2 and intermolecular TM1-TM2/TM1-TM2, have been analysed in this work. A third interface, intermolecular TM3-TM4/TM3-TM4, is predicted.
Table 1 Human tetraspanin sequences analyzed in this study.
SWISS-PROT locus name Accession number Protein name(s)
TSN1_HUMAN O60635 NET-1, Tspan-1
CD82_HUMAN P27701 CD82 antigen NM_030927
CD37_HUMAN P11049 CD37 antigen
CD53_HUMAN P19397 CD53 antigen
T4S7_HUMAN O14817 NAG-2, Tspan-4
TNE5_HUMAN O75954 NET-5
BAB55318 AK027715 -
TSN2_HUMAN O60636 Tspan-2
T4S9_HUMAN P62079 NET-4, Tspan-5
TM4B_HUMAN Q9UKR8 TM4B
- NM_030927 -
C151_HUMAN P48509 CD151 antigen, PETA-3
CD9_HUMAN P21926 CD9 antigen, p24, MRP-1
CD81_HUMAN P60033 CD81 antigen, TAPA-1
T4S3_HUMAN P19075 CO-029 antigen
T4S8_HUMAN O60637 Tspan-3, TM4-A
T412_HUMAN O95859 NET-2
SAS_HUMAN Q12999 Sarcoma amplified sequence (SAS)
T413_HUMAN O95857 NET-6
TNE7_HUMAN O95858 NET-7
CD63_HUMAN P08962 CD63 antigen, LAMP-3
T4S2_HUMAN P41732 A15, TALLA-1, CD231 antigen
T4S6_HUMAN O43657 Tspan-6, TM4-D
OCSP_HUMAN Q9H1Z9 Oculospanin (Ocsp)
UPKA_HUMAN O00322 Uroplakin Ia (UPIa)
UPKB_HUMAN O75841 Uroplakin Ib (UPIb)
RDS_HUMAN P23942 Peripherin, Retinal degeneration slow protein (RDS)
ROM1_HUMAN Q03395 Rod outer segment membrane protein-1 (ROM-1)
Table 2 Vertebrate CD9 orthologs analyzed in this study.
Organism SWISS-PROT or GenBank locus name Accession number
Homo sapiens (Human) CD9_HUMAN P21926
Cercopithecus aethiops (African green monkey) CD9_CERAE P30409
Sus scrofa (Pig) CD9_PIG Q8WMQ3
Bos taurus (Cow) CD9_BOVIN P30932
Felis catus (Cat) CD9_FELCA P40239
Rattus norvegicus (Norway rat) CD9_RAT P40241
Mus musculus (House mouse) CD9_MOUSE P40240
Gallus gallus (Chicken) NP_990093 (GenBank) NP_990093
Petromyzon marinus (Sea lamprey) AAN64299 (GenBank) AAN64299
Danio rerio (Zebrafish) AAH59691 (GenBank) AAH59691
==== Refs
Stipp CS Kolesnikova TV Hemler ME Functional domains in tetraspanin proteins Trends Biochem Sci 2003 28 106 112 12575999 10.1016/S0968-0004(02)00014-2
Hemler ME Tetraspanin proteins mediate cellular penetration, invasion and fusion events, and define a novel type of membrane microdomain Ann Rev Cell Dev Biol 2003 19 397 422 14570575 10.1146/annurev.cellbio.19.111301.153609
Berditchevski F Complexes of tetraspanins with integrins: more than meets the eye J Cell Sci 2001 114 4143 4151 11739647
Tarrant JM Robb L van Spriel AB Wright MD Tetraspanins: molecular organisers of the leukocyte surface Trends Immunol 2003 24 610 617 14596886 10.1016/j.it.2003.09.011
Le Naour F Rubinstein E Jasmin C Prenant M Boucheix C Severely reduced female fertility in CD9-deficient mice Science 2000 287 319 321 10634790 10.1126/science.287.5451.319
Kaji K Oda S Shikano T Ohnuki T Uematsu Y Sakagami J Tada N Miyazaki S Kudo A The gamete fusion process is defective in eggs of CD9-deficient mice Nature Genet 2000 24 279 282 10700183 10.1038/73502
Miyado K Yamada G Yamada S Hasuma H Nakamura Y Ryu F Suzuki K Kosai K Inoue K Ogura A Okabe M Mekada E Requirement of CD9 on the egg plasma membrane for fertilization Science 2000 287 321 324 10634791 10.1126/science.287.5451.321
Maecker HT Levy S Normal lymphocyte development but delayed humoral immune response in CD81-null mice J Exp Med 1997 185 1505 1510 9126932 10.1084/jem.185.8.1505
Miyazaki T Muller U Campbell KS Normal development but differentially altered proliferative responses of lymphocytes in mice lacking CD81 EMBO J 1997 16 4217 4225 9250665 10.1093/emboj/16.14.4217
Wright MD Geary SM Fitter S Moseley GW Lau LM Sheng KC Apostolopoulos V Stanley EG Jackson DE Ashman LK Characterization of mice lacking the tetraspanin superfamily member CD151 Mol Cell Biol 2004 24 5978 5988 15199151 10.1128/MCB.24.13.5978-5988.2004
Kazarov AR Yang X Stipp CS Sehgal B Hemler ME An extracellular site on tetraspanin CD151 determines α 3 and α 6 integrin-dependent cellular morphology J Cell Biol 2002 158 1299 1309 12356873 10.1083/jcb.200204056
Zhu GZ Miller BJ Boucheix C Rubinstein E Liu CC Hynes RO Myles DG Primakoff P Residues SFQ (173–175) in the large extracellular loop of CD9 are required for gamete fusion Development 2002 129 1995 2002 11934865
Kitadokoro K Bordo D Galli G Petracca R Falugi F Abrignani S Grandi G Bolognesi M CD81 extracellular domain 3D structure: insight into the tetraspanin superfamily structural motifs EMBO J 2001 20 12 18 11226150 10.1093/emboj/20.1.12
Seigneuret M Delaguillaumie A Lagaudriere-Gesbert C Conjeaud H Structure of the tetraspanin main extracellular domain. A partially conserved fold with a structurally variable domain insertion J Biol Chem 2001 276 40055 40064 11483611 10.1074/jbc.M105557200
Maecker HT Todd SC Levy S The tetraspanin superfamily: molecular facilitators FASEB J 1997 11 428 442 9194523
Rubinstein E Le Naour F Lagaudrière-Gesbert C Billard M Conjeaud H Boucheix C CD9, CD63, CD81, and CD82 are components of a surface tetraspan network connected to HLA-DR and VLA antigens Eur J Immunol 1996 26 2657 2665 8921952
Claas C Stipp CS Hemler ME Evaluation of prototype TM4SF protein complexes and their relation to lipid rafts J Biol Chem 2001 276 7974 7984 11113129 10.1074/jbc.M008650200
Charrin S Manie S Billard M Ashman L Gerlier D Boucheix C Rubinstein E Multiple levels of interactions within the tetraspanin web Biochem Biophys Res Commun 2003 304 107 112 12705892 10.1016/S0006-291X(03)00545-X
Berditchevski F Odintsova E Sawada S Gilbert E Expression of the palmitoylation-deficient CD151 weakens the association of alpha 3beta 1 integrin with the tetraspanin-enriched microdomains and affects integrin-dependent signalling J Biol Chem 2002 277 36991 37000 12110679 10.1074/jbc.M205265200
Charrin S Manie S Oualid M Billard M Boucheix C Rubinstein E Differential stability of tetraspanin/tetraspanin interactions: role of palmitoylation FEBS Lett 2002 516 139 144 11959120 10.1016/S0014-5793(02)02522-X
Yang X Claas C Kraeft SK Chen LB Wang Z Kreidberg JA Hemler ME Palmitoylation of tetraspanin proteins: modulation of CD151 lateral interactions, subcellular distribution, and integrin-dependent cell morphology Mol Biol Cell 2002 13 767 781 11907260 10.1091/mbc.01-05-0275
Berditchevski F Gilbert E Griffiths MR Fitter S Ashman L Jenner SJ Analysis of the CD151-alpha3beta1 integrin and CD151-tetraspanin interactions by mutagenesis J Biol Chem 2001 276 41165 41174 11479292 10.1074/jbc.M104041200
Cannon KS Cresswell P Quality control of transmembrane domain assembly in the tetraspanin CD82 EMBO J 2001 20 2443 2453 11350933 10.1093/emboj/20.10.2443
Toyo-Oka K Yashiro-Ohtani Y Park CS Tai XG Miyake K Hamaoka T Fujiwara H Association of a tetraspanin CD9 with CD5 on the T cell surface: role of particular transmembrane domains in the association Int Immunol 1999 11 2043 2052 10590270 10.1093/intimm/11.12.2043
Kovalenko OV Yang X Kolesnikova TV Hemler ME Evidence for specific tetraspanin homodimers: inhibition of palmitoylation makes cysteine residues available for cross-linking Biochem J 2004 377 407 417 14556650 10.1042/BJ20031037
Boucheix C Thien Duc GH Jasmin C Rubinstein E Tetraspanins and malignancy Exp Rev Mol Med 2001
Hemler ME Specific tetraspanin functions J Cell Biol 2001 155 1103 1107 11756464 10.1083/jcb.200108061
Bienstock RJ Barrett JC KAI1, a prostate metastasis suppressor: prediction of solvated structure and interactions with binding partners; integrins, cadherins, and cell-surface receptor proteins Mol Carcinog 2001 32 139 153 11746826 10.1002/mc.1073
Burkhard P Stetefeld J Strelkov SV Coiled coils: a highly versatile protein folding motif Trends Cell Biol 2001 11 82 88 11166216 10.1016/S0962-8924(00)01898-5
Langosch D Heringa J Interaction of transmembrane helices by a knobs-into-holes packing characteristic of soluble coiled coils Proteins 1998 31 150 159 9593189 10.1002/(SICI)1097-0134(19980501)31:2<150::AID-PROT5>3.0.CO;2-Q
Lupas A Coiled coils: new structures and new functions Trends Biochem Sci 1996 21 375 382 8918191 10.1016/0968-0004(96)10052-9
Walshaw J Woolfson DN Extended knobs-into-holes packing in classical and complex coiled-coil assemblies J Struct Biol 2003 144 349 361 14643203 10.1016/j.jsb.2003.10.014
O'Shea EK Klemm JD Kim PS Alber T X-ray structure of the GCN4 leucine zipper, a two-stranded, parallel coiled coil Science 1991 254 539 544 1948029
Zacharias DA Violin JD Newton AC Tsien RY Partitioning of lipid-modified monomeric GFPs into membrane microdomains of live cells Science 2002 296 913 916 11988576 10.1126/science.1068539
Bowie JU Helix packing in membrane proteins J Mol Biol 1997 272 780 789 9368657 10.1006/jmbi.1997.1279
Snapp EL Hegde RS Francolini M Lombardo F Colombo S Pedrazzini E Borgese N Lippincott-Schwartz J Formation of stacked ER cisternae by low affinity protein interactions J Cell Biol 2003 163 257 269 14581454 10.1083/jcb.200306020
Eilers M Shekar SC Shieh T Smith SO Fleming PJ Internal packing of helical membrane proteins Proc Natl Acad Sci USA 2000 97 5796 5801 10823938 10.1073/pnas.97.11.5796
Javadpour MM Eilers M Groesbeek M Smith SO Helix packing in polytopic membrane proteins: role of glycine in transmembrane helix association Biophys J 1999 77 1609 1618 10465772
Russ WP Engelman DM The GxxxG motif: a framework for transmembrane helix-helix association J Mol Biol 2000 296 911 919 10677291 10.1006/jmbi.1999.3489
Senes A Gerstein M Engelman DM Statistical analysis of amino acid patterns in transmembrane helices: the GxxxG motif occurs frequently and in association with beta-branched residues at neighboring positions J Mol Biol 2000 296 921 936 10677292 10.1006/jmbi.1999.3488
Lemmon MA Flanagan JM Hunt JF Adair BD Bormann BJ Dempsey CE Engelman DM Glycophorin A dimerization is driven by specific interactions between transmembrane alpha-helices J Biol Chem 1992 267 7683 7689 1560003
Lemmon MA Treutlein HR Adams PD Brunger AT Engelman DM A dimerization motif for transmembrane alpha-helices Nature Struct Biol 1994 1 157 163 7656033 10.1038/nsb0394-157
MacKenzie KR Prestegard JH Engelman DM A transmembrane helix dimer: structure and implications Science 1997 276 131 133 9082985 10.1126/science.276.5309.131
Senes A Ubarretxena-Belandia I Engelman DM The Calpha – H...O hydrogen bond: a determinant of stability and specificity in transmembrane helix interactions Proc Natl Acad Sci USA 2001 98 9056 9061 11481472 10.1073/pnas.161280798
Deber CM Khan AR Li Z Joensson C Glibowicka M Wang J Val→Ala mutations selectively alter helix-helix packing in the transmembrane segment of phage M13 coat protein Proc Natl Acad Sci USA 1993 90 11648 11652 8265602
Overton MC Chinault SL Blumer KJ Oligomerization, biogenesis, and signaling is promoted by a glycophorin A-like dimerization motif in transmembrane domain 1 of a yeast G protein-coupled receptor J Biol Chem 2003 278 49369 49377 14506226 10.1074/jbc.M308654200
Li R Gorelik R Nanda V Law PB Lear JD DeGrado WF Bennett JS Dimerization of the transmembrane domain of integrin alphaIIb subunit in cell membranes J Biol Chem 2004 279 26666 26673 15067009 10.1074/jbc.M314168200
Gerber D Sal-Man N Shai Y Two motifs within a transmembrane domain, one for homodimerization and the other for heterodimerization J Biol Chem 2004 279 21177 21182 14985340 10.1074/jbc.M400847200
Kleiger G Grothe R Mallick P Eisenberg D GXXXG and AXXXA: common alpha-helical interaction motifs in proteins, particularly in extremophiles Biochemistry 2002 41 5990 5997 11993993 10.1021/bi0200763
Liu Y Engelman DM Gerstein M Genomic analysis of membrane protein families: abundance and conserved motifs Genome Biol 2002 3 research0054 12372142
Lear JD Stouffer A Gratkowski H Nanda V DeGrado WF Association of a model transmembrane peptide containing Gly in a heptad sequence motif Biophys J 2004 87 3421 3429 15315956 10.1529/biophysj.103.032839
Cosson P Bonifacino JS Role of transmembrane domain interactions in the assembly of class II MHC molecules Science 1992 258 659 662 1329208
Adamian L Liang J Helix-helix packing and interfacial pairwise interactions of residues in membrane proteins J Mol Biol 2001 311 891 907 11518538 10.1006/jmbi.2001.4908
Lee AG Ca2+-ATPase structure in the E1 and E2 conformations: mechanism, helix-helix and helix-lipid interactions Biochim Biophys Acta 2002 1565 246 266 12409199
Doyle DA Morais CJ Pfuetzner RA Kuo A Gulbis JM Cohen SL Chait BT MacKinnon R The structure of the potassium channel: molecular basis of K+ conduction and selectivity Science 1998 280 69 77 9525859 10.1126/science.280.5360.69
Zhou FX Cocco MJ Russ WP Brunger AT Engelman DM Interhelical hydrogen bonding drives strong interactions in membrane proteins Nature Struct Biol 2000 7 154 160 10655619 10.1038/81919
Choma C Gratkowski H Lear JD DeGrado WF Asparagine-mediated self-association of a model transmembrane helix Nature Struct Biol 2000 7 161 166 10655620 10.1038/72440
Gratkowski H Lear JD DeGrado WF Polar side chains drive the association of model transmembrane peptides Proc Natl Acad Sci USA 2001 98 880 885 11158564 10.1073/pnas.98.3.880
Ruan W Lindner E Langosch D The interface of a membrane-spanning leucine zipper mapped by asparagine-scanning mutagenesis Protein Sci 2004 13 555 559 14739334 10.1110/ps.03357404
Partridge AW Therien AG Deber CM Polar mutations in membrane proteins as a biophysical basis for disease Biopolymers 2002 66 350 358 12539263 10.1002/bip.10313
Sanders CR Myers JK Disease-related misassembly of membrane proteins Annu Rev Biophys Biomol Struct 2004 33 25 51 15139803 10.1146/annurev.biophys.33.110502.140348
Cai SJ Khorchid A Ikura M Inouye M Probing catalytically essential domain orientation in histidine kinase EnvZ by targeted disulfide crosslinking J Mol Biol 2003 328 409 418 12691749 10.1016/S0022-2836(03)00275-4
Hamdan FF Ward SD Siddiqui NA Bloodworth LM Wess J Use of an in situ disulfide cross-linking strategy to map proximities between amino acid residues in transmembrane domains I and VII of the M3 muscarinic acetylcholine receptor Biochemistry 2002 41 7647 7658 12056896 10.1021/bi016029c
Guan L Murphy FD Kaback HR Surface-exposed positions in the transmembrane helices of the lactose permease of Escherichia coli determined by intermolecular thiol cross-linking Proc Natl Acad Sci USA 2002 99 3475 3480 11904412 10.1073/pnas.052703699
Roy R Laage R Langosch D Synaptobrevin transmembrane domain dimerization-revisited Biochemistry 2004 43 4964 4970 15109254 10.1021/bi0362875
Luo B-H Springer TA Takagi J A specific interface between integrin transmembrane helices and affinity for ligand PLoS Biol 2004 2 e153 15208712 10.1371/journal.pbio.0020153
Yauch RL Hemler ME Specific interactions among transmembrane 4 superfamily (TM4SF) proteins and phosphatidylinositol 4-kinase Biochem J 2000 351 629 637 11042117 10.1042/0264-6021:3510629
Zhang XA Bontrager AL Hemler ME TM4SF proteins associate with activated PKC and link PKC to specific beta1 integrins J Biol Chem 2001 276 25005 25013 11325968 10.1074/jbc.M102156200
Li W Metcalf D Gorelik R Li R Mitra N Nanda V Law PB Lear JD DeGrado WF Bennett JS A push-pull mechanism for regulating integrin function Proc Natl Acad Sci USA 2005 102 1424 1429 15671157 10.1073/pnas.0409334102
Weiner SJ Kollman PA Case DA Singh UC Ghio C Alagona G Profeta S Weiner P A new force field for molecular mechanical simulation of nucleic acids and proteins J Am Chem Soc 1984 106 765 10.1021/ja00315a051
Kuhlman B Baker D Native protein sequences are close to optimal for their structures Proc Natl Acad Sci USA 2000 97 10383 10388 10984534 10.1073/pnas.97.19.10383
Desmet J De Maeyer M Hazes B Lasters I The dead-end elimination theorem and its use in protein side-chain positioning Nature 1992 356 539 542 10.1038/356539a0
|
15985154
|
PMC1190194
|
CC BY
|
2021-01-04 16:03:52
|
no
|
BMC Struct Biol. 2005 Jun 28; 5:11
|
utf-8
|
BMC Struct Biol
| 2,005 |
10.1186/1472-6807-5-11
|
oa_comm
|
==== Front
BMC UrolBMC Urology1471-2490BioMed Central London 1471-2490-5-121608350610.1186/1471-2490-5-12Research ArticleGenetic polymorphism of the N-acetyltransferase 2 gene, and susceptibility to prostate cancer: a pilot study in north Indian population Srivastava Daya SL [email protected] Rama D [email protected] Department of Urology Sanjay Gandhi post Graduate Institute of Medical Sciences, Lucknow-226014, Uttar Pradesh, India2005 6 8 2005 5 12 12 17 1 2005 6 8 2005 Copyright © 2005 Srivastava and Mittal; licensee BioMed Central Ltd.2005Srivastava and Mittal; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
N-acetyltransferase 2 is phase II metabolizing enzyme that participates in the bioconversion of heterocyclic arylamines into electrophilic nitrenium ions, which are important ultimate carcinogens that are directly implicated in tumor initiation process. This study was conducted to examine; (1) whether the N-acetyltransferase 2 (NAT2) genotype is a risk factor for prostate cancer, (2) to study effect of NAT2 genotype on modifying prostate cancer risk from tobacco use.
Methods
The case control study was undertaken over a period of 28 months and included 130 prostate cancer patients (CaP) and 140 controls. The NAT2 genotypes were identified by PCR-RFLP method in DNA extracted from peripheral blood. Genotype frequencies and the association of genotypes with patients and control groups were assessed by logistic regression model.
Results
We observed non-significant association of rapid acetylator genotype NAT2 (OR = 1.452, 95% CI: 0.54–1.87, P = 0.136) in prostate cancer patients. However significant association was observed between rapid acetylator genotype NAT2 and CaP tobacco users (OR = 3.43, 95% CI: 1.68–7.02, P-value < 0.001) when compared with controls.
Conclusion
The data suggests that the NAT2 rapid acetylator genotypes may play an important role in determining the risk of developing prostate cancer particularly in the tobacco users of north Indian population.
==== Body
Background
The human N-acetylation polymorphism is a genetic trait phenotypically allied by variation in N-acetyltransferases 2 (NAT2) activity with therapeutic agents. Acetylation polymorphism arises from the allelic variations in human arylamine N-acetyltransferases 2 resulting in the production of NAT2 proteins with variable enzyme activity or stability. Certain drugs and chemicals may contribute to the occurrence of adverse drug effects and act as susceptibility factors for certain malignancies such as prostate and bladder cancer [1,2]. NAT2 is one of the phase II enzyme that participate in the bioconversion of heterocyclic arylamines into electrophilic nitrenium ions, which are important ultimate carcinogens that are directly implicated in tumor initiation process [3]. It is expressed at high level in liver and encoded by a polymorphic gene presenting several nucleotide substitutions. Consequently the presence of the different alleles in each individual genome produce a broad range of metabolic phenotypes that vary from fully active rapid metabolizers to the less active alleles of slow metabolizers [4]. Many chemical and dietary carcinogens, such as nitrosoamines and arylamines derived from dietary fat as well as tobacco users product, acquire bioactivation and inactivation by enzymes. This suggests that polymorphism of genes encoding metabolic enzymes may represent potential risk factors [5-7].
Some studies indicated that genetically variable NATs, CYP P450 and GSTs are involved in the metabolism of drugs, carcinogens and natural products; and may be responsible for cancer susceptibility [7,8]. It has been reported that rapid acetylators genotypes of NAT2 may be at increased risk of liver and colon cancer [9], hepatocellular carcinoma [10] and colorectal cancer [11] when exposed to environmental arylamines carcinogens, due to NAT2 rapid acetylator mediated O-acetylation. Recent molecular epidemiological studies have analyzed the relationship between various metabolic enzymes, such as N-acetyltransferases (NATs) and cytochrome P450 (CYP) in etiology of prostate cancer [12,13].
It is known that human express two forms of N-acetyltransferases: NAT1 and NAT2; both genes are polymorphic. A recent review reported the nucleotide and amino-acid changes associated with various alleles and deduced phenotype from genotype. It also summarized results of molecular epidemiologic studies assessing the association of NAT1 and NAT2 genotypes with cancer risk of bladder, colon, breast, lung, head and neck and prostate [14]. A review by Chen, (2001) in prostate cancer suggests that the frequencies of some polymorphisms in certain genes differ among different racial and ethnic groups [15]. Whether these genetic variants can help explain part of the large differences in prostate cancer risk in various populations await further clarification.
The present study was undertaken to study the following objectives, i) To observe the frequencies of rapid and slow acetylators (NAT2) in CaP and control individuals ii) to study effect of NAT2 genotype on modifying prostate cancer risk from tobacco use.
Methods
Patient selection
The study group consisted of 130 prostate cancer patients mean age (63.3 ± 9.9) and 140 normal healthy controls mean age (56.7 ± 13.9). The criteria for the patient selection was based on clinical proforma, pathological, and histo-pathological records from the outpatient department of Sanjay Gandhi postgraduate institute of medical science, Lucknow. This study was approved by ethical committee of the Institute. Only histologically confirmed prostate cancer patients were included in the study. All cancer patients had higher Gleason scores (6–9) and was detected at advance stage due to lack of structured screening program under any health scheme in our country. Informed consent was obtained from each participant. The inclusion criteria for the controls were absence of any prior history of cancer or pre-cancerous lesions. Serological (prostate serum antigen, < 4 ng/dl), physical (digital rectal examination) and radiological examinations were performed in all control individuals in order to exclude the possibility of malignancy. The consumption of tobacco in any form (cigarette/ bidi smoking, chewing tobacco in beetle leaf, pan-masala/ gutka etc.) in both groups (cases and controls) was noted through a detailed questionnaire. The criteria of non-users are persons who never use tobacco related material like cigarette/ bidi smoking, pan-masala/ gutka, or chewing tobacco in beetle leaf whereas tobacco user were those who used all these carcinogenic material. These questionnaires were published in our other studies [24-26].
PCR-RFLP and alleles genotyping
Genomic DNA was isolated from peripheral leucocytes by Proteinase -K digestion and phenol/chloroform method [16]. The NAT2 genotypes were determined using the PCR-RFLP as described previously [17]. PCR product of 1093 bp was generated by polymerase chain reaction using the following primer:
Forward 5'-TCTAGCATGAATCACTCTGC-3'
Reverse 5'- GGAACA AATTGG AC TTGG -3'.
Genomic DNA 200 ng was added to a PCR mixture, comprising 18.5 pmol of each primer, 200 μmol dNTP, 1.5 unit of Taq polymerase, and 5 μl, 10X PCR buffer (10 mmol/ml Tris HCl pH = 8.4, 50 mmol/ml KCl and 2.5 mol/ml MgCl2) in a total volume of 50 μl. PTC-100 thermocycler (MJ Research, U.S.A.) for polymerase chain reaction was employed. The reaction mixture was subjected to initial denaturation at 94°C for 5 min, followed by 35 cycles (94°C, 1 min), annealing (58.5°C, 1 min) and extension (72°C, 1 min). The final extension was done at 72°C for 10 min. Following PCR, 7 μl of PCR products were digested with four separate enzymes including Kpn1 for NAT2*5 allele, at 37°C for 2 hrs; Taq1 for NAT2*6 allele, at 56°C for 4 hrs; BamH1 for NAT2*7 allele at 37°C for 2 hrs; and Alu1 for NAT2*14 allele at 37°C for 2 hrs. Digested product was run on 2% agarose gel for NAT2*5, NAT2*7, NAT2*14 alleles and 3% agarose gels for NAT2*6 allele.
NAT2 have many alleles but more common alleles studied frequently are NAT2*5, NAT2*7, NAT2*14 and NAT2*6 allele as described in the present study also. However, we could not genotypes many other additional alleles of NAT2, which could be the limitation in our study.
Estimating the frequency of rapid and slow acetylator
The variant and non-variant NAT2 alleles were recorded and rapid or slow acetylator phenotype assignments were deduced on the basis of NAT2 genotype [4]. Genotypes possessing two variant alleles (NAT2*5, NAT2*6, NAT2*7, or NAT2*14) were assigned as slow acetylator phenotype whereas others were assigned as rapid acetylator phenotype.
Statistical analysis
Statistical analysis was done with SPSS 11.5 software program. Differences in genotype prevalence and association between case and control groups were assessed by binary logistic regression model. Odds ratios (OR) and its 95% confidence interval (CI) were obtained by summarizing data over two habit strata (tobacco users/ non-users). We evaluated age adjusted (confounder OR) and age unadjusted odds ratios, and 95% CI using logistic regression models. Univariant analysis, odds ratios, and 95% CI were used to describe the strength of association.
Results
Comparative details of the observed frequency of different alleles are presented in (Table 1); which indicates significant interethnic variation in NAT2 genotypes in different populations. The NAT2*5 and NAT2*6 allele is most commonly present in our population and also in South Indian and Caucasian- American population but is rare in Japanese population, whereas the NAT2*14 allele is only found in African – American population.
Table 1 Frequency of NAT2 alleles in of north Indian control and other ethnic population.
Population Allelic frequency of NAT2 Reference
NAT2*5 NAT2*6 NAT2*7 NAT2*14
North Indians (n = 140) 0.50 0.30 0. 25 0.0 Present study
South Indians (n = 166) 0. 22 0.37 0. 25 0.0 [20]
Caucasian-American (n = 372) 0. 45 0.28 0.02 0.0 [22]
African – American (n = 128) 0.30 0. 22 0.02 0.09 [22]
Japanese (n = 173) 0.01 0.20 0.13 – [1]
The distribution of genotypes of NAT2 in control and cancer patients is shown in (Table 2). Higher frequency of NAT2 rapid acetylator was observed (64.6%) among the patient groups as compared to the controls (55.7%). However, this was statistically non significant (OR = 1.452, 95% CI: 0.54–1.87, P = 0.136).
Table 2 Frequency distribution of NAT2 genotypes in prostate cancer patients and controls.
Patients NAT2 genotype P – value Unadjusted OR (95% CI) Adjusted OR
Slow – acetylators Rapid-acetylators
Controls (n = 140) 62 (44.29%) 78 (55.71%) 1.0 (Reference) 1.0
Prostate cancer (n = 130) 46 (35.38%) 84 (64.62%) 0.136 1.452 (0.54–1.87) 1.348 (0.39–2.28)
Adjusted OR = Age adjusted odds ratio.
The association between tobacco users and NAT2 genotypes are summarized in (Table 3). The OR for the rapid acetylator genotypes verses slow acetylator genotypes was 3.43 fold higher for the susceptibility of prostate cancer as compared to the controls (OR = 3.43, 95% CI: 1.68–7.02, P-value = 0.001) (Table 3).
Table 3 Association between NAT2 acetylator genotypes with tobacco users and prostate cancer patients.
Tobacco users / non -users Controls (n = 140) Ca-Prostate (n = 130) P-value Unadjusted OR (95% CI) Adjusted OR (95% CI)
Non-users
Slow – acetylator 43(42.16) 29 (42.03) 1.0(Ref) 1.0(Ref)
Rapid-acetylator 59(57.84) 40(57.97) 0.987 1.01(0.49–2.11) 1.04(0.54–2.03)
Tobacco Users
Slow – acetylator 19 (50%) 17(27.87) 0.492 1.33(0.49–2.47) 1.45 (0.61–3.46)
Rapid-acetylator 19 (50%) 44(72.13) 0.001 3.43(1.68–7.02) 4.37(2.02–9.45)
Adjusted OR = Age adjusted odds ratio.
We categorized prostate serum antigen value (PSA) into three group (PSA = <10 ng/ml, >10 ng/ml,>20 ng/ml) and Gleason score into two group (GS = 6–7 and 8–9). And we observed that NAT2 genotypes were non-significant with PSA (P = 0.090) or Gleason score (= 0.678) for risk of prostate cancer in our population (Fig 1 and Fig 2).
Figure 1 <10 = PSA (prostate serum antigen) value less than 10 ng/ml, >10 = PSA value in between 10–20 ng/ml and >20 = PSA value greater than 20 ng/ml.
Figure 2 6–7 = Gleason score 6 and 7, 8–9 = Gleason score 8 and 9 of prostate tumor.
Discussion
The present NAT2 genotyping study based on molecular methods in discriminating the acetylator genotypes both in controls and prostate cancer patients is the first of its kind in north Indian population. Rapid acetylator genotypes were comparatively predominant (55.7% and 64.6%) as compared to the slow acetylator genotypes (44.3% and 35.4%).
The results observed in the present study suggest, that NAT2 genotype has a trend of association for prostate cancer risk when considered alone (OR = 1.452, 95% CI: 0.54–1.87, P = 0.136) but is statistically non-significant (Table 2). However, no association could be established between NAT2 genotype and PSA and/or Gleason score (Fig 1 and Fig 2). Our findings agree with previous studies that showed significant association of prostate cancer for NAT2 rapid acetylator genotype in American study (23) whereas non significant association was observed in Swedish, Danish [12]) and Spanish population [13]. On the contrary, there was significant association reported with slow acetylator genotypes of NAT2 in Japanese population [1]. The discrimination in association study from our observation could be related to ethnic variation.
However, in the present study we observed significant association between the NAT2 genotypes in tobacco users as compared to the controls (Table 3). The synergistic interaction between the rapid acetylators genotype with tobacco users obtained in our study implies that exposure to tobacco, activates heterocyclic amines that are substrates for NAT2 which may increase the risk for prostate cancer. These observations are in agreement with the reports published by the investigators in liver and colon. In the liver heterocyclic amines may be N-hydroxylated by the hepatic CYP1A2, and in turn O-acetylated by NAT enzymes to an active form that can develop DNA adducts [5,9]. NAT2 genotypes studied in hepatocellular carcinoma [10] and colorectal cancer [11] have indicated the prevalence of rapid acetylator among patient population. Thus, the present study suggests that rapid acetylator genotype could be associated with the susceptibility to prostate cancer especially in tobacco users. The mechanisms behind this association indicate that the slow acetylator genotypes should decrease the generation of critical intracellular concentration of such ultimate carcinogens, and thus reduce tumorogenesis upon environmental exposure (tobacco users). However, rapid acetylator genotypes should increase the generation of such ultimate carcinogens and enhance tumorogenesis by the pathway of O-acetylation. The results found in the present study reveal a markedly increased frequency of allele encoding the active genotype, among the patients cohort that entirely fit the above model, and are consistent with genotype / phenotype based studied that demonstrate an excess of rapid acetylator among the prostate cancer patients.
In the controls, slow allele of NAT2 is present up to 90% in Arab population [21], 40–60% in Caucasians including Indian [18] and 5–25% East Asian [19]. We observed 44% of slow acetylator genotypes; however, another study from South Indian population indicated 74% of slow acetylator genotype [20]. Thus it indicates that frequency of slow allele observed in our population matched with studies in Caucasians population [18]. Differences of distribution of slow allele of NAT2 between our and south Indian population is due to the different ethnic /or geographical environment.
Conclusion
In conclusion, this study indicates that NAT2 rapid acetylator genotype exhibit a trend of association with the risk of developing prostate cancer, and more so in case of patients who are tobacco users.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
DSL Srivastava carried out the molecular genetic studies, participated in analyzing the data & drafted the manuscript. RD Mittal participated in designing of the study & manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
The authors wish to thank Dr. A. Mandhani for providing the clinical samples and Dr. B. Mittal for helpful suggestions. One of the authors (D.S.L. Srivastava) is thankful to Council of Scientific and Industrial Research, New Delhi, for awarding senior research fellowship and financial aid.
==== Refs
Hamasaki T Inatomi H Katoh T Aono H Ikuyama T Muratani T Matsumoto T NAT2 gene polymorphism as a possible biomarker for susceptibility to prostate cancer in Japanese Int J Urol 2003 10 167 173 12622714 10.1046/j.1442-2042.2003.00586.x
Inatomi H Katoh K Kawamoto T Matsumoto T NAT2 gene polymorphism as a possible biomarker for susceptibility to bladder cancer in Japanese Int J Urol 1999 6 446 54 10510890 10.1046/j.1442-2042.1999.00096.x
Hein DW Doll MA Gray K Rustan TD Ferguson RJ Metabolic activation of N-hydroxy-2-acetylaminofluorene by NAT1 and NAT2 in colon cytosols of Syrian hamster congenic at the NAT2 locus Cancer Res 1993 53 509 514 8425184
Vatsis KP Weber WW Bell DA Nomenclature for N-acetyltransferase Pharmacogenetics 1995 5 1 17 7773298
Hein DW Doll MA Rustan TD Gray K Feng Y Ferguson RJ Grant DM Metabolic activation and deactivation of arylamine carcinogens by recombinant human NAT1 and polymorphic NAT2 acetyltransferases Carcinogenesis 1993 14 1633 1638 8353847
Cascorbi I Brockmoller J Mrozikiewicz PM Muller A Roots I Arylamines N-acetyltransferase activity in man Drug Metab Rev 1999 31 489 502 10335449 10.1081/DMR-100101932
Caporaso N Goldstein A Cancer genes: single and susceptibility: exposing the difference Pharmacogenetics 1995 5 59 63 7663529
Guengerich FP Influence of nutrients and dietary materials on cytochrome P-450 enzymes Am J Clin Nutr 1995 61 651S 658S 7879733
Turesky RJ Lang NP Butler MA Teitel CH Kadlubar FF Metabolic activation of carcinogenic heterocyclic amines by human liver and colon Carcinogenesis 1991 12 1839 845 1934265
Huang YS Chern HD Wu JC Chao Y Huang YH Chang FY Lee SD Polymorphism of the N-acetyltransferase 2 gene, red meat intake, and the susceptibility of hepatocellular carcinoma Am J Gastroenterol 2003 98 1417 1422 12818290 10.1111/j.1572-0241.2003.07452.x
Gill JPG Lechner MC Increased frequency of wild type arylamines-N-acetyltransferase allele NAT2*4 homozygotes in Portuguese patients with colorectal cancer Carcinogenesis 1998 19 37 41 9472690 10.1093/carcin/19.1.37
Wadelius M Autrup JL Stubbins MJ Andersson SO Johansson JE Wadelius C Wolf CR Autrup H Rane A Polymorphism in NAT2, CYP2D6, CYP2C19 and GSTP1 and their association in prostate cancer Pharmacogenetics 1999 9 333 340 10471065
Agundez JA Martinez C Olivera M Gallardo L Ladero JM Rosado C Prados J Rodriguez-Molina J Resel L Benitez J Expression in human prostate of drug and carcinogen metabolizing enzymes: association with prostate cancer risk Br J Cancer 1998 78 1361 1367 9823980
Hein DW Doll MA Fretland AJ Leff MA Webb SJ Xiao GH Devanaboyina US Nangju NA Feng Y Molecular genetics of NAT1 and NAT2 acetylation polymorphisms Cancer Epidemiol Biomarkers Prev 2000 9 29 42 10667461
Chen C Risk of prostate cancer in relation to polymorphism of metabolic genes Epidemiol Rev 2001 23 30 35 11588852 10.1093/aje/153.1.30
Sambrook J Fritsch E Maniatis T Molecular Cloning-A Laboratory Manual 2nd Ed 1989 Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press
Hsieh FI Pu YS Chern HD Hsu LI Chiou HY Chen CJ Genetic polymorphisms of N-acetyltransferase 1 and 2 and risk of cigarette smoking-related bladder cancer Br J Cancer 1999 81 537 541 10507782 10.1038/sj.bjc.6690727
Xie HG Xu ZH Ou-Yang DS Shu Y Yang DL Wang JS Yan XD Huang SL Wang W Zhou HH Meta-analysis of phenotype and genotype of NAT2 deficiency in Chinese population Pharmacogenetics 1997 7 503 514 9429236
Whoolhouse NM Qureshi MM Bastaki SMA Polymorphic N-acetyltransferase (NAT2) genotyping of Emiratis Pharmacogenetics 1997 7 73 82 9110365
Anitha A Banerjee M Arylamine N-acetyltransferase 2 polymorphism in the ethnic population of South Indians Int J Mol Med 2003 11 125 131 12469231
Lin HJ Han CY Lin BK Hardy S Ethnic distribution of slow acetylator mutation in the polymorphic N-acetyltransferase (NAT2) gene Pharmacogenetics 1994 4 125 134 7920692
Bell DA Taylor JA Butler MA Stephens EA Wiest J Brubaker LH Kadlubar FF Lucier GW Genotype/phenotype discordance for human arylamine N-acetyltransferase (NAT2) reveals a new slow acetylator allele common in African-Americans Carcinogenesis 1993 14 1689 92 8102597
Wang CY Jones RF Debiec-Rychter M Soos G Haas GP Correlation of the genotypes for N-acetyltransferases 1 and 2 with double bladder and prostate cancers in a case-comparison study Anticancer Res 2002 22 3529 35 12552951
Mittal RD Srivastava DS Kumar A Mittal B Polymorphism of GSTM1 and GSTT1 genes in prostate cancer: A Study from North India Indian J Cancer 2004 41 115 119 15472409
Srivastava DS Kumar A Mittal B Mittal RD Polymorphism of GSTM1 and GSTT1 genes in bladder cancer: a study from North India Arch Toxicol 2004 78 430 434 15057507 10.1007/s00204-004-0559-y
Srivastava DS Mandhani A Mittal B Mittal RD Genetic polymorphism of glutathione S-transferase genes (GSTM1, GSTT1 and GSTP1) and susceptibility to prostate cancer in Northern India BJU Int 2005 95 170 173 15638917 10.1111/j.1464-410X.2005.05271.x
|
16083506
|
PMC1190195
|
CC BY
|
2021-01-04 16:30:02
|
no
|
BMC Urol. 2005 Aug 6; 5:12
|
utf-8
|
BMC Urol
| 2,005 |
10.1186/1471-2490-5-12
|
oa_comm
|
==== Front
Cancer Cell IntCancer Cell International1475-2867BioMed Central London 1475-2867-5-231604580210.1186/1475-2867-5-23Primary ResearchTherapeutic Electromagnetic Field (TEMF) and gamma irradiation on human breast cancer xenograft growth, angiogenesis and metastasis Cameron Ivan L [email protected] Lu-Zhe [email protected] Nicholas [email protected] W Elaine [email protected] C Douglas [email protected] Department of Cellular and Structural Biology, University of Texas Health Science Center at San Antonio, 7703 Floyd Curl Drive, San Antonio, Texas 78229, USA2 Pennington Biomedical Research Center, Louisiana State University, 6400 Perkins Road, Baton Rouge, Louisiana 70808, USA3 EMF Therapeutics, Inc., P.O. Box 679, Signal Mountain, Tennessee 37377, USA2005 26 7 2005 5 23 23 5 7 2004 26 7 2005 Copyright © 2005 Cameron et al; licensee BioMed Central Ltd.2005Cameron et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The effects of a rectified semi-sinewave signal (15 mT amplitude, 120 pulses per second, EMF Therapeutics, Inc.) (TEMF) alone and in combination with gamma irradiation (IR) therapy in nude mice bearing a human MDA MB231 breast cancer xenograft were tested. Green fluorescence protein transfected cancer cells were injected into the mammary fat pad of young female mice. Six weeks later, mice were randomly divided into four treatment groups: untreated controls; 10 minute daily TEMF; 200 cGy of IR every other day (total 800 cGy); IR plus daily TEMF. Some mice in each group were euthanized 24 hours after the end of IR. TEMF treatment continued for 3 additional weeks. Tumor sections were stained for: endothelial cells with CD31 and PAS or hypoxia inducible factor 1α (HIF).
Results
Most tumors <35 mm3 were white but tumors >35 mm3 were pink and had a vascularized capsule. The cortex within 100 microns of the capsule had little vascularization. Blood vessels, capillaries, and endothelial pseudopods were found at >100 microns from the capsule (subcortex). Tumors >35 mm3 treated with IR 24 hours previously or with TEMF had decreased blood vessels in the subcortex and more endothelial pseudopods projecting into hypoxic, HIF positive areas than tumors from the control group. Mice that received either IR or TEMF had significantly fewer lung metastatic sites and slower tumor growth than did untreated mice. No harmful side effects were attributed to TEMF.
Conclusion
TEMF therapy provided a safe means for retarding tumor vascularization, growth and metastasis.
electromagnetic fieldbreast cancerionizing irradiationangiogenesismetastasis
==== Body
Background
In a previously published experimental research report, it was found that exposing a transplantable murine mammary adenocarcinoma to a 15 mT EMF given at 120 pulses per second for 10 minutes per day significantly reduced tumor growth and vascularization and resulted in an increased survival time [1]. This published report appears to be the only literature available on the use of pulsating magnetic fields to reduce tumor angiogenesis. The authors of this report suggested that the magnetic field treatment used acted to reduce tumor angiogenesis and might have value as an alternative therapeutic modality for treatment of patients with tumors. The study reported here was designed to further investigate the potential of the same EMF therapy to inhibit growth and angiogenesis of a human breast cancer xenograft and to compare the effects of: 1) a commonly used course of gamma irradiation (IR) involving exposure to 200 cGy every second day for a total of 800 cGy, 2) daily exposure to TEMF, and 3) a combination of these two therapeutic treatment regimens on tumor growth, tumor angiogenesis, tumor metastasis, and of the side effects of each treatment regimen. Although this study used whole body IR therapy, most IR therapy of human patients is restricted to localized targeted regions of the body to avoid general side effects of IR treatment. The MDA MB231 cancer cell line transfected with and expressing a green fluorescent protein (GFP) gene was used to facilitate study of metastases of cancer cells from the site of the primary tumor [2].
Our study results demonstrate the potential of TEMF therapy to retard tumor: growth, angiogenesis, and metastasis, without harmful side effects.
Results
Body Weight
Once the mice were divided into treatment groups the body weight of each mouse was measured every 3 to 4 days for the remainder of the experiment. As illustrated in Fig. 1, the two groups that received IR therapy every second day for 8 days demonstrated a mean body weight loss beginning during IR therapy and lasting until about 8 or 9 days after the end of IR therapy. After completion of the IR therapy, the irradiated mice again began to regain their weight toward the mean weight of the two groups of mice not subjected to IR therapy. The group of mice that received only EMF therapy demonstrated a continuous increase in mean body weight similar to the group of mice given no therapy.
Figure 1 Body weights during the course of the experiment. The two groups of mice that received gamma irradiation both lost body weight during and for a few days after the course of exposure, but the body weights later recovered towards the weights of the two groups of mice not exposed to gamma irradiation.
Tumor Growth
Fig. 2 illustrates mean tumor volume change for each of the four treatment groups starting at the beginning of IR and/or EMF therapy. All tumors in each therapy group were less than 35 mm3 at the start of treatment period. To statistically assess tumor growth rate, the data on each tumor in each group of mice was subjected to linear regression analysis. Tumor volume gave a good fit to a linear regression model. The slope (growth rate) derived from the linear regression of each tumor volume was used to determine any statistical differences in growth rates between treatment groups (Fig. 2). Growth rate of tumors from the untreated group was significantly faster (p < 0.001) than any of the three groups of treated mice. The tumor growth rate of the group of mice treated solely with EMF therapy was significantly less than the tumor growth rate of the untreated group. The tumor growth rates of the two groups of mice treated with gamma irradiation were significantly less than the tumor growth rates of the untreated or TEMF only treated groups.
Figure 2 Illustrates the effect on tumor growth of: the 8-day course of gamma irradiation therapy (IR), the daily exposure to the TEMF therapy (TEMF), a combination of gamma irradiation and TEMF therapy (TEMF/IR), and no treatment (Control). The tumors were all 35 mm3 or less at the start of treatment. The growth rate is the slope of the linear regression of the tumor volumes at each time. It can be seen that tumor growth was beginning to deviate from the linear fit about 18 days after the start of TEMF or radiation treatments however there was still a reasonable linear fit. ANOVA was used to determine any statistical differences in growth rates between treatment groups. Different letters demonstrate significant differences. Gamma irradiation alone (IR) or in combination with TEMF (TEMF/IR) reduced the tumor growth. The TEMF therapy resulted in a slower tumor growth rate than the untreated controls but not as slow as the gamma irradiated mice.
Results of an evaluation of the tumor growth rates from day 7 though day 15 post IR treatment are reported in Fig. 3. As illustrated in Fig. 3, the mean tumor growth rate of mice given neither IR nor TEMF treatment was significantly higher than the three groups given IR and TEMF therapy either alone or in combination. The group of mice given both IR and TEMF followed by daily TEMF had a mean tumor growth rate significantly lower than that of the other three groups. Clearly, the continued application of daily TEMF therapy following IR therapy had an additive inhibitory effect on the tumor growth rate during the 7 to 15 days after the course of IR therapy.
Figure 3 Mean ± SEM tumor growth rate data from day 7 to day 15 post IR therapy in each treatment group. ANOVA was used to determine any statistical differences in growth rates between treatment groups. Different letters demonstrate significant differences. Tumors in the untreated control group grew significantly faster than did tumors in the other three groups. Tumors in the IR/TEMF group had a growth rate significantly lower than that of the other three groups.
Tumor vasculature
While making measurements on tumor volume, tumor color was observed through the skin of the nude mice. After recording tumor color and volume of each mouse, it became clear that smaller tumors were white while larger tumors were pink. Fig. 4 illustrates the relationship between tumor color and tumor volume. Almost all tumors less than 35 mm3 were white whereas tumors greater than approximately 35 mm3 were pink in color. Application of the pressure from one's finger to a pink tumor followed by the rapid removal of the finger pressure showed a rapid color change from white to pink.
Figure 4 Relationship between color of the tumor as observed through the skin and the measured tumor volume. Tumors <35 mm3 were mostly white while tumors >35 mm3 were pink.
Histological examination of midsections of PAS stained tumors was done to assess tumor vascularization patterns. Tumors with a volume >35 mm3 demonstrated a connective tissue capsule with blood vessels (Fig. 5A). At higher magnification the cortical area within about 100 μm of the capsule revealed few blood vessels while the area greater than 100 μm from the capsule showed evidence of considerable blood vessel and capillaries with many endothelial pseudopods extending away from the capillaries (Fig. 5B&5C). The general direction of the pseudopods was parallel to the tumor capsule surface. Immunohistochemical localization of CD-31, used as a specific marker of endothelium, demonstrated a positive reaction of pseudopods (Fig. 5D). Areas of tumor necrosis were observed, below the subcortical area (Fig. 5E&5F). Immunohistochemical localization of hypoxia-inducible factor 1-α (HIF) reveals the subcortical area of the tumor to contain HIF positive cells while the tumor capsule, cortex and necrotic areas of the tumor demonstrate no evidence of HIF. Thus, the area found to be HIF positive were enriched in endothelial pseudopods.
Figure 5 Photomicrographs illustrating the pattern of the tumor vascular network (A and B), of endothelial pseudopods stained with PAS (C) and with CD-31 specific endothelial markers (D), and of the localization of HIF-α (E and F). A, The tumor capsule (left) reveals blood vessels. The cortex under the capsule reveals no blood vessels and few endothelial pseudopods while the subcortical area to the right has more pseudopods. B, The subcortical area of the tumor reveals a small blood capillary with multiple endothelial pseudopods protruding at right angles into the tumor mass. C, At high magnification endothelial pseudopods are seen to branch and occasionally have a vacuole/lumen (arrow). D, The endothelial pseudopods react positively to the CD-31 specific endothelial marker. The CD-31 antibody identifies endothelial cells using the avidin-biotin peroxidase complex method. E, Viable cell area can be seen beneath tumor capsule (left). Necrotic area can be seen to the right. F, Enlarged subcortical area from E. In E the hypoxic area between the viable and the necrotic tissue is stained brown.
Ocular grid intercept couting was used to quantify tumor vascularization in 8 μm thick PAS stained histological sections of the tumors. The numbers of ocular grid intercepts were scored over the area of: blood vessels and capillaries, endothelial pseudopods and over the area with no indication of these structures. This method has been shown to be a usable measure of volume density occupied by recognizable structures. The results of the scoring of blood vessels and of endothelial pseudopods in the subcortical regions of the tumors are summarized in Fig. 6A&6B. Statistical analysis of the mean blood vessel volume density versus the pseudopod volume density reveals a significant exponential correlation coefficient of 0.966. Thus, as the blood vessel volume density decreased, the endothelial pseudopod volume density increased. As illustrated in Fig. 6A&6B, IR resulted in significantly reduced blood vessel volume density but a significantly increased volume density of pseudopods at one day after the last dose of irradiation, but the blood vessel volume density increased and the pseudopod volume density decreased to the level of the untreated control by 22 days after the last dose of irradiation. However, the groups of mice treated solely with EMF had significantly less blood vessel volume density and a significantly higher pseudopod volume density at one day after the end of IR. The low blood vessel volume density and the high pseudopods volume density in the tumors of the EMF treated mice remained the same at 22 days after the end of the IR treatment.
Figure 6 Quantification of changes in tumor vascularization between treatment groups. A and B, The percent of areas (volume density) of blood vessels and the percent of area of endothelial pseudopods were determined using an ocular grid intercept counting method. The mean ± SEM of each treatment group is graphed. Columns that do not share a common letter within a graph are significantly different (p < 0.01). The data indicate that gamma irradiation and EMF alone or in combination decreased the total area of blood vessels and increased the total area of pseudopods compared to the control. Data from the untreated mice (Control) and the EMF treated mice (TEMF) were pooled from mice from the early and from the late sacrifice because there was no significant time of sacrifice difference within these groups.
Metastasis
The number of lung metastatic colonies per mouse in each of the four treatment groups is summarized in Fig. 7. Both the mean number of colonies in the lungs per mouse as well as the incidence of metastasis was significantly higher in the untreated group than in the three groups treated with IR and/or EMF therapy. There were no significant differences between the three treatment groups in number or in incidence of lung metastatic colonies per mouse.
Figure 7 Metastasis of GFP-expressing human breast MDA MB231 cells from the site of inoculation in the inguinal mammary fat pad to lodge in the lungs of mice in each of the four treatment groups. The lungs were removed and smashed between microscopic slides and microcolonies of GFP cells were observed by epilumination using blue light and detected by presence of green fluorescence microcolonies observed using both 3.5× and 10× objective lens. The untreated mice had a significantly higher number of GFP positive microcolonies than the groups of mice that received gamma irradiation therapy or those mice that received EMF therapy. There were no other significant differences between groups in either mean number of microcolonies per lung or in incidence. Horizontal lines indicate mean values for each treatment group.
Side Effects of Gamma Irradiation Therapy and of TEMF Therapy
Mice were euthanized at both one day and at 22 days after the last IR exposure. Data on liver and spleen weights are summarized in Table 1, and data on blood counts are summarized in Table 2. There were no statistically significant differences between treatment groups in mean liver weights at the earlier or later kills. On the other hand, the spleen weights of mice that were euthanized one day after their last exposure to IR were significantly less than the spleen weights of the non-gamma irradiated groups. The mean spleen weights of the IR mice recovered to the level of the untreated groups by 22 days after their last IR exposure. There was no significant loss of spleen weight at the time of early or at the time of late euthanasia due to TEMF therapy.
Table 1 Gamma Irradiation Therapy (IR) and Therapeutic Electromagnetic Field Therapy (TEMF) on Liver and Spleen Weight (Means ± SEM)
Therapy Group n Liver weight (grams) Spleen weight (grams)
Early1
Control 5 0.94 ± 0.04 0.107 ± 0.009
IR 5 0.87 ± 0.07 0.031 ± 0.004
TEMF 5 1.04 ± 0.05 0.110 ± 0.006
TEMF/IR 5 0.93 ± 0.04 0.033 ± 0.004
Late2
Control 9 1.26 ± 0.07 0.166 ± 0.008
IR 20 1.17 ± 0.03 0.146 ± 0.008
TEMF 10 1.16 ± 0.05 0.145 ± 0.012
TEMF/IR 10 1.15 ± 0.05 0.151 ± 0.014
1Mice sacrificed one day after last irradiation treatment.
2Mice sacrificed 22 days after last irradiation treatment.
Results of Statistical Analyses:
1. Liver weight: No significant differences among treatment groups.
2. Spleen weight: At early kill, groups treated with gamma irradiation show significant (p < 0.01) loss of weight compared to groups not treated with gamma irradiation. In late kill, no significant differences were detected.
Table 2 Gamma Irradiation Therapy (IR) and Electromagnetic Field Therapy (TEMF) on Blood Counts(Means ± SEM)
Therapy Group n WBC × 103/μL RBC × 106/μL Platelets × 103/μL Micronuclei (%RBC)
Early1
Control 4 2.91 ± 0.83 9.08 ± 0.07 584 ± 76 1.3 ± 0.04
IR 4 0.11 ± 0.01 7.50 ± 0.16 389 ± 17 1.2 ± 0.03
TEMF 5 2.89 ± 0.60 9.05 ± 0.12 505 ± 61 1.3 ± 0.04
TEMF/IR 4 0.23 ± 0.06 7.68 ± 0.13 443 ± 34 1.2 ± 0.03
Late2
Control 8 1.85 ± 0.37 8.30 ± 0.14 551 ± 83 1.6 ± 0.06
IR 13 1.48 ± 0.332 7.06 ± 0.16 973 ± 98 0.7 ± 0.01
TEMF 11 1.15 ± 0.29 8.85 ± 0.22 613 ± 62 0.7 ± 0.01
TEMF/IR 10 1.07 ± 0.13 6.58 ± 0.11 1007 ± 141 0.6 ± 0.01
1Mice sacrificed one day after last irradiation treatment.
2Mice sacrificed 22 days after last irradiation treatment.
Results of Statistical Analyses:
1. At the end of irradiation therapy plus one day, gamma irradiation caused significant decreases in WBC, RBC, and platelet counts. No other significant differences at p < 0.01.
2. At the end of irradiation therapy plus 22 days, gamma irradiation caused significant decreases in WBC and RBC counts and a significant increase in platelet counts. No other significant differences at p < 0.01.
There was a significant decrease in WBC, RBC, and platelets counts attributed to gamma irradiation exposure at 1 day after the end of radiation treatment (Table 2). There was no significant difference in WBC, RBC, and platelets counts attributed to TEMF treatment. At 22 days after the end of the IR treatment, the WBC and RBC counts in the two IR groups were not quite as low but remained significantly lower than in the two non-irradiated groups while the IR groups had significantly higher platelet counts than any of the other groups. Clearly, platelet counts have made a significant compensatory rebound by 22 days after the end of IR therapy. Statistical analysis of micronuclei counts indicated that there was no significant genotoxic damage due to either treatment at either time of euthanasia.
Another measure of possible side effects was the scoring of mitotic activity in the duodenal crypts. Fig. 8 summarizes results of mitotic activity in the duodenal crypts of mice sacrificed one day after the end of the IR therapy regimen. The two groups of mice treated with IR 24 hours previously had significantly fewer metaphase figures per crypt than the non-treated group or to the group of mice treated solely with the course of TEMF therapy.
Figure 8 Number of metaphase figures per midaxial histological section of duodenal crypts. Column height indicates mean ± SEM. The data are from the mice sacrificed one day after the last gamma irradiation exposure. The columns that do not share a common letter are significantly different. Gamma irradiation but not EMF decreased the number of metaphase figures in the duodenal crypts.
All mice treated with IR demonstrated a tan skin discoloration immediately after the 8-day course of IR therapy, but the tan skin color returned to the normal skin color by 22 days after the course of IR therapy was terminated. The skin color in the two non-irradiated groups of mice remained normal in appearance throughout the course of the study.
Discussion
On tumor vascularization
The immunohistochemical localization of hypoxia induced factor (HIF) in the tumors helped explain the vascular pattern of the encapsulated tumors. The HIF positive reactivity in the tumor was localized to the subcortical areas of the tumor in the same area found to have the most blood vessels, capillaries and pseudopods. These observations suggest that this subcortical area of the tumor is hypoxic leading to production of angiogenesis growth factors, which in turn acts on the host endothelial cells to sprout pseudopods. Sprouting of cultured endothelial cells in response to vascular endothelial growth factors is known to occur [4]. Feraud et al. also reported that this endothelial cell sprouting could be inhibited by several angiostatic agents. The presence of pseudopodal like structures in viable areas of a tumor immediately adjacent to necrotic areas of the tumor has been previously reported [5]. The lack of HIF reactivity in the cortical area beneath the well vascularized tumor capsule suggest this cortical area is not hypoxic and not in need of an extensive vasculature. It appears that the PSA staining of endothelial cell sprouting pseudopods in tumors gives quantifiable spatial information on where in the tumor that HIF and vascular endothelial growth factors are located.
Action of IR and of EMF therapies
EMF therapy has been reported to slow tumor growth and tumor angiogenesis and to increase the survival time of tumor bearing mice [1]. An experiment was therefore designed to test if the combining of IR therapy and EMF therapy in a tumor bearing mouse model might give an addictive therapeutic index.
Over half of all cancer patients are given ionizing radiation (IR) during their course of treatment [6]. The interactions between IR and tumor vascularization are complex [7,8]. It has been shown that IR therapy is dependent on tumor vascularization and that IR therapy initially interferes with tumor vascularization leading to tumor hypoxia [9]. However hypoxic areas are radio-resistant until they can induce angiogenesis and re-oxygenation. Hypoxic areas of tumors produce HIF leading to production of angiogenesis growth factors, which stimulate angiogenesis to allow further tumor growth. This known sequence of events following an IR treatment of a tumorous patient has lead to the combined use of IR therapy coupled to use of an anti-angiogenesis agent. By blocking angiogenic factors, like VEGF, and by blocking the formation of new vessels in combination with IR therapy researchers have dramatically increased the efficacy of IR therapy in a number of tumor types [7-9]. The use of anti-angiogenic agents has proved useful in the treatment of cancer progression in most preclinical and clinical trials [10,11]. However the prevention of metastasis (Inoue et al.) by anti-angiogenic agents [11,12] has not gone unchallenged [[13] and see rebuttal by Kieran et al. [14]].
The experimental results indicate that either IR therapy or EMF therapy suppressed tumor growth. The IR therapy proved to be more effective at suppressing tumor growth than did the EMF therapy while the IR therapy but not the EMF therapy had harmful side effects. Both IR therapy and EMF therapy reduced blood vessel volume density within the tumor when measured one day after the end of the IR therapy but blood vessel volume density in tumors of the IR treated mice returned to pre IR treatment levels. The continual use of daily EMF therapy in the IR treated mice did suppress the return of blood vessel volume density within the tumor following IR therapy. The combination of both therapeutic modalities therefore led to a sustained and significant reduction in the extent of tumor blood vessel volume density.
That pseudopod volume density was relatively high at 22 day after IR in the group of mice that were on sustained daily EMF therapy suggests that major areas of these tumors were anoxic and thus were producing HIF and VEGF's resulting in the sprouting of the endothelial cell pseudopods. This indicates that these early events in the angiogenesis process were not suppressed by the continued use of EMF therapy. Because continued use of EMF did suppress tumor blood vessel volume density at 22 days after IR therapy the EMF therapy may be acting to suppress one of more of the later steps in the process of tumor vascularization such as: endothelial cell proliferation, lumen formation, generation of vascular tubes and initiation of blood flow [15].
As illustrated in Fig. 6A, the continued daily use of TEMF following the course of IR therapy suppressed tumor blood vessel volume density compared to mice given the IR therapy alone. The data on tumor growth rate, beginning a week after IR therapy, revealed that the continuation of TEMF therapy suppressed the mean tumor growth rate of the tumors compared to those of mice treated only with IR therapy. In fact, linear regression analysis of the data indicates that the slope of the tumor growth rate of the IR plus TEMF treated mice did not differ from a slope of zero. Thus, these tumor growth rate results indicate a significant treatment advantage for continuous daily TEMF therapy following a standard course of IR therapy.
Since the extent of tumor vascularization has been positively linked to the likelihood of tumor cell metastasis [12,16-18] it was anticipated that a decrease in the blood vessel volume density in the tumor, as observed in the experiments reported herein, would decrease the chance for tumor cells to metastasize to distant sites in the tumor bearing mouse model used in the study. Inspection of the incidence of metastasis observed in the lungs of human breast cancer xenographs revealed: that IR therapy alone reduced metastasis by 49%, that EMF therapy alone reduced metastasis by 54% while the combination of IR and EMF therapies when followed by daily EMF therapy resulted in a 73% reduction in incidence of metastasis to the lung. Although all three of these therapy regimens caused a significant decrease in incidence of metastasis of tumor cells to the lungs the greatest decrease, although not statistically significant, was with the combination of IR and EMF therapies.
In summary the results indicate that sustained EMF therapy by itself reduced: the extent of tumor blood vessel volume density, the tumor growth rate and tumor cell metastasis to the lung, without harmful side effects. TEMF therapy alone may therefore be a useful alternate therapy for cancer patients who decide not to undergo standard radiation or for chemotherapy. The IR therapy alone markedly reduced: the tumor growth rate, reduced tumor metastasis but only transitly reduced tumor blood vessel volume density. The IR therapy did however have harmful side effects. These side effects could be minimized by targeted IR exposure. The combination of IR therapy and continued TEMF therapy markedly suppressed the return of blood vessel volume density and tumor growth following IR therapy and resulted in the lowest incidence of tumor metastasis. Thus these findings support the ideas that TEMF of the type used in this experiment may be an effective antiangiogenic therapy and that a combination of an anti-vascular modality, such as TEMF, with IR therapy may enhance the therapeutic index of IR therapy for cancer.
Materials and methods
Animals
Female athymic nude mice (purchased from the Harlan Sprague Dawley, Inc., Indianapolis, IN), 6 weeks of age, were used for this study. The animals were housed under pathogen-free conditions and fed an AIN-76 semipurified diet slightly altered to contain 10% w/w corn oil.
Cell Lines
Human breast cancer cell line MDA-MB-231 was obtained from the American Type Culture Collection. To determine the effect of the treatments on the metastatic potential of the MDA-MB-231 cells they were stably transfected with the enhanced GFP expression plasmid, pEGFP-N1 (Clontech Laboratories, Inc.). The expression of GFP allowed for detection of micrometastatic colonies on the whole flattened lungs with a fluorescence epilumination microscope. The cell lines were cultured in McCoys's 5A medium supplemented with pyruvate, vitamins, amino acids, antibiotics, and 10% fetal bovine serum [2]. HDMECs and the culture medium EGM-2 Mv were maintained at 37°C in a humidified incubator with 5% CO2.
In Vivo Tumor Growth
The MDA-MB-231/GFP cells were harvested from exponential cultures and inoculated at 2 × 106 cells/inoculum in the inguinal mammary fat pad area of female athymic nude mice, 6 weeks of age. When the tumors grew to an average diameter of about 3 mm after 5 weeks, the mice were divided into 4 groups such that the mean and median of tumor volume of the four groups were closely matched. Growth of each xenograft was monitored by externally measuring tumors in three dimensions using digital calipers 3 times per week. Xenograft volume (V) was determined by the following equation: V = (L × W2) × 0.5, where L is the length and W is the width of a xenograft.
Gamma Irradiation (IR) Therapy
Eighty tumor-bearing mice received a cumulative dose of 800 cGy of IR (200 cGy each third day for 4 cycles). The 200 cGy/day dosage is based on the results from a preliminary dose response study using the same mouse model and on the fact that 180 to 250 cGy/day is the commonly used acute dose for radiotherapy of humans. About 815 cGy whole body radiation is the LD50/30 for mice (50% lethal within 30 days). Mice were irradiated in a 137Cs Gamma Cell-40 Irradiator (Atomic Energy of Canada) facility in our Department of Radiology. Dosimetric analyses for this instrument is performed monthly for calculation of a precise 200 cGy exposure. Mice were transferred to a circular cage with individual compartments for each mouse during irradiation. Mice were euthanized at one day or 23 days after the last radiation treatment. The peripheral blood, duodenum, lungs, liver, spleen and tumor were harvested for analyses and the carcasses frozen in liquid nitrogen and stored at -20°C for later analysis of metastasis.
Magnetic Field Therapy
A therapeutic electromagnetic field (TEMF) system having a proprietary signal devised by EMF Therapeutics, Inc. (Chattanooga, TN, USA) was used. The system generates a pulsating half sinewave magnetic field with a frequency of 120 pulses per second [1]. An ellipsoidal coil with 21" major axis and 14" minor axis was used to generate the magnetic field. In the experiment reported here, the magnetic flux density measured in the exposure chamber was 15 mT. This value of magnetic field flux density was chosen based upon our previous experience and data [1]. A thorough 3-D mapping of the magnetic filed was performed for the entire space covered by the coil. The flux density of the magnetic field in the exposure chamber (25 cm long, 10 cm wide and 13 cm high) was consistent within the entire volume of the chamber. The walls of the exposure chamber were perforated to allow air exchange and to limit the temperature change inside the exposure chamber during the TEMF treatment was less than 1°C.
Euthanasia and Tissue Handling
Mice were deeply anesthetized using a ketamine/rompun mixture prepared by the UTHSCSA Laboratory Animal veterinarian, cervically dislocated, then were exsanguinated by cardiac puncture. Blood was collected into an EDTA containing microtube for complete blood counts.
The tumor, duodenum, spleen and liver were removed at the time of euthanasia while the carcass (including the lungs) was frozen and stored at -20°F for later analyses. Sections of the duodenum, the tumor, and a lobe of liver were fixed in Omni Fix II (Mt. Vernon N.Y.) and paraffin embedded. Embedded tissues were cut 4 μm or 8 μm thick, cut sections were placed on microscope slides then deparafinized and stained with hematoxylin and eosin or with periodic acid-Schiff (PAS) for morphological analysis. Additional sections were prepared for immunohistochemistry.
Metastasis studies
Lungs were removed from the thawed carcass to examine any spontaneous metastasis. The lungs were placed on a 25 × 75 mm microscope slide then smashed with a second slide to an area of about 100 mm2. The smashed lungs were secured in the flattened condition by rapping Magic tape (Scotch) around each end of the smashed lung preparation. The GFP-expressing metastatic cancer cell colonies, if any, were identified and counted using a Zeiss fluorescence microscope (TE-200) with a 3.5× and a 10× objective. GFP micrometastatic sites with three or more cells were scored as positive. No large metastatic sites were observed in the lungs.
Tumor Vascularity
To determine the effect of treatment on tumor angiogenesis, we measured the vascularity of excised tumors. Tumor tissues were fixed and imbedded in paraffin. Mid tumor sections of 8 μm were cut from the embedded tissue and stained with periodic acid Schiff (PAS). Sections were examined by light microscopy. CD31 immunostaining for mouse blood vessels was performed by incubating tumor sections with a rat antimouse CD-31 (PECAM-1) monoclonal antibody (PharMingen) at 5 μg/ml for 30 min at 37°C. Sections were then incubated with a biotin-labeled goat antirat IgG (Zymed; 1:200 dilution) for 30 min at room temperature, followed by ABC reagent kit (Vector Laboratories) for 30 min at room temperature. Color reaction was performed with 3, 3'-diaminobenzidine (Vector Laboratories) and counterstained with hematoxylin. Hypoxia-inducible factor-1 alpha immunohistochemistry was done following the instruction for antigen retrieval (Biogenex protocol) and iso-IHC (inno-Genex Mouse-on-mouse iso-IHC kit) with these changes: Dewaxing was with 3 changes of xylene, 10 minutes/change. Rehydration in 100%, 90%, and 70% ethanol, 10 minutes each. Initial dilution of the antibody was 1:200. Antibody: Stressgen anti-HIF-1 alpha, product # OSA-601. All sections were coded, then treated as above and the extent of blood vessels, endothelial cell pseudopods and total area volume density was scored using an ocular grid. The number of grid line intercepts over blood vessels and endothelial pseudopods gave a measure of the total volume density of these structures.
Assays Performed
Complete and differential blood counts
A blood cell counter with veterinary pack was used for counts of red cells, white cells and platelets in EDTA anticoagulated blood of the mice. Our Laboratory Animal Resources division performed this test.
Peripheral blood micronuclei
At least 1000 acridine orange stained cells were counted and the percentage of erythrocytes containing micronuclei was determined [3].
Histological analyses of duodenum
Fixed specimens of duodenum were trimmed, processed and oriented for paraffin embedding. Four μm thick sections of the paraffin blocks were mounted on slides. Complete midaxially sectioned crypts on H&E stained slides were selected for analyses. Complete crypts were defined as those with: 1) the crypt base at the muscularis mucosa, 2) an open lumen from mouth to base and 3) a single column of epithelial cells up each side of the crypt. The numbers of metaphase figures per midaxial crypt section was counted for 10 crypts of each mouse.
Statistical analyses
SAS computer software was used for statistical analyses. Tests for normality (basic statistics) was used on each data set. One way and two way analyses of variance followed by Student-Newman-Keuls multiple range tests, as appropriate, was used to determine if there were statistically significant (p ≤ 0.05) differences in any measured parameter due to the therapies. Tumor volume change for each tumor was determined using least squares linear regression analysis of tumor volume using Prism™ (Graphpad Software, Inc.). Differences in the proportions of mice with lung metastatic sites were assessed using the Fischer exact test analyses.
The abbreviations used
GFP, green fluorescent protein; TEMF, therapeutic electromagnetic field; PAS, periodic acid Schiff; HIF, hypoxia inducible factor 1- alpha; CD-31, monoclonal antibody against endothelial cells; IR, ionizing irradiation
Conflict of Interest
The author(s) declare that they have no competing interests.
Acknowledgements
This work was supported by EMF Therapeutics, Inc. and NIH grant CA 75253.
==== Refs
William CD Markov MS Hardman WE Cameron IL Therapeutic electromagnetic field effects on angiogenesis and tumor growth Anticancer Res 2001 21 3887 3892 11911264
Bandyopadhyay A López-Casillas F Malik SN Montiel JL Mendoz V Yang J Sun LZ Antitumor activity of a recombinant soluble betaglycan in human cancer xenograft Can Res 2002 62 4690 4695
Vijayalaxmi Frei MR Dusch SJ Gue V Meltz ML Jauchem JR Frequency of micronuclei in the peripheral blood and bone marrow of cancer-prone mice chronically exposed to 2450 MHz radiofrequency radiation Radiat Res 1997 147 495 500 9092931
Feraud O Cao Y Vittet D Embryonic stem cell-derived embryoid bodies development in collagen gels recapitulates sprouting angiogenesis Lab Invest 2001 81 1669 1681 11742037
Yang M Li L Jiang P Moossam AR Penman S Hoffman RM Dual-color fluorescence imaging distinguishes tumor cells from induced host angiogenic vessels and stromal cells Proc Natl Acad Sci USA 2003 100 14259 14262 14614130 10.1073/pnas.2436101100
Owen JB Coia LR Hanks Recent patterns of growth in radiation therapy facilities in the United States: a pattern of care study report Int J Radiat Onc Bio Phys 1992 24 983 986
Wachsberger P Burd R Dicken AP Tumor response to ionizing radiation combined with antiangiogenesis or vascular targeting agents: exploring mechanisms of interaction Clin Can Res 2003 9 1957 1971
Wachsberger P Burd R Dicken AP Ionizing radiation and anti-vascular therapy: exploring mechanisms of tumor response Clin Can Res
Gorski DH Beckett MA Jasowiak NT Calvin DP Mauceri HJ Salloum RM Seetharam S Koons A Hari DM Kufe DW Weichselbaum RR Blockade of the vascular endothelial growth factor stress response increases antitumor effects of ionizing radiation Can Res 1999 59 3374 3378
Weidner N New paradigm for vessel intravasation by tumor cells Amer J Path 2002 160 1937 1939 12057896
Inoue K Chikayawa M Fukatas Yoshikawa C Shuin T Frequent administration of angiogenesis inhibitor TNP-470 (AGM-1470) at an optimal biological dose inhibits tumor growth and metastasis of metastatic human transitional cell carcinoma in the urinary bladder Clin Can Res 2002 8 2389 2398
Crepin M DiBenedetto M Bagheri-Yarmand R Starzec A Vassy R Perret G Prevention of breast tumor angiogenesis and metastasis by cytostatic molecules in relevant mouse models Breast Can Res 2003 5 52 10.1186/bcr711
Steeg PS Angiogenesis inhibitors: motivators of metastasis? Nat Med 2003 9 822 823 12835693 10.1038/nm0703-822
Kieran MW Folkman J Haymach J Angiogenesis inhibitors and hypoxia Nat Med 2003 9 1104 12949515 10.1038/nm0903-1104a
Ribatti D Vacca A De Falco G Roccaro A Roncali L Dammacco F Angiogenesis, angiogenic factors expression and hematological malignances Anticancer Res 2001 21 4333 4340 11908688
Jaeger TM Weidner N Chew K Moore DH Kerschmann RL Waldman FM Carroll PR Tumor angiogenesis correlates with lymph node metastases in invasive bladder cancer J Urol 1995 154 69 71 7539869 10.1097/00005392-199507000-00026
Weidner N Semple JP Welch WR Tumor angiogenesis and metastasis-correlation in invasive breast cancer carcinoma N Eng J Med 1991 324 1 8
Leek RD The prognostic role of angiogenesis in breast cancer Anticancer Res 2001 21 4325 4332 11908687
|
16045802
|
PMC1190196
|
CC BY
|
2021-01-04 16:24:16
|
no
|
Cancer Cell Int. 2005 Jul 26; 5:23
|
utf-8
|
Cancer Cell Int
| 2,005 |
10.1186/1475-2867-5-23
|
oa_comm
|
==== Front
Cancer Cell IntCancer Cell International1475-2867BioMed Central London 1475-2867-5-231604580210.1186/1475-2867-5-23Primary ResearchTherapeutic Electromagnetic Field (TEMF) and gamma irradiation on human breast cancer xenograft growth, angiogenesis and metastasis Cameron Ivan L [email protected] Lu-Zhe [email protected] Nicholas [email protected] W Elaine [email protected] C Douglas [email protected] Department of Cellular and Structural Biology, University of Texas Health Science Center at San Antonio, 7703 Floyd Curl Drive, San Antonio, Texas 78229, USA2 Pennington Biomedical Research Center, Louisiana State University, 6400 Perkins Road, Baton Rouge, Louisiana 70808, USA3 EMF Therapeutics, Inc., P.O. Box 679, Signal Mountain, Tennessee 37377, USA2005 26 7 2005 5 23 23 5 7 2004 26 7 2005 Copyright © 2005 Cameron et al; licensee BioMed Central Ltd.2005Cameron et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The effects of a rectified semi-sinewave signal (15 mT amplitude, 120 pulses per second, EMF Therapeutics, Inc.) (TEMF) alone and in combination with gamma irradiation (IR) therapy in nude mice bearing a human MDA MB231 breast cancer xenograft were tested. Green fluorescence protein transfected cancer cells were injected into the mammary fat pad of young female mice. Six weeks later, mice were randomly divided into four treatment groups: untreated controls; 10 minute daily TEMF; 200 cGy of IR every other day (total 800 cGy); IR plus daily TEMF. Some mice in each group were euthanized 24 hours after the end of IR. TEMF treatment continued for 3 additional weeks. Tumor sections were stained for: endothelial cells with CD31 and PAS or hypoxia inducible factor 1α (HIF).
Results
Most tumors <35 mm3 were white but tumors >35 mm3 were pink and had a vascularized capsule. The cortex within 100 microns of the capsule had little vascularization. Blood vessels, capillaries, and endothelial pseudopods were found at >100 microns from the capsule (subcortex). Tumors >35 mm3 treated with IR 24 hours previously or with TEMF had decreased blood vessels in the subcortex and more endothelial pseudopods projecting into hypoxic, HIF positive areas than tumors from the control group. Mice that received either IR or TEMF had significantly fewer lung metastatic sites and slower tumor growth than did untreated mice. No harmful side effects were attributed to TEMF.
Conclusion
TEMF therapy provided a safe means for retarding tumor vascularization, growth and metastasis.
electromagnetic fieldbreast cancerionizing irradiationangiogenesismetastasis
==== Body
Background
In a previously published experimental research report, it was found that exposing a transplantable murine mammary adenocarcinoma to a 15 mT EMF given at 120 pulses per second for 10 minutes per day significantly reduced tumor growth and vascularization and resulted in an increased survival time [1]. This published report appears to be the only literature available on the use of pulsating magnetic fields to reduce tumor angiogenesis. The authors of this report suggested that the magnetic field treatment used acted to reduce tumor angiogenesis and might have value as an alternative therapeutic modality for treatment of patients with tumors. The study reported here was designed to further investigate the potential of the same EMF therapy to inhibit growth and angiogenesis of a human breast cancer xenograft and to compare the effects of: 1) a commonly used course of gamma irradiation (IR) involving exposure to 200 cGy every second day for a total of 800 cGy, 2) daily exposure to TEMF, and 3) a combination of these two therapeutic treatment regimens on tumor growth, tumor angiogenesis, tumor metastasis, and of the side effects of each treatment regimen. Although this study used whole body IR therapy, most IR therapy of human patients is restricted to localized targeted regions of the body to avoid general side effects of IR treatment. The MDA MB231 cancer cell line transfected with and expressing a green fluorescent protein (GFP) gene was used to facilitate study of metastases of cancer cells from the site of the primary tumor [2].
Our study results demonstrate the potential of TEMF therapy to retard tumor: growth, angiogenesis, and metastasis, without harmful side effects.
Results
Body Weight
Once the mice were divided into treatment groups the body weight of each mouse was measured every 3 to 4 days for the remainder of the experiment. As illustrated in Fig. 1, the two groups that received IR therapy every second day for 8 days demonstrated a mean body weight loss beginning during IR therapy and lasting until about 8 or 9 days after the end of IR therapy. After completion of the IR therapy, the irradiated mice again began to regain their weight toward the mean weight of the two groups of mice not subjected to IR therapy. The group of mice that received only EMF therapy demonstrated a continuous increase in mean body weight similar to the group of mice given no therapy.
Figure 1 Body weights during the course of the experiment. The two groups of mice that received gamma irradiation both lost body weight during and for a few days after the course of exposure, but the body weights later recovered towards the weights of the two groups of mice not exposed to gamma irradiation.
Tumor Growth
Fig. 2 illustrates mean tumor volume change for each of the four treatment groups starting at the beginning of IR and/or EMF therapy. All tumors in each therapy group were less than 35 mm3 at the start of treatment period. To statistically assess tumor growth rate, the data on each tumor in each group of mice was subjected to linear regression analysis. Tumor volume gave a good fit to a linear regression model. The slope (growth rate) derived from the linear regression of each tumor volume was used to determine any statistical differences in growth rates between treatment groups (Fig. 2). Growth rate of tumors from the untreated group was significantly faster (p < 0.001) than any of the three groups of treated mice. The tumor growth rate of the group of mice treated solely with EMF therapy was significantly less than the tumor growth rate of the untreated group. The tumor growth rates of the two groups of mice treated with gamma irradiation were significantly less than the tumor growth rates of the untreated or TEMF only treated groups.
Figure 2 Illustrates the effect on tumor growth of: the 8-day course of gamma irradiation therapy (IR), the daily exposure to the TEMF therapy (TEMF), a combination of gamma irradiation and TEMF therapy (TEMF/IR), and no treatment (Control). The tumors were all 35 mm3 or less at the start of treatment. The growth rate is the slope of the linear regression of the tumor volumes at each time. It can be seen that tumor growth was beginning to deviate from the linear fit about 18 days after the start of TEMF or radiation treatments however there was still a reasonable linear fit. ANOVA was used to determine any statistical differences in growth rates between treatment groups. Different letters demonstrate significant differences. Gamma irradiation alone (IR) or in combination with TEMF (TEMF/IR) reduced the tumor growth. The TEMF therapy resulted in a slower tumor growth rate than the untreated controls but not as slow as the gamma irradiated mice.
Results of an evaluation of the tumor growth rates from day 7 though day 15 post IR treatment are reported in Fig. 3. As illustrated in Fig. 3, the mean tumor growth rate of mice given neither IR nor TEMF treatment was significantly higher than the three groups given IR and TEMF therapy either alone or in combination. The group of mice given both IR and TEMF followed by daily TEMF had a mean tumor growth rate significantly lower than that of the other three groups. Clearly, the continued application of daily TEMF therapy following IR therapy had an additive inhibitory effect on the tumor growth rate during the 7 to 15 days after the course of IR therapy.
Figure 3 Mean ± SEM tumor growth rate data from day 7 to day 15 post IR therapy in each treatment group. ANOVA was used to determine any statistical differences in growth rates between treatment groups. Different letters demonstrate significant differences. Tumors in the untreated control group grew significantly faster than did tumors in the other three groups. Tumors in the IR/TEMF group had a growth rate significantly lower than that of the other three groups.
Tumor vasculature
While making measurements on tumor volume, tumor color was observed through the skin of the nude mice. After recording tumor color and volume of each mouse, it became clear that smaller tumors were white while larger tumors were pink. Fig. 4 illustrates the relationship between tumor color and tumor volume. Almost all tumors less than 35 mm3 were white whereas tumors greater than approximately 35 mm3 were pink in color. Application of the pressure from one's finger to a pink tumor followed by the rapid removal of the finger pressure showed a rapid color change from white to pink.
Figure 4 Relationship between color of the tumor as observed through the skin and the measured tumor volume. Tumors <35 mm3 were mostly white while tumors >35 mm3 were pink.
Histological examination of midsections of PAS stained tumors was done to assess tumor vascularization patterns. Tumors with a volume >35 mm3 demonstrated a connective tissue capsule with blood vessels (Fig. 5A). At higher magnification the cortical area within about 100 μm of the capsule revealed few blood vessels while the area greater than 100 μm from the capsule showed evidence of considerable blood vessel and capillaries with many endothelial pseudopods extending away from the capillaries (Fig. 5B&5C). The general direction of the pseudopods was parallel to the tumor capsule surface. Immunohistochemical localization of CD-31, used as a specific marker of endothelium, demonstrated a positive reaction of pseudopods (Fig. 5D). Areas of tumor necrosis were observed, below the subcortical area (Fig. 5E&5F). Immunohistochemical localization of hypoxia-inducible factor 1-α (HIF) reveals the subcortical area of the tumor to contain HIF positive cells while the tumor capsule, cortex and necrotic areas of the tumor demonstrate no evidence of HIF. Thus, the area found to be HIF positive were enriched in endothelial pseudopods.
Figure 5 Photomicrographs illustrating the pattern of the tumor vascular network (A and B), of endothelial pseudopods stained with PAS (C) and with CD-31 specific endothelial markers (D), and of the localization of HIF-α (E and F). A, The tumor capsule (left) reveals blood vessels. The cortex under the capsule reveals no blood vessels and few endothelial pseudopods while the subcortical area to the right has more pseudopods. B, The subcortical area of the tumor reveals a small blood capillary with multiple endothelial pseudopods protruding at right angles into the tumor mass. C, At high magnification endothelial pseudopods are seen to branch and occasionally have a vacuole/lumen (arrow). D, The endothelial pseudopods react positively to the CD-31 specific endothelial marker. The CD-31 antibody identifies endothelial cells using the avidin-biotin peroxidase complex method. E, Viable cell area can be seen beneath tumor capsule (left). Necrotic area can be seen to the right. F, Enlarged subcortical area from E. In E the hypoxic area between the viable and the necrotic tissue is stained brown.
Ocular grid intercept couting was used to quantify tumor vascularization in 8 μm thick PAS stained histological sections of the tumors. The numbers of ocular grid intercepts were scored over the area of: blood vessels and capillaries, endothelial pseudopods and over the area with no indication of these structures. This method has been shown to be a usable measure of volume density occupied by recognizable structures. The results of the scoring of blood vessels and of endothelial pseudopods in the subcortical regions of the tumors are summarized in Fig. 6A&6B. Statistical analysis of the mean blood vessel volume density versus the pseudopod volume density reveals a significant exponential correlation coefficient of 0.966. Thus, as the blood vessel volume density decreased, the endothelial pseudopod volume density increased. As illustrated in Fig. 6A&6B, IR resulted in significantly reduced blood vessel volume density but a significantly increased volume density of pseudopods at one day after the last dose of irradiation, but the blood vessel volume density increased and the pseudopod volume density decreased to the level of the untreated control by 22 days after the last dose of irradiation. However, the groups of mice treated solely with EMF had significantly less blood vessel volume density and a significantly higher pseudopod volume density at one day after the end of IR. The low blood vessel volume density and the high pseudopods volume density in the tumors of the EMF treated mice remained the same at 22 days after the end of the IR treatment.
Figure 6 Quantification of changes in tumor vascularization between treatment groups. A and B, The percent of areas (volume density) of blood vessels and the percent of area of endothelial pseudopods were determined using an ocular grid intercept counting method. The mean ± SEM of each treatment group is graphed. Columns that do not share a common letter within a graph are significantly different (p < 0.01). The data indicate that gamma irradiation and EMF alone or in combination decreased the total area of blood vessels and increased the total area of pseudopods compared to the control. Data from the untreated mice (Control) and the EMF treated mice (TEMF) were pooled from mice from the early and from the late sacrifice because there was no significant time of sacrifice difference within these groups.
Metastasis
The number of lung metastatic colonies per mouse in each of the four treatment groups is summarized in Fig. 7. Both the mean number of colonies in the lungs per mouse as well as the incidence of metastasis was significantly higher in the untreated group than in the three groups treated with IR and/or EMF therapy. There were no significant differences between the three treatment groups in number or in incidence of lung metastatic colonies per mouse.
Figure 7 Metastasis of GFP-expressing human breast MDA MB231 cells from the site of inoculation in the inguinal mammary fat pad to lodge in the lungs of mice in each of the four treatment groups. The lungs were removed and smashed between microscopic slides and microcolonies of GFP cells were observed by epilumination using blue light and detected by presence of green fluorescence microcolonies observed using both 3.5× and 10× objective lens. The untreated mice had a significantly higher number of GFP positive microcolonies than the groups of mice that received gamma irradiation therapy or those mice that received EMF therapy. There were no other significant differences between groups in either mean number of microcolonies per lung or in incidence. Horizontal lines indicate mean values for each treatment group.
Side Effects of Gamma Irradiation Therapy and of TEMF Therapy
Mice were euthanized at both one day and at 22 days after the last IR exposure. Data on liver and spleen weights are summarized in Table 1, and data on blood counts are summarized in Table 2. There were no statistically significant differences between treatment groups in mean liver weights at the earlier or later kills. On the other hand, the spleen weights of mice that were euthanized one day after their last exposure to IR were significantly less than the spleen weights of the non-gamma irradiated groups. The mean spleen weights of the IR mice recovered to the level of the untreated groups by 22 days after their last IR exposure. There was no significant loss of spleen weight at the time of early or at the time of late euthanasia due to TEMF therapy.
Table 1 Gamma Irradiation Therapy (IR) and Therapeutic Electromagnetic Field Therapy (TEMF) on Liver and Spleen Weight (Means ± SEM)
Therapy Group n Liver weight (grams) Spleen weight (grams)
Early1
Control 5 0.94 ± 0.04 0.107 ± 0.009
IR 5 0.87 ± 0.07 0.031 ± 0.004
TEMF 5 1.04 ± 0.05 0.110 ± 0.006
TEMF/IR 5 0.93 ± 0.04 0.033 ± 0.004
Late2
Control 9 1.26 ± 0.07 0.166 ± 0.008
IR 20 1.17 ± 0.03 0.146 ± 0.008
TEMF 10 1.16 ± 0.05 0.145 ± 0.012
TEMF/IR 10 1.15 ± 0.05 0.151 ± 0.014
1Mice sacrificed one day after last irradiation treatment.
2Mice sacrificed 22 days after last irradiation treatment.
Results of Statistical Analyses:
1. Liver weight: No significant differences among treatment groups.
2. Spleen weight: At early kill, groups treated with gamma irradiation show significant (p < 0.01) loss of weight compared to groups not treated with gamma irradiation. In late kill, no significant differences were detected.
Table 2 Gamma Irradiation Therapy (IR) and Electromagnetic Field Therapy (TEMF) on Blood Counts(Means ± SEM)
Therapy Group n WBC × 103/μL RBC × 106/μL Platelets × 103/μL Micronuclei (%RBC)
Early1
Control 4 2.91 ± 0.83 9.08 ± 0.07 584 ± 76 1.3 ± 0.04
IR 4 0.11 ± 0.01 7.50 ± 0.16 389 ± 17 1.2 ± 0.03
TEMF 5 2.89 ± 0.60 9.05 ± 0.12 505 ± 61 1.3 ± 0.04
TEMF/IR 4 0.23 ± 0.06 7.68 ± 0.13 443 ± 34 1.2 ± 0.03
Late2
Control 8 1.85 ± 0.37 8.30 ± 0.14 551 ± 83 1.6 ± 0.06
IR 13 1.48 ± 0.332 7.06 ± 0.16 973 ± 98 0.7 ± 0.01
TEMF 11 1.15 ± 0.29 8.85 ± 0.22 613 ± 62 0.7 ± 0.01
TEMF/IR 10 1.07 ± 0.13 6.58 ± 0.11 1007 ± 141 0.6 ± 0.01
1Mice sacrificed one day after last irradiation treatment.
2Mice sacrificed 22 days after last irradiation treatment.
Results of Statistical Analyses:
1. At the end of irradiation therapy plus one day, gamma irradiation caused significant decreases in WBC, RBC, and platelet counts. No other significant differences at p < 0.01.
2. At the end of irradiation therapy plus 22 days, gamma irradiation caused significant decreases in WBC and RBC counts and a significant increase in platelet counts. No other significant differences at p < 0.01.
There was a significant decrease in WBC, RBC, and platelets counts attributed to gamma irradiation exposure at 1 day after the end of radiation treatment (Table 2). There was no significant difference in WBC, RBC, and platelets counts attributed to TEMF treatment. At 22 days after the end of the IR treatment, the WBC and RBC counts in the two IR groups were not quite as low but remained significantly lower than in the two non-irradiated groups while the IR groups had significantly higher platelet counts than any of the other groups. Clearly, platelet counts have made a significant compensatory rebound by 22 days after the end of IR therapy. Statistical analysis of micronuclei counts indicated that there was no significant genotoxic damage due to either treatment at either time of euthanasia.
Another measure of possible side effects was the scoring of mitotic activity in the duodenal crypts. Fig. 8 summarizes results of mitotic activity in the duodenal crypts of mice sacrificed one day after the end of the IR therapy regimen. The two groups of mice treated with IR 24 hours previously had significantly fewer metaphase figures per crypt than the non-treated group or to the group of mice treated solely with the course of TEMF therapy.
Figure 8 Number of metaphase figures per midaxial histological section of duodenal crypts. Column height indicates mean ± SEM. The data are from the mice sacrificed one day after the last gamma irradiation exposure. The columns that do not share a common letter are significantly different. Gamma irradiation but not EMF decreased the number of metaphase figures in the duodenal crypts.
All mice treated with IR demonstrated a tan skin discoloration immediately after the 8-day course of IR therapy, but the tan skin color returned to the normal skin color by 22 days after the course of IR therapy was terminated. The skin color in the two non-irradiated groups of mice remained normal in appearance throughout the course of the study.
Discussion
On tumor vascularization
The immunohistochemical localization of hypoxia induced factor (HIF) in the tumors helped explain the vascular pattern of the encapsulated tumors. The HIF positive reactivity in the tumor was localized to the subcortical areas of the tumor in the same area found to have the most blood vessels, capillaries and pseudopods. These observations suggest that this subcortical area of the tumor is hypoxic leading to production of angiogenesis growth factors, which in turn acts on the host endothelial cells to sprout pseudopods. Sprouting of cultured endothelial cells in response to vascular endothelial growth factors is known to occur [4]. Feraud et al. also reported that this endothelial cell sprouting could be inhibited by several angiostatic agents. The presence of pseudopodal like structures in viable areas of a tumor immediately adjacent to necrotic areas of the tumor has been previously reported [5]. The lack of HIF reactivity in the cortical area beneath the well vascularized tumor capsule suggest this cortical area is not hypoxic and not in need of an extensive vasculature. It appears that the PSA staining of endothelial cell sprouting pseudopods in tumors gives quantifiable spatial information on where in the tumor that HIF and vascular endothelial growth factors are located.
Action of IR and of EMF therapies
EMF therapy has been reported to slow tumor growth and tumor angiogenesis and to increase the survival time of tumor bearing mice [1]. An experiment was therefore designed to test if the combining of IR therapy and EMF therapy in a tumor bearing mouse model might give an addictive therapeutic index.
Over half of all cancer patients are given ionizing radiation (IR) during their course of treatment [6]. The interactions between IR and tumor vascularization are complex [7,8]. It has been shown that IR therapy is dependent on tumor vascularization and that IR therapy initially interferes with tumor vascularization leading to tumor hypoxia [9]. However hypoxic areas are radio-resistant until they can induce angiogenesis and re-oxygenation. Hypoxic areas of tumors produce HIF leading to production of angiogenesis growth factors, which stimulate angiogenesis to allow further tumor growth. This known sequence of events following an IR treatment of a tumorous patient has lead to the combined use of IR therapy coupled to use of an anti-angiogenesis agent. By blocking angiogenic factors, like VEGF, and by blocking the formation of new vessels in combination with IR therapy researchers have dramatically increased the efficacy of IR therapy in a number of tumor types [7-9]. The use of anti-angiogenic agents has proved useful in the treatment of cancer progression in most preclinical and clinical trials [10,11]. However the prevention of metastasis (Inoue et al.) by anti-angiogenic agents [11,12] has not gone unchallenged [[13] and see rebuttal by Kieran et al. [14]].
The experimental results indicate that either IR therapy or EMF therapy suppressed tumor growth. The IR therapy proved to be more effective at suppressing tumor growth than did the EMF therapy while the IR therapy but not the EMF therapy had harmful side effects. Both IR therapy and EMF therapy reduced blood vessel volume density within the tumor when measured one day after the end of the IR therapy but blood vessel volume density in tumors of the IR treated mice returned to pre IR treatment levels. The continual use of daily EMF therapy in the IR treated mice did suppress the return of blood vessel volume density within the tumor following IR therapy. The combination of both therapeutic modalities therefore led to a sustained and significant reduction in the extent of tumor blood vessel volume density.
That pseudopod volume density was relatively high at 22 day after IR in the group of mice that were on sustained daily EMF therapy suggests that major areas of these tumors were anoxic and thus were producing HIF and VEGF's resulting in the sprouting of the endothelial cell pseudopods. This indicates that these early events in the angiogenesis process were not suppressed by the continued use of EMF therapy. Because continued use of EMF did suppress tumor blood vessel volume density at 22 days after IR therapy the EMF therapy may be acting to suppress one of more of the later steps in the process of tumor vascularization such as: endothelial cell proliferation, lumen formation, generation of vascular tubes and initiation of blood flow [15].
As illustrated in Fig. 6A, the continued daily use of TEMF following the course of IR therapy suppressed tumor blood vessel volume density compared to mice given the IR therapy alone. The data on tumor growth rate, beginning a week after IR therapy, revealed that the continuation of TEMF therapy suppressed the mean tumor growth rate of the tumors compared to those of mice treated only with IR therapy. In fact, linear regression analysis of the data indicates that the slope of the tumor growth rate of the IR plus TEMF treated mice did not differ from a slope of zero. Thus, these tumor growth rate results indicate a significant treatment advantage for continuous daily TEMF therapy following a standard course of IR therapy.
Since the extent of tumor vascularization has been positively linked to the likelihood of tumor cell metastasis [12,16-18] it was anticipated that a decrease in the blood vessel volume density in the tumor, as observed in the experiments reported herein, would decrease the chance for tumor cells to metastasize to distant sites in the tumor bearing mouse model used in the study. Inspection of the incidence of metastasis observed in the lungs of human breast cancer xenographs revealed: that IR therapy alone reduced metastasis by 49%, that EMF therapy alone reduced metastasis by 54% while the combination of IR and EMF therapies when followed by daily EMF therapy resulted in a 73% reduction in incidence of metastasis to the lung. Although all three of these therapy regimens caused a significant decrease in incidence of metastasis of tumor cells to the lungs the greatest decrease, although not statistically significant, was with the combination of IR and EMF therapies.
In summary the results indicate that sustained EMF therapy by itself reduced: the extent of tumor blood vessel volume density, the tumor growth rate and tumor cell metastasis to the lung, without harmful side effects. TEMF therapy alone may therefore be a useful alternate therapy for cancer patients who decide not to undergo standard radiation or for chemotherapy. The IR therapy alone markedly reduced: the tumor growth rate, reduced tumor metastasis but only transitly reduced tumor blood vessel volume density. The IR therapy did however have harmful side effects. These side effects could be minimized by targeted IR exposure. The combination of IR therapy and continued TEMF therapy markedly suppressed the return of blood vessel volume density and tumor growth following IR therapy and resulted in the lowest incidence of tumor metastasis. Thus these findings support the ideas that TEMF of the type used in this experiment may be an effective antiangiogenic therapy and that a combination of an anti-vascular modality, such as TEMF, with IR therapy may enhance the therapeutic index of IR therapy for cancer.
Materials and methods
Animals
Female athymic nude mice (purchased from the Harlan Sprague Dawley, Inc., Indianapolis, IN), 6 weeks of age, were used for this study. The animals were housed under pathogen-free conditions and fed an AIN-76 semipurified diet slightly altered to contain 10% w/w corn oil.
Cell Lines
Human breast cancer cell line MDA-MB-231 was obtained from the American Type Culture Collection. To determine the effect of the treatments on the metastatic potential of the MDA-MB-231 cells they were stably transfected with the enhanced GFP expression plasmid, pEGFP-N1 (Clontech Laboratories, Inc.). The expression of GFP allowed for detection of micrometastatic colonies on the whole flattened lungs with a fluorescence epilumination microscope. The cell lines were cultured in McCoys's 5A medium supplemented with pyruvate, vitamins, amino acids, antibiotics, and 10% fetal bovine serum [2]. HDMECs and the culture medium EGM-2 Mv were maintained at 37°C in a humidified incubator with 5% CO2.
In Vivo Tumor Growth
The MDA-MB-231/GFP cells were harvested from exponential cultures and inoculated at 2 × 106 cells/inoculum in the inguinal mammary fat pad area of female athymic nude mice, 6 weeks of age. When the tumors grew to an average diameter of about 3 mm after 5 weeks, the mice were divided into 4 groups such that the mean and median of tumor volume of the four groups were closely matched. Growth of each xenograft was monitored by externally measuring tumors in three dimensions using digital calipers 3 times per week. Xenograft volume (V) was determined by the following equation: V = (L × W2) × 0.5, where L is the length and W is the width of a xenograft.
Gamma Irradiation (IR) Therapy
Eighty tumor-bearing mice received a cumulative dose of 800 cGy of IR (200 cGy each third day for 4 cycles). The 200 cGy/day dosage is based on the results from a preliminary dose response study using the same mouse model and on the fact that 180 to 250 cGy/day is the commonly used acute dose for radiotherapy of humans. About 815 cGy whole body radiation is the LD50/30 for mice (50% lethal within 30 days). Mice were irradiated in a 137Cs Gamma Cell-40 Irradiator (Atomic Energy of Canada) facility in our Department of Radiology. Dosimetric analyses for this instrument is performed monthly for calculation of a precise 200 cGy exposure. Mice were transferred to a circular cage with individual compartments for each mouse during irradiation. Mice were euthanized at one day or 23 days after the last radiation treatment. The peripheral blood, duodenum, lungs, liver, spleen and tumor were harvested for analyses and the carcasses frozen in liquid nitrogen and stored at -20°C for later analysis of metastasis.
Magnetic Field Therapy
A therapeutic electromagnetic field (TEMF) system having a proprietary signal devised by EMF Therapeutics, Inc. (Chattanooga, TN, USA) was used. The system generates a pulsating half sinewave magnetic field with a frequency of 120 pulses per second [1]. An ellipsoidal coil with 21" major axis and 14" minor axis was used to generate the magnetic field. In the experiment reported here, the magnetic flux density measured in the exposure chamber was 15 mT. This value of magnetic field flux density was chosen based upon our previous experience and data [1]. A thorough 3-D mapping of the magnetic filed was performed for the entire space covered by the coil. The flux density of the magnetic field in the exposure chamber (25 cm long, 10 cm wide and 13 cm high) was consistent within the entire volume of the chamber. The walls of the exposure chamber were perforated to allow air exchange and to limit the temperature change inside the exposure chamber during the TEMF treatment was less than 1°C.
Euthanasia and Tissue Handling
Mice were deeply anesthetized using a ketamine/rompun mixture prepared by the UTHSCSA Laboratory Animal veterinarian, cervically dislocated, then were exsanguinated by cardiac puncture. Blood was collected into an EDTA containing microtube for complete blood counts.
The tumor, duodenum, spleen and liver were removed at the time of euthanasia while the carcass (including the lungs) was frozen and stored at -20°F for later analyses. Sections of the duodenum, the tumor, and a lobe of liver were fixed in Omni Fix II (Mt. Vernon N.Y.) and paraffin embedded. Embedded tissues were cut 4 μm or 8 μm thick, cut sections were placed on microscope slides then deparafinized and stained with hematoxylin and eosin or with periodic acid-Schiff (PAS) for morphological analysis. Additional sections were prepared for immunohistochemistry.
Metastasis studies
Lungs were removed from the thawed carcass to examine any spontaneous metastasis. The lungs were placed on a 25 × 75 mm microscope slide then smashed with a second slide to an area of about 100 mm2. The smashed lungs were secured in the flattened condition by rapping Magic tape (Scotch) around each end of the smashed lung preparation. The GFP-expressing metastatic cancer cell colonies, if any, were identified and counted using a Zeiss fluorescence microscope (TE-200) with a 3.5× and a 10× objective. GFP micrometastatic sites with three or more cells were scored as positive. No large metastatic sites were observed in the lungs.
Tumor Vascularity
To determine the effect of treatment on tumor angiogenesis, we measured the vascularity of excised tumors. Tumor tissues were fixed and imbedded in paraffin. Mid tumor sections of 8 μm were cut from the embedded tissue and stained with periodic acid Schiff (PAS). Sections were examined by light microscopy. CD31 immunostaining for mouse blood vessels was performed by incubating tumor sections with a rat antimouse CD-31 (PECAM-1) monoclonal antibody (PharMingen) at 5 μg/ml for 30 min at 37°C. Sections were then incubated with a biotin-labeled goat antirat IgG (Zymed; 1:200 dilution) for 30 min at room temperature, followed by ABC reagent kit (Vector Laboratories) for 30 min at room temperature. Color reaction was performed with 3, 3'-diaminobenzidine (Vector Laboratories) and counterstained with hematoxylin. Hypoxia-inducible factor-1 alpha immunohistochemistry was done following the instruction for antigen retrieval (Biogenex protocol) and iso-IHC (inno-Genex Mouse-on-mouse iso-IHC kit) with these changes: Dewaxing was with 3 changes of xylene, 10 minutes/change. Rehydration in 100%, 90%, and 70% ethanol, 10 minutes each. Initial dilution of the antibody was 1:200. Antibody: Stressgen anti-HIF-1 alpha, product # OSA-601. All sections were coded, then treated as above and the extent of blood vessels, endothelial cell pseudopods and total area volume density was scored using an ocular grid. The number of grid line intercepts over blood vessels and endothelial pseudopods gave a measure of the total volume density of these structures.
Assays Performed
Complete and differential blood counts
A blood cell counter with veterinary pack was used for counts of red cells, white cells and platelets in EDTA anticoagulated blood of the mice. Our Laboratory Animal Resources division performed this test.
Peripheral blood micronuclei
At least 1000 acridine orange stained cells were counted and the percentage of erythrocytes containing micronuclei was determined [3].
Histological analyses of duodenum
Fixed specimens of duodenum were trimmed, processed and oriented for paraffin embedding. Four μm thick sections of the paraffin blocks were mounted on slides. Complete midaxially sectioned crypts on H&E stained slides were selected for analyses. Complete crypts were defined as those with: 1) the crypt base at the muscularis mucosa, 2) an open lumen from mouth to base and 3) a single column of epithelial cells up each side of the crypt. The numbers of metaphase figures per midaxial crypt section was counted for 10 crypts of each mouse.
Statistical analyses
SAS computer software was used for statistical analyses. Tests for normality (basic statistics) was used on each data set. One way and two way analyses of variance followed by Student-Newman-Keuls multiple range tests, as appropriate, was used to determine if there were statistically significant (p ≤ 0.05) differences in any measured parameter due to the therapies. Tumor volume change for each tumor was determined using least squares linear regression analysis of tumor volume using Prism™ (Graphpad Software, Inc.). Differences in the proportions of mice with lung metastatic sites were assessed using the Fischer exact test analyses.
The abbreviations used
GFP, green fluorescent protein; TEMF, therapeutic electromagnetic field; PAS, periodic acid Schiff; HIF, hypoxia inducible factor 1- alpha; CD-31, monoclonal antibody against endothelial cells; IR, ionizing irradiation
Conflict of Interest
The author(s) declare that they have no competing interests.
Acknowledgements
This work was supported by EMF Therapeutics, Inc. and NIH grant CA 75253.
==== Refs
William CD Markov MS Hardman WE Cameron IL Therapeutic electromagnetic field effects on angiogenesis and tumor growth Anticancer Res 2001 21 3887 3892 11911264
Bandyopadhyay A López-Casillas F Malik SN Montiel JL Mendoz V Yang J Sun LZ Antitumor activity of a recombinant soluble betaglycan in human cancer xenograft Can Res 2002 62 4690 4695
Vijayalaxmi Frei MR Dusch SJ Gue V Meltz ML Jauchem JR Frequency of micronuclei in the peripheral blood and bone marrow of cancer-prone mice chronically exposed to 2450 MHz radiofrequency radiation Radiat Res 1997 147 495 500 9092931
Feraud O Cao Y Vittet D Embryonic stem cell-derived embryoid bodies development in collagen gels recapitulates sprouting angiogenesis Lab Invest 2001 81 1669 1681 11742037
Yang M Li L Jiang P Moossam AR Penman S Hoffman RM Dual-color fluorescence imaging distinguishes tumor cells from induced host angiogenic vessels and stromal cells Proc Natl Acad Sci USA 2003 100 14259 14262 14614130 10.1073/pnas.2436101100
Owen JB Coia LR Hanks Recent patterns of growth in radiation therapy facilities in the United States: a pattern of care study report Int J Radiat Onc Bio Phys 1992 24 983 986
Wachsberger P Burd R Dicken AP Tumor response to ionizing radiation combined with antiangiogenesis or vascular targeting agents: exploring mechanisms of interaction Clin Can Res 2003 9 1957 1971
Wachsberger P Burd R Dicken AP Ionizing radiation and anti-vascular therapy: exploring mechanisms of tumor response Clin Can Res
Gorski DH Beckett MA Jasowiak NT Calvin DP Mauceri HJ Salloum RM Seetharam S Koons A Hari DM Kufe DW Weichselbaum RR Blockade of the vascular endothelial growth factor stress response increases antitumor effects of ionizing radiation Can Res 1999 59 3374 3378
Weidner N New paradigm for vessel intravasation by tumor cells Amer J Path 2002 160 1937 1939 12057896
Inoue K Chikayawa M Fukatas Yoshikawa C Shuin T Frequent administration of angiogenesis inhibitor TNP-470 (AGM-1470) at an optimal biological dose inhibits tumor growth and metastasis of metastatic human transitional cell carcinoma in the urinary bladder Clin Can Res 2002 8 2389 2398
Crepin M DiBenedetto M Bagheri-Yarmand R Starzec A Vassy R Perret G Prevention of breast tumor angiogenesis and metastasis by cytostatic molecules in relevant mouse models Breast Can Res 2003 5 52 10.1186/bcr711
Steeg PS Angiogenesis inhibitors: motivators of metastasis? Nat Med 2003 9 822 823 12835693 10.1038/nm0703-822
Kieran MW Folkman J Haymach J Angiogenesis inhibitors and hypoxia Nat Med 2003 9 1104 12949515 10.1038/nm0903-1104a
Ribatti D Vacca A De Falco G Roccaro A Roncali L Dammacco F Angiogenesis, angiogenic factors expression and hematological malignances Anticancer Res 2001 21 4333 4340 11908688
Jaeger TM Weidner N Chew K Moore DH Kerschmann RL Waldman FM Carroll PR Tumor angiogenesis correlates with lymph node metastases in invasive bladder cancer J Urol 1995 154 69 71 7539869 10.1097/00005392-199507000-00026
Weidner N Semple JP Welch WR Tumor angiogenesis and metastasis-correlation in invasive breast cancer carcinoma N Eng J Med 1991 324 1 8
Leek RD The prognostic role of angiogenesis in breast cancer Anticancer Res 2001 21 4325 4332 11908687
|
16098228
|
PMC1190197
|
CC BY
|
2021-01-04 16:38:31
|
no
|
Cardiovasc Ultrasound. 2005 Aug 12; 3:20
|
latin-1
|
Cardiovasc Ultrasound
| 2,005 |
10.1186/1476-7120-3-20
|
oa_comm
|
==== Front
Cerebrospinal Fluid ResCerebrospinal Fluid Research1743-8454BioMed Central London 1743-8454-2-51604580610.1186/1743-8454-2-5ResearchIn normal rat, intraventricularly administered insulin-like growth factor-1 is rapidly cleared from CSF with limited distribution into brain Nagaraja Tavarekere N [email protected] Padma [email protected] Martin [email protected] Peter D [email protected] Clifford S [email protected] Joseph D [email protected] Department of Anesthesiology, Henry Ford Health System, Detroit, MI 48202, USA2 Department of Medicine, Mt. Sinai School of Medicine, New York, NY 10029, USA3 Department of Neurology, University of Pennsylvania, Philadelphia, PA 19104, USA2005 26 7 2005 2 5 5 14 2 2005 26 7 2005 Copyright © 2005 Nagaraja et al; licensee BioMed Central Ltd.2005Nagaraja et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Putatively active drugs are often intraventricularly administered to gain direct access to brain and circumvent the blood-brain barrier. A few studies on the normal central nervous system (CNS) have shown, however, that the distribution of materials after intraventricular injections is much more limited than presumed and their exit from cerebrospinal fluid (CSF) is more rapid than generally believed. In this study, we report the intracranial distribution and the clearance from CSF and adjacent CNS tissue of radiolabeled insulin-like growth factor-1 after injection into one lateral ventricle of the normal rat brain.
Methods
Under barbiturate anesthesia, 125I-labeled insulin-like growth factor-1 (IGF-1) was injected into one lateral ventricle of normal Sprague-Dawley rats. The subsequent distribution of IGF-1 through the cerebrospinal fluid (CSF) system and into brain, cerebral blood vessels, and systemic blood was measured over time by gamma counting and quantitative autoradiography (QAR).
Results
Within 5 min of infusion, IGF-1 had spread from the infused lateral ventricle into and through the third and fourth ventricles. At this time, 25% of the infused IGF-1 had disappeared from the CSF-brain-meningeal system; the half time of this loss was 12 min. The plasma concentration of cleared IGF-1 was, however, very low from 2 to 9 min and only began to rise markedly after 20 min. This delay between loss and gain plus the lack of radiotracer in the cortical subarachnoid space suggested that much of the IGF-1 was cleared into blood via the cranial and/or spinal nerve roots and their associated lymphatic systems rather than periventricular tissue and arachnoid villi. Less than 10% of the injected radioactivity remained in the CSF-brain system after 180 min. The CSF and arteries and arterioles within the subarachnoid cisterns were labeled with IGF-1 within 10 min. Between 60 and 180 min, most of the radioactivity within the cranium was retained within and around these blood vessels and by periaqueductal gray matter. Tissue profiles at two sites next to ventricular CSF showed that IGF-1 penetrated less than 1.25 mm into brain tissue and appreciable 125I-activity remained at the tissue-ventricular CSF interface after 180 min.
Conclusion
Our findings suggest that entry of IGF-1 into normal brain parenchyma after lateral ventricle administration is limited by rapid clearance from CSF and brain and slow movement, apparently by diffusion, into the periventricular tissue. Various growth factors and other neuroactive agents have been reported to be neuroprotective within the injured brain after intraventricular administration. It is postulated that the delivery of such factors to neurons and glia in the injured brain may be facilitated by abnormal CSF flow. These several observations suggest that the flow of CSF and entrained solutes may differ considerably between normal and abnormal brain and even among various neuropathologies.
==== Body
Background
Insulin-like growth factor-1 (IGF-1) is present in brain and cerebrospinal fluid (CSF) [1,2]. Its expression and the distribution of its receptors have been shown to change dynamically during development and differentiation [3], implying that IGF-1 is involved in these processes within the central nervous system (CNS). Hinting at some neuropathological role, CSF levels of IGF-1 have been shown to rise in several diseases, most notably with pituitary tumors [4,5]. Recently, IGF-1 has been used in some studies for its putative neuroprotective effects and has been suggested to be a potential therapeutic agent in many disorders of the nervous system including amyotrophic lateral sclerosis, Alzheimer's disease, and cerebral ischemia [6].
The routes of IGF-1 administration have varied among experimental studies and conditions, but intraventricular injections have often been employed, especially for the treatment of ischemic injury [7-12]. The intraventricular approach bypasses the blood-brain barrier (BBB) and implicitly assumes direct access of the injected material to most, if not all, brain tissue. Calling this assumption into question, however, are reports that show rapid, nearly complete clearance of intraventricularly injected radiolabeled sucrose, polyethylene glycol (PEG4000; MW = 4000 Da) and 40-amino acid amyloid peptide (Aβ 1–40) from CSF and brain into blood [13,14]. These studies also indicated that the small amount of radiolabeled material remaining in brain after 1–3 hr (<10% of injected) was mostly in the walls and/or perivascular spaces of the pial arteries and arterioles within the subarachnoid cisterns and in the tissue around the aqueduct of Sylvius [13,14]. The efficacy of delivery into brain via the CSF, has also been challenged by the finding that diffusion of higher molecular weight compounds from ventricular fluid into brain is restricted by the ependyma [15]. Smaller compounds (MW<5000 Da), however, readily permeate the ependyma, but their subsequent penetration into the parenchyma is limited by factors such as their restricted rate of diffusion through the tortuous interstitium, transcapillary loss, and cellular uptake and binding [16-18].
The neurobiological effects of intraventricularly injected substances are widely accepted, but the preceding observations suggest that the pathways, rates of distribution and sites of action within the neuraxis of such agents may be more complex, perhaps more limited, than assumed. Understanding the nature of such distribution may help explain the normal pathways of CSF flow and its entrained substances and the function and malfunction of this specialized brain fluid system, the so-called "third circulation."
In the present study, the distribution of 125I labeled IGF-1 (MW = 7430 Da) among CSF, brain, and blood between 2 min and 3 hr after a bolus intraventricular injection was investigated in rats. The hypothesis to be tested was that most of the intraventricularly injected peptide, IGF-1, would be quickly cleared from the CSF and brain into blood and that the remaining amount of IGF-1 would not be widely or uniformly distributed within brain.
Methods
For these studies, 125I-IGF-1 (specific activity, 2 mCi/mmol; MW = 7430 Da) and 125I microscales were purchased from Amersham Life Sciences, USA. Male Sprague-Dawley rats weighing 300–350 g were obtained from Charles River (Cambridge, MA). All surgical procedures and experiments were performed according to National Institutes of Health guidelines under an approved protocol from the Institutional Animal Care and Use Committee of the Henry Ford Health System.
Intraventricular injections
Intraventricular injections were made according to the methods previously described [13,14]. Briefly, one femoral vein and artery were cannulated under halothane anesthesia for subsequent injection of anesthetic agent (40 mg/kg; Nembutal sodium, Abbott Laboratories, Chicago, IL) and for sampling blood, respectively. After firmly positioning the head in a stereotaxic frame, 200–500 nCi of 125I-IGF-1 in either 0.5 or 1.0 μL saline was infused at the rate of 1.0 μL/min into the anterior horn of one lateral ventricle via a syringe pump (Sage Instruments: Model 351, Cambridge, MA). Stereotaxic coordinates for the infusion were: antero-posterior = 0.0 relative to bregma; lateral = 1.5 mm to the midline; and depth = 4.5 mm down from the surface of the skull. After infusion, the cannula was left in place for the duration of the experiment (n = 4–5 for each duration). Arterial blood samples were collected at pre-designated intervals. Animals were decapitated at times ranging from 2–180 min after injection, and the whole head was instantly frozen in 2-methylbutane cooled to minus 45°C.
Despite careful usage of the above stereotaxic coordinates, the infusate was sometimes delivered mainly into the brain parenchyma. Such infusions became obvious upon viewing the autoradiograms (ARG's). Any parenchymal injections (i.e., those missing the ventricle) were due to manual errors and were unintentional (n = 2–3). Data from such experiments were analyzed separately from those with intraventricular injections and some of these findings will be reported.
Assessing organ and fluid radioactivity
The brains with accompanying CSF and blood plus surrounding meninges were dissected from the frozen head without thawing. Working within a chest freezer set at -20°C, the upper part of the skull was carefully removed with a rongeur. This frozen specimen contained virtually all of the intracranial contents and radioactivity. Immediately after removal, the frozen brain-meningeal-CSF specimens were placed in bottles chilled to -80°C and counted in a gamma counter (Wallac 1480, Turku, Finland). These data were used to calculate the clearance of 125I-radioactivity from the CSF-brain-meningeal system. Plasma was obtained from the blood samples, and radioactivity per unit volume of plasma determined. These samples were employed to track the appearance of radioactivity in blood following administration. Samples of urine, kidney, liver, and muscle were also assayed for radioactivity by gamma counting. In all instances, the counts were corrected for decay of 125I-activity. In a separate group of anesthetized rats (n = 4), subarachnoidal CSF was collected from the cisterna magna at several times after 125I-IGF-1 infusion. Acid precipitability of radioactivity was assessed on these specimens as well as on samples of plasma and urine from all the animals. All data obtained by gamma counting was calculated in units of dpm. Data obtained by quantification if autoradiographic images were calculated in nCi/g.
The procedure for obtaining the venous input rate, referred to hereafter as the emergence function, was similar to one used by Patlak and Pettigrew [19] to establish intravenous infusion schedules that yield specific arterial time-course [20,21]. First, a very rapid intravenous bolus injection was made and a carefully chosen series of well-timed blood samples were obtained: every 5–10 seconds for the first minute, every minute over the next 4 min, and then at longer intervals over the next hour or more. This curve depended on the distribution and clearance of the material throughout the body as well as the intravenous input function. A transfer function or transform that links the bolus intravenous injection to the experimentally determined arterial time-course was obtained. Then the reverse was done, namely, the transfer function was applied via the convolution integral to the arterial time-course measured after intraventricular infusion, thus generating the emergence function. In this operation, the transfer function took into account the whole body distribution and fate of the cleared material once it has entered the veins draining the site of injection including plasma protein binding and urinary excretion.
As just indicated, to obtain the emergence function, it was first necessary to determine the arterial time-course of radioactivity after a certain input into a systemic vein. To accomplish this, two rats were given an intravenous bolus of 125I-IGF-1 (~1.2 μCi). Eighteen blood samples of 20–25 μl each were drawn at a preset times between 5 sec and 60 min after the bolus. Plasma was obtained from the blood samples, and radioactivity per unit volume of plasma was determined.
Autoradiography
After measuring total radioactivity as indicated above, the frozen brain-CSF-meningeal specimens were covered in cold embedding matrix (M-1; Lipshaw, Pittsburgh, PA) and stored at -80°C in sealed plastic bags until the time of sectioning. Using a cryostat set at -19°C and starting at the caudal end, the embedded brains were serially cut at 400 μm intervals into sets of five 20 μm thick sections. The first and fifth sections were picked up on a slide for subsequent Nissl staining; these histologies were used to identify the areas of interest. The second, third and fourth sections in the series were collected on chilled, numbered coverslips for autoradiography. The next fifteen sections were discarded, and the cycle was repeated up to the beginning of the olfactory tubercles. The coverslips were placed on a slide warmer at 60°C, which instantly dried and affixed the frozen sections. The coverslips plus a set of 125I-microscales (standards calibrated in nCi/g) were placed into an x-ray cassette along with a sheet of Kodak Biomax MR-1 film. After sufficient time of exposure (20 – 25 days), the films were developed.
Image analysis
The autoradiographic images were digitized and visualized with an imaging system (Model AIS, Imaging Research Inc., St. Catharines, Canada). The matching Nissl stained sections were simultaneously displayed via a microfiche reader (Eye Communication Systems, Inc., Hartland, WI). Areas with appreciable amounts of radioactivity were located. These brain regions were demarcated on the histologies, and their exact location identified with a rat brain atlas [22].
Using the known values of the 125I-standards, a standard curve of optical density versus radioactivity in nCi/g was constructed for each autoradiograph. The optical densities of the tissue sections on the autoradiographs were converted into radioactivity units (nCi/g) using the standard curve and the radioactivity for the regions of interest was calculated. As indicated above, the quantity of injected radioactivity was varied among the experiments, with the greater amount infused for the longer times. This procedure led to adequate levels of radioactivity for accurate assaying in all experiments. For data analysis and presentation, all the values were normalized to 50 nCi per initial infusion.
To determine the distribution from lateral ventricle and aqueduct into brain, profiles of tissue radioactivity were constructed. A rectangle was drawn on the digitized image perpendicular to the ependymal border and extending 2 mm into the adjacent tissue. Radioactivity along this rectangle was measured at every 0.25 mm and a plot of radioactivity versus the distance was created. The areas under the curve for such plots were calculated using the trapezoid rule.
All data are reported as the mean ± SD.
Results
Clearance of 125I-IGF-1 from CSF and brain
Following accurate intraventricular administration (Fig. 1A, B), total intracranial radioactivity (brain-CSF-meningeal specimens) at 2 min after administration was approximately 93% of the amount infused, indicating complete or near complete delivery (Fig. 2). About 25% of the injected label was cleared from the system during the first 5 min and nearly 80% by 30 min. During the next 150 min, an additional 12% was cleared. When the infusion into one lateral ventricle was on target, the disappearance of 125I-radioactivity from the intracranial compartment was seemingly biphasic with half times and clearance fractions of 10 and 180 min and 79% and 21%, respectively. Almost all the radioactivity in the CSF drawn from the live, anesthetized rats was TCA-precipitable and presumed to be linked to IGF-1.
Figure 1 Nissl-stained histologies (A and C) and adjacent autoradiograms (ARG's; B and D) showing the lateral (LV) and the third (3 V) ventricles 30 min after injection into one lateral ventricle (A and B, respectively) or into the parenchyma (C and D, respectively). 1A. This histological section is at bregma -0.2 mm and cuts through the cerebral cortex (not labeled), LV, caudate-putamen (CPU), 3 V, anterior commissure (not labeled), and the optic chiasm (missing on the histology, but evident on the ARG, 1B). 1B. The lateral ventricle choroid plexus and the walls of the LV and 3 V on the ipsilateral side are darkly labeled on this ARG, but little radioactivity is evident in the ventricles. The flattened loop of moderate darkness at the very bottom of the ARG demarcates the subarachnoid space around the optic chiasm. The dark half-ring or crescent above the optic chiasm partially surrounds the medial nucleus of the preoptic area, a peculiar pre-hypothalamic structure; this crescent appears to be an extension of the ventral part of the 3 V but may be an artifact. 1C. This section is at bregma -0.3 mm and passes through the same structures as 1A above. Notable on this histology are the needle tract and corpus callosum, which on the ipsilateral side below the tract is expanded, pale, and edematous. 1D. Most of the radioactivity in this ARG is contained in the tissue near the main site of the intraparenchymal infusion, but some is in the choroid plexus and CSF (both shown by the tail below the main patch of blackness on the ARG). Much of the tissue radioactivity is in the corpus callosum and some of it extends via this white matter tract across the midline. Scale bar = 1.5 mm.
Figure 2 The percentage of radioactivity remaining in the entire system of frozen brain, blood, CSF, and meninges over time following either intraventricular or intraparenchymal infusion of 125I-IGF-1. There were 4–5 experiments per mean and time for the intraventricular infusions but only 2–3 for the intraparenchymal ones, which were accidental and not intended. Data are shown as mean ± SD. The initial phase dominated the first 30 min of clearance and accounted for most of the loss from the system for both sets of infusion data. From 45 min onward, virtually no radioactivity was lost from the system with the intraparenchymal infusion, whereas 125I-activity declined very slowly with intraventricular infusion (40% decline from 45 to 180 min).
In 12 of 75 infusions, much of the radioactivity was accidentally deposited directly into the parenchyma (Fig. 1C, D). In these cases, the subsequent clearance of 125I-activity from the brain-CSF-meningeal specimen was different than that following direct intraventricular administration (Fig. 2). The data from these 12 rats suggest a rapid and sizable decrease (30% loss) of intracranial radioactivity over the first 5 min (the first time of sampling for this group of rats; n = 1) and a slower but larger decrease (40% loss) over the next 40 min. It is likely that most of this clearance was of IGF-1 delivered into the CSF. There was essentially no further loss of radioactivity from the brain-CSF-meningeal system over the remaining 135 min (Fig. 2). During this phase, most of the radioactivity was located in the tissue around the injection site dorsal to the lateral ventricle. Infusion into the parenchyma caused some swelling of the corpus callosum, the white matter structure that received much of the infusate (Fig. 1C). These observations are indicative of the distribution and tissue injury obtained when agents and tracers are deliberately or accidentally administered into the parenchyma [23,24].
With both intraventricular and intraparenchymal infusions, low but significantly greater than background amounts of radioactivity were detected in the blood at 5 and 9 min (Fig. 3). Plasma 125I-activity rose sharply after 20 min in both cases and levelled off after 90 min, albeit at a much lower concentration for the intraparenchymal than intraventricular infusions (600 and 900 dpm/ml, respectively). Between 90 and180 min, the concentration in blood was essentially constant.
Figure 3 Plasma radioactivity over time after infusions into the lateral ventricle and parenchyma. Data are shown as mean ± SD. Plasma concentrations were fairly similar for both infusions up to 20 min, when the intraventricular infusion began to yield higher radioactivity; plasma concentrations for intraventricular infusions were about 50% higher than those for intraparenchymal infusion at 90–180 min. These data plus those in Fig. 2 clearly show the much greater clearance of intraventricularly infused 125I-IGF-1.
For the arterial time-course following an intravenous bolus of 125I-IGF-1 (required to find the emergence function along with the preceding data), plasma radioactivity had begun to rise by 15 sec, reached a peak around 25 sec, dropped precipitously over the next several min, and slowly declined from 10–60 min (Fig. 4A). The rate of 125I-appearance in venous blood (emergence function, dpm per min) was low at 5 min (the earliest time with significant plasma radioactivity) but it was higher for intraparenchymal (221 dpm/min) than intraventricular (185 dpm/min) injections (Fig. 4B; rate given on the ordinate). After 7.5 min, the rate increased markedly, reaching a peak at approximately 40 min of 1665 dpm/min with intraventricular injection and 924 dpm/min with intraparenchymal administration. The emergence rates fell continuously thereafter, but good estimates beyond 60 min were not possible because of the low radioactivity and the flatness of the post-bolus blood curve from 45 min onward (Fig. 4A).
Figure 4 Calculation and graphing of the emergence function, i.e., the rate of appearance of 125I-activity in venous blood. 4A. The time-course of radioactivity concentration in arterial plasma after an intravenous bolus injection of 125I-IGF-1. 4B. The emergence function calculated from the arterial plasma concentration-time curve following intravenous injection plus the arterial plasma curves following intraventricular and intraparenchymal infusions. The emergence function is the rate of appearance of the radioactivity in the venous system (dpm/min) and was calculated for 2.5, 7.5, 15, 25, 37.5, 52.5, and 75 min; the breakpoints on the graphs demarcate these times after infusion. The curves were not smoothed although the rates continually change throughout the period of interest.
The lag between the rapid disappearance from the brain-CSF-meningeal specimen from 2 min to 30 min (Fig. 2) and the marked rise in blood concentration after 20 min (Fig. 3) is a bit surprising. This discrepancy suggests that most of the 125I-IGF-1 clearance from the intracranial compartment did not directly and immediately pass into blood across the capillaries of the structures around and within the ventricles, such as the choroid plexuses, subependymal zone, and circumventricular organs (e.g., the subfornical organ and median eminence).
Among the three tissues sampled and urine, 125I-activity was at or near background at 2 and 5 min in all four (Fig. 5) and remained very low up to 20 min for liver, muscle, and urine. The radioactivity was somewhat elevated in the kidney at 9 and 20 min and thereafter arose slightly in the liver and much more in the kidney, and urine. Uptake of 125I by muscle, the largest organ in the body, was slow and slight. None of the 125I-activity in urine was TCA-precipitable, but 30–50% of that in plasma was. These findings demonstrate that little of the 125I-activity cleared from the brain-CSF-meningeal samples over the first 10 min passed immediately into blood or into non brain tissues and urine. Therefore the low plasma levels of radioactivity from 2–9 min (Fig. 3) were not the result of rapid systemic tissue uptake and urinary excretion.
Figure 5 IGF-1 radioactivity in kidney, liver, skeletal muscle and urine after its intraventricular injection. Kidneys were always seen to preferentially sequester IGF-1 from blood in the first 90 min after injection. Appreciable amounts of radioactivity began to appear in the urine after an initial lag period of ~5 min. After 20 min, radioactivity in urine rose almost linearly with time whereas that in the kidneys began dropping after 90 min. Liver and skeletal muscle had comparatively very little radioactivity throughout the period of study. The inset shows the tissue radioactivity for the first 20 min that may not be evident on the main graph. All data are mean ± SD at the various times.
Distribution of IGF-1 radioactivity within the CSF system
Two min after infusion into one lateral ventricle, CSF-contained radioactivity was already present in the third ventricle (Fig. 6A) and aqueduct (Fig. 6B), but the optical densities at these two sites were too high for accurate quantitation. This was also true for the ipsilateral lateral ventricle. The concentrations of 125I-IGF-1 at 5 min were ~10,000 nCi/g in the CSF retained within the ipsilateral lateral ventricle and ~6000 nCi/g in that within the third ventricle and aqueduct (Fig. 7). Indicative of relatively rapid CSF flow, radioactivity fell sharply thereafter in these three CSF compartments, reaching low levels by 30 min, and becoming very low (<200 nCi/g) by 180 min. Between 2 and 5 min, most of the intracranial 125I-IGF-1 was present in the CSF within the lateral and third ventricles and the aqueduct. There was some early mixing of the injected radioactivity into the contralateral lateral ventricle (data not shown). The concentration of 125I IGF-1 within the fourth ventricle was appreciable at 5 min (Fig. 7), reached a maximum (>2000 nCi/g) at 10–20 min, and fell gradually thereafter.
Figure 6 Nissl-stained histologies (A and C) and adjacent autoradiograms (ARG's; B and D) showing the third ventricle (A and B, respectively; bregma -3.3 mm) and aqueduct of Sylvius (C and D, respectively; bregma -7.6 mm) 2 min after infusion of 125I-IGF-1 into one lateral ventricle (LV). 6A. This histological section cuts through the parietal cortex, hippocampus, and thalamus (the large, light gray mass below the hippocampus) as well as the third ventricle (3 V), which is torn at its ventral border. The latter is formed by the very thin median eminence, which is one of the circumventricular organs and often tears during sectioning. 6B. The ARG indicates considerable delivery of radioactivity to the dorsal (very dark and very full) and ventral parts of the 3 V within two min of ending the infusion. The lateral streak of black to the left of the dorsal 3 V is actually part of the latter that extends over the stria medullaris thalami, the white matter that forms the dorso-medial border of the thalamus. Some spreading of 125I-activity into the tissue adjacent to the ventral part of the 3 V is evident on the ARG at this time. The most ventral portion of this part of the 3 V flairs over the median eminence; the CSF in these little pockets explains, in part, the dark streaks radiating from the bottom of the ventral part of the 3 V and over the median eminence. The streak on the left seems, however, strangely long and has seldom been seen by us before. It may be real; it may be an artifact. 6C. This histological section passes through the pineal gland, the superior colliculus (midbrain) and subiculum (cortical end of the hippocampus). The tiny dots surrounding the midbrain are clusters of small arteries and veins in the subarachnoid space and cisterns; at this level, the latter are the quadrigeminal (dorsal) and ambient (lateral) cisterns. 6D. This ARG shows a very high level of radioactivity in the CSF within the aqueduct after only 2 min of circulation and indicates the speed at which CSF flows from the lateral ventricle through the 3 V system, which includes many recesses, and into the aqueduct. Scale bar = 1.5 mm.
Figure 7 Distribution of 125I-IGF-1 radioactivity in five CSF compartments as a function of time after ending the infusion. The data are shown as mean ± SD at the various times. The first two sets of points are at 5 and 9 min. By 5 min, the peak concentrations have already passed through the lateral and third ventricles and the aqueduct; at 30 min, 125I-IGF-1 in these three compartments has fallen to <750 nCi/g. Radioactivity in the fourth ventricle reached a plateau over the 9–20 min period and dropped thereafter. In the midbrain cisterns, 125I-IGF-1 sharply rose from 20 to 45 min and then fell continuously thereafter. The peak radioactivity at 45 min is notable, exceeding that in the 4th ventricle by 5-fold. These curves show that 125I-IGF-1 moved relatively briskly through the ventricular system.
Radioactivity in the quadrigeminal, ambient, and interpeduncular cisterns, parts of the midbrain subarachnoidal system, was essentially zero for the first five minutes but was detectable, albeit low, from 10–20 min (Figs. 7 and 8B); it rose rapidly from 20 to 45 min (~5700 nCi/g) and then fell over the next 135 min (Fig. 7). Over the 20–180 min period, much of the intracranial radioactivity was localized around the arteries and arterioles within these and other cisterns (Fig. 8A and 8B). This radioactivity could be in either the perivascular sheath, which contains CSF, or the vessel wall or both.
Figure 8 Nissl-stained histologies (A and C) and adjacent autoradiograms (ARG's) at 20 min after 125I-IGF-1 infusion (A and B, respectively; bregma -7.8 mm) and 90 min (C and D, respectively; bregma -6.8 mm). 8A. This section passes through the pineal gland, inferior colliculus (midbrain), aqueduct, and hippocampus and is just slightly caudal to Fig. 5C. Patches of small arteries and veins are scattered throughout the midbrain cisterns. 8B. The adjacent ARG shows some penetration of radioactivity into periaqueductal gray matter at 20 min and a minute amount within the midbrain cisterns. 8C. About 1.0 mm rostral to 7A, this section cuts through the superior colliculus, aqueduct, midbrain cisterns, dentate gyrus of the hippocampus, and the pontine nuclei (the large mass hanging below the midbrain). 8D. The matching ARG, obtained at 90 min, indicates considerable retention of 125I-IGF-1 by the tissue immediately around the aqueduct and on both sides of the midbrain cisterns. The arteries and veins in the subarachnoid space also retain much radioactivity at this time. Scale bar = 1.5 mm.
As we have reported for 14C-sucrose, 14C-PEG4000, and 125I-amyloid β peptide (Aβ1–40) [13,14], the radioactivity in the midbrain subarachnoid cisterns during the first 20 min appeared to come mostly from the third and fourth ventricles via the velum interpositum (Fig. 9A) and anterior medullary velum (Fig. 9B), respectively. These two velae are extensions of the subarachnoid space and form part of the membranous covering of these two ventricles. Once the CSF-entrained radioactivity moved from ventricle into these velae, it flowed from the velum interpositum and the superior medullary velum into the quadrigeminal, ambient and the interpeduncular cisterns. The sharp rise in midbrain cisternal 125I-IGF-1 from 20–45 min (Fig. 7) probably is mainly due to CSF flow via the "classic" route; that is, from the lateral recesses of the fourth ventricle through the foramina of Magendie and Luschka and into the cisterna magna and the subarachnoidal cisterns around the brain stem.
Figure 9 Nissl-stained histologies (A and C) and adjacent autoradiograms (ARG's) at 30 min after 125I-IGF-1 infusion for the two subarachnoid velae, the velum interpositum (A and B, respectively; bregma -4.3 mm) and the superior (anterior) medullary velum (C and D, respectively; bregma -9.2 mm). 9A. Starting at the subfornical organ and running caudad along the roof and upper sides of the third ventricle (3 V), the velum interpositum separates the 3 V from the subarachnoid space within it. The 3 V is clear and featureless, whereas the velum interpositum is filled with typical arachnoid tissue, e.g., trabeculae and blood vessels. The thalamus is the brain structure comprising the lower half of the figure. 9B. At 30 min, there was some radioactivity in the 3 V and the contralateral velum (left side) and much more in the ipsilateral velum. The dark spot at the bottom of the ARG came from 125I-activity within the mammillary recess of the 3 V. 9C. The superior medullary velum (SMV) forms the posterior wall of the recess of the inferior colliculus (this recess is an outpouching ofthe aqueduct and not shown) and the anterior roof of the fourth ventricle, mostly in the vicinity of the cerebellum; it contains subarachnoid CSF and tissue. 9D. The very dark "A-shaped" figure in the middle of this ARG indicates radioactivity that has collected over 30 min in the CSF and tissue of the SMV. The grayness within the legs of the upright "A" indicates some diffusion into the cerebellar lobe from the SMV. The legs and crossbar of the inverted "A" in the lower middle of the ARG represent radioactivity within the SMV on the ventral side of the cerebellar lobe (the SMV is not visible on the histology at this magnification, Fig. 9C; it can be seen to be subarachnoid tissue in Fig. 6 of Ghersi-Egea et al. [14]. The lighter spot within the lower part of the inverted "A" arises from CSF signal within the fourth ventricle. The faint figures in the rest of this ARG indicate radioactivity in several midbrain cisterns of the subarachnoid system at 30 min. Scale bar = 0.6 mm.
At longer times (e.g., 90 min; Fig. 8D), an appreciable uptake and retention of 125I-IGF-1 was evident for some of the tissues adjacent to the array of midbrain cisterns; among these tissues are the medial geniculate (a thalamic nucleus), the basal part of the cerebral peduncles, and the dentate gyrus of the hippocampus. Again as in our other studies, no radioactivity could be seen in the subarachnoid space over the cerebral cortices.
Periventricular and periaqueductal tissue profiles
The tissue profiles of radioactivity indicated that the ependyma is not a significant barrier to the movement of IGF-1 from CSF into gray matter. This is evidenced by the sizable amounts of radiolabel that moved within the first 5 min as far as 500 μm into periventricular tissue, exemplified in this study by the caudate-putamen (Fig. 10A), and 250 μm into periaqueductal gray matter (Fig. 10B). These distances are reasonable for free diffusion through the tortuous extracellular space of the brain for such a duration and molecule. The differences in the amounts of radioactivity and its spread into the parenchyma between these two areas is caused by the 1–2 min delay in delivery to the aqueduct and the lower 125I-concentration within it (Fig. 7). The latter is the result of dilution within the third ventricle by inflowing unlabeled CSF from the contralateral lateral ventricle and by production of fresh CSF by the third ventricle choroid plexus.
Figure 10 Tissue radioactivities as a function of distance (profiles) away from the CSF-brain interface for the caudate-putamen (A) and periaqueductal gray matter (B) at four times. The shapes of the curves and the concentration-distance integrals indicate two different tissue distribution dynamics. To illustrate this, we will consider only the 5 and 9 min profiles and avoid the matter of sizable tissue uptake from blood that affects the later times. For the caudate-putamen (10A), the 5 min points all lie above the 9 min ones out to 1.5 mm, where they both approach zero. The areas under these curves are 2757 (nCi/g) × mm at 5 min and 2002 (nCi/g) × mm at 9 min; clearly there was a loss of radioactivity, probably back into the CSF, over the 5–9 min period. In contrast, periaqueductal gray matter radioactivities (10B) were similar at the edge of the tissue (x = 0) at these two times but the 9 min points were higher than the 5 min ones from 0.25 to 0.75 mm, where both became essentially zero. The area under the 5 min curve was less than that under the 9 min one, 662 vs. 873 (nCi/g) × mm, respectively; periaqueductal gray matter, thus, continued to take up 125I-IGF-1 even as concentration in aqueductal CSF was falling (Fig. 7). One explanation for this is greater binding or trapping of IGF-1 in periaqueductal gray matter than in caudate-putamen. Data are shown as mean ± SD.
As the CSF concentration falls within the ventricular system over time, the profiles flatten for both tissues sites. This decline is driven by a combination of backflux of unbound radiolabel from tissue to adjacent CSF and movement further into the tissue (see the legend to Fig. 10 for more on this). At all times and both tissue sites, 125I-activity decreased over distance into the brain and reached a plateau around 1.5–2.0 mm for caudate-putamen (Fig 10A) and 1.0–1.5 mm for periaqueductal gray matter (Fig. 10B). Indicating a net loss of IFG-1 over time, the mean area under the tissue radioactivity-distance curve over 2.0 mm for the caudate-putamen was highest at 5 min (2757 nCi mm g-1) and fell continuously thereafter reaching 1293 nCi mm g-1 at 180 min. In contrast, the mean area under the curve for periaqueductal gray was lowest at 5 min (652 nCi mm g-1), rose to 955 nCi mm g-1 by 30 min, and remained high at 180 min (922 nCi mm g-1). Regional differences in the uptake and retention of IGF-1 are obvious from these two tissue profiles.
Discussion
Experimental procedure
The intraventricular route of IGF-1 administration was selected because it has been used for many studies. Bolus injections into one lateral ventricle of IGF-1 have been reported, for instance, to improve neurological outcome for both permanent and transient cerebral ischemia in rats and other small laboratory animals [9-12]. A small bolus was chosen for injection in order to cause a minimal and transient disturbance of CSF pressure and flow.
Freezing the whole head immediately after decapitation has been shown to prevent collapse of the ventricles and preserve pre-mortem blood and CSF distribution within the cranial cavity [25]. A key part of this procedure is keeping the head and its contents frozen while removing the brain-CSF-meningeal specimen. The procedure is arduous, but these conditions minimize loss and post-mortem movement of radioactivity. In contrast to our procedure, others have removed the brain before freezing [15,26] or have perfusion-fixed the brain before sectioning and autoradiography [27].
The emergence function is based on the assumption of tracer kinetics; that is, the concentration of the radiotracer is presupposed to be very low with respect to the cold substrate, in this case, IGF-1. With this assumption, the distribution of 125I-IGF-1 is considered to be a linear function of its concentration in CSF, brain, and blood. There are IGF-1 binding proteins (IGFBP's) in the CNS and blood [6] and undoubtedly a state of dynamic equilibrium exists between IGF-1 and its binding proteins. Although 125I-IGF-1 concentrations were somewhat high in the ventricular and intravenous infusates, it would be diluted to tracer levels in plasma within seconds as it mixes in the circulating blood and binds to plasma proteins such as the IGFBP's. The tracer assumption for the intravenous infusion may be violated for the first 30 seconds or so (systemic circulation time in the normal rat is about 15 sec) but not thereafter. This dilution-mixing process is probably slower in the CSF, but the amount of 125I-IGF-1 intraventricularly infused was considerably lower, namely, around 25% of the intravenous dose, and that enhances the tracer assumption on the CNS side.
Peculiarities in cerebrospinal fluid flow
Recent experiments indicate that about 25% of CSF and CSF-borne radiolabeled sucrose [14], PEG4000 [13], soluble amyloid beta peptide (sAβ1–40) [13], and IGF-1 (Fig. 9) flow into the subarachnoid extensions of the velum interpositum (from the dorsal part of third ventricle) and superior medullary velum (from the rostral part of the fourth ventricle). Within 4–5 min, intravelar CSF appears in the basal and midbrain cisterns and the arteries and arterioles contained within them [14]. The rest of the CSF flows from the sites of production within the ventricles to and through the lateral recesses of the fourth ventricle and the foramina of Luschka and Magendie into the cisterna magna. From there, this CSF passes into the spinal and cranial subarachnoid space. It was also noted in these three studies that very little of these radiolabel materials reached the cortical surface of the normal rat brain even after three hours of circulation.
Within the basal cisterns, all four of these radiolabeled materials accumulate in the pial arteries and arterioles. Anatomically, these blood vessels seem to be surrounded by a specialized pia-arachnoid membrane complex that functions to trap CSF-borne substrates. In addition, the highly vascular choroid plexuses were also observed to take up and retain some IGF-1 for the first 30 min (Fig. 1B and 5B). It should be noted that IGF-1 receptors are present in rat choroid plexus [28]. It may be that one or more of these vascular tissues are involved in the putative neuroactive effects of IGF-1 (Figs. 1 and 7).
Of relevance to vascular involvement, gene transfer to the adventitia and leptomeninges of arteries and arterioles along the ventral surface of the brain has been accomplished in mice via a recombinant adenovirus injected into either one lateral ventricle or the cisterna magna [29,30]. Such transfer was not, however, achieved on the dorsal surface of the brain, which may be due to the very low delivery of CSF-entrained substances over the cortex as observed in other studies [13,14] as well as the present one.
Differences in the CSF-brain-blood distribution of peptides
Before proceeding further, an evaluation of the robustness of these techniques and data is in order. The gamma counting technique for the plasma and tissue radioactivity is very straightforward and highly accurate. The delay of several minutes before the elevation of both the plasma (Figs. 3) and systemic tissue (Figs. 3, 4, 5) 125I-activity is probably not an artefact. One possibility to consider is that not all of the IGF-1 reached the lateral ventricle CSF at injection and the initial rapid decline is caused by that shortcoming. At 2 min, however, the earliest time of sampling, around 93% of the radioactivity was still in the system, and most of the decrease occurred subsequently. Similar studies with 14C-sucrose [14] and 14C-PEG4000 [13] showed that virtually all the infused radioactivity was present within the cranium for the first 5 min. On the occasions when it was obvious from the autoradiograms that the IGF-1 was partially or completely infused into the brain parenchyma, those experiments were eliminated from the successful group. It, thus, seems that there is little or no problem with the accuracy of the intraventricular infusion method.
In the present study, IGF-1 exited from the brain-CSF-meningeal compartment at an unexpectedly fast rate. The clearance of ~50% of it by 15 min and ~80% by 30 min is far faster than that of sucrose [14] or PEG4000 [13] which had no clearance over the first 5 min and apparent half-lives ~60 min for both. Of the molecules studied to date, only 125I-labeled soluble amyloid β peptide (I-sAβ1–40) with a ~50% loss within 8 min [13], is cleared more rapidly than IGF-1. In another study, the half time of disappearance from brain of insulin has been calculated as 26 min in mice [26].
Rapid clearance of IGF-1 and I-sAβ1–40 from the brain-CSF-meningeal system would be expected to take place directly into the circulating cerebral blood and lead to systemic distribution. In the study with I-sAβ1–40 [13], this was found to be the case: when arterial blood concentration of I-sAβ1–40 was measured 3.5 min after intraventricular infusion, about 30% of the infused I-sAβ1–40 had already been cleared from the brain-CSF-meningeal system, and the blood level of 125I was high. With further clearance over time, the concentration of I-sAβ1–40 in blood continued to rise slowly, increasing by 25% between 3.5 and 30 min. Consistent with this, the tissue levels of 125I-activity were appreciable in liver at 3.5 and 10 min and even higher in the kidney at these two times. Although emergence function analysis was not done on these data, it was evident that the I-sAβ1–40 cleared from the CSF-brain-meningeal system passed immediately into the circulating blood and distributed throughout the body. Because of the speed of this process, the clearance of I-sAβ1–40 was assumed to be directly across the capillaries of the choroid plexuses, the subependymal zone, and other periventricular structures such as the circumventricular organs. This mechanism has been suggested for not only sAβ1–40 but also other peptides and immunoglobulins [13,31,32].
In contrast, this study found an unexpected delay between the loss of IGF-1 from the CSF-brain-meningeal system and its appearance in blood. Although 35% of the infused radioactivity was gone from the system after 9 min, the plasma concentration was very low (less than 8% of the maximum) at this time and at the preceding times of sampling (Fig. 3). Thereafter the clearance of IGF-1 from the CSF-brain-meningeal system continued, with only 20% remaining by 30 min. Plasma radioactivity increased between 9 and 20 min, more sharply between 20 and 30 min, and even more steeply between 30 and 45 min, finally reaching a plateau at 90 min (Fig. 3). Reflecting this delayed, slow increase in blood radioactivity, the emergence rate for IGF-1 (Fig. 4) began to increase between 9 and 12 min and then climbed markedly between 12 and 20 min.
A study on the clearance of a biologically inert and extracellular compound 14C-PEG4000 (MW = 4000 Da) following intraventricular infusion [13], showed that the compound was retained in higher concentration in the CSF system at early time points than for IGF-1, but a similar delay occurred before it appeared in blood. Since the PEG clearance pathway is probably non-specific and it is most likely lost by CSF bulk flow and absorption, it is possible that a sizable portion of the infused IGF-1 could also have been cleared via this route. However, the delay in appearance in blood in both studies indicates that the compounds lost from the brain-CSF-meningeal system at the early time points did not go directly into the circulating cerebral blood.
Hence the temporal mismatch of IGF-1 clearance from the CSF-brain-meningeal and its appearance in systemic blood remains enigmatic. A small part of the initial clearance of IGF-1 does seem to be directly into blood via the periventricular tissue and choroid plexuses as occurs with sAβ1–40. A much larger portion of it may flow rather rapidly out of the cranial space via the PEG-CSF pathway. The time-course and magnitude of this is suggested by the steep rise in the rate of appearance in systemic blood that starts after 9 min and peaks at 40 min (Fig. 4B). Much of this bulk CSF clearance may be by way of the cranial nerve sheaths and the cranial and cervical lymphatic systems as has been shown for albumin [33,34] or by passage from the cisterna magna into the spinal subarachnoid space, which is outside of our sampling field, and hence into the venous system. One or the other or both of these putative routes could be the physiological explanation of the delay in the IGF-1 appearance in blood.
Tissue profiles
The tissue profiles showed some penetration of IGF-1 into the brain immediately adjacent to the CSF (Fig. 9A, caudate-putamen; Fig. 9B, periaqueductal gray matter) over the first 9 min when blood-contained radioactivity was low (Fig. 3). At these times, little or no radioactivity had, however, moved beyond 1.0 mm in the caudate-putamen and 0.5 mm in periaqueductal gray matter. At 30 and 180 min, the concentrations of radioactivity in the ependymal and subependymal tissue were much higher than in the adjacent CSF (Fig. 6), and the radioactivity at the peripheral zone was three-times higher than the deeper tissue plateau concentration. Collectively, these observations suggest that 125I-IGF or some labeled metabolite(s) is "trapped" in this tissue, perhaps to a receptor or within some periventricular intracellular compartment.
The implication of this is that for intraventricularly injected IGF-1 to be centrally effective it either interacts with receptors or compartments that are immediately around the ventricular system or works at deep active sites at the low levels indicated by the 30–180 min 125I-IGF-1 plateaux (Fig. 9). Recent work has shown the presence of neuronal progenitor cells in the subependymal layer (a.k.a., the subventricular zone), especially of the lateral ventricles and dentate gyrus of the hippocampus [35,36]. Such cells are well positioned with respect to circulating CSF and could be the target of CSF-borne growth factors.
Intraventricular IGF-1 delivery following brain injury
Injury to the brain might result in changes to the flow of CSF and affect the distribution of intraventricularly administered peptides into brain. For example, 2 hr after hypoxic-ischemic injury, tritiated IGF-1 was intraventricularly infused for 30 min and immediately thereafter was found in the ipsilateral cerebral cortex [37]; this cortical radioactivity decreased over the next 6 hr and reached background (the contralateral level) after 12 hr. Microautoradiography indicated that 3H-IGF-1 was also present in the corpus callosum and its ventral extension, the external capsule, plus the perivascular (Virchow-Robin) spaces that surround the penetrating arteries and arterioles of the cortex 30 min after ending the infusion. This contrasts with our findings for normal brain of little or no IGF distribution to the cerebral cortex and underlying white matter. On the other hand, in our experimental "mistakes" in which much of the infusate was placed in the parenchyma causing trauma, the white matter was also heavily labeled with 125I-IGF-1 at all times of sampling (30 min data showed in Fig. 1D).
As to clearance into systemic blood, the serum concentration of 3H-IGF-1 in the studies of Guan et al [37] was high at 30 min after beginning the infusion in both hypoxia-ischemia injured rats and uninjured controls and rose 3-fold and 5-fold, respectively, over the next 3 hr. This resembles the plasma data given in Figure 3 for the intraparenchymal (accidental) and intraventricular infusion groups.
In a later study with the same hypoxia-ischemia model and intraventricular infusion protocol, Guan et al [38] investigated the cellular distribution of 3H-IGF-1 and its co-localization with immunoreactive IGFBP-2 at 30 min and 6 hr after administration. Thirty min after ending the infusion of 3H-IGF-1, it was mainly found in the pia mater, the perivascular spaces (extensions of the external subarachnoid space), and subcortical white matter structures such as the corpus callosum. The addition of cold IGF-1 to the infusate diminished the distribution of radiotracer to white matter tracts but not to the pia mater and perivascular spaces. Within subcortical white matter, tritiated IGF-1 co-localized with IGFBP-2 immunoreactivity and was seen at the microautoradiographic level to be associated with fiber tracts and some oligodendrocytes. In the cerebral cortex 30 min after completing the infusion, the 3H-signal was diffusely distributed but evident on the neurons of layers II-V, glia (astrocytes or microglia were not specified), and the extracellular matrix. On the basis of these data and the earlier study [37], the authors suggested for this model of hypoxia-ischemia that: 1) such patterns of IGF-1 distribution were not compatible with simple diffusion and were likely to be abnormal; 2) bulk flow of CSF through the ependyma and along white matter tracts and also through the subarachnoidal perivascular spaces carried IGF-1, in some cases in association with IGFBP-2, to the neurons and glia of the injured cerebral cortex; and 3) this abnormal flow of CSF promotes the delivery of intraventricularly administered therapeutic agents to damaged tissue sites.
Our findings are consistent with this postulate in several somewhat indirect ways. The observations on normal brain indicate little or no distribution of intraventricularly infused IGF-1 to subcortical white matter, cerebral cortex, and perivascular spaces over 3 hr (Figs. 1A, 1B, and 5); the patterns reported by Guan et al [37] are, thus, most probably abnormal and due to the pathological state. In support of this, in the experiments where the IGF-1 was infused into the parenchyma and injured the tissue, much of the radioactivity was found in the edematous corpus callosum (Fig. 1C and 1D). The data in Figure 9 show that IGF-1 does not move very far into normal gray matter and illustrate the limits of such movement, probably mostly diffusional, in control conditions. Seemingly, bulk flow of CSF plus any edema fluid formed within the parenchyma is needed to transport IGF-1 as widely as appears to be the case in the hypoxia-ischemia model [37]. Of some relevance to this, convection enhanced delivery of chemotherapeutic agents infused directly into the parenchyma at appreciable rates of volume flow are currently under investigation in several laboratories and neurosurgical settings [39,40].
Metabolism of IGF-1 in brain
Finally, the retention of the radiolabel on the parent compound is always a problem in the interpretation of data from studies such as ours. Proteases in many organs of the body including brain have been shown to cleave IGF-1 into des(1–3)IGF-1 and the tripeptide, glycine-proline-glutamate (GPE) [41]. It is possible that one or more of these proteolytic products is no longer radiolabeled, is not rapidly cleared from the system, does not permeate the BBB, can diffuse great distances in brain tissue and reach the appropriate receptors, and is active. Fitting with this scheme, some reports have shown the beneficial effects of intraventricularly injected active IGF-1 fragment in experimental studies [42,43]. It was also felt that the tripeptide GPE was more effective than its prohormone IGF-1 [43]. If substantiated, then rapid removal of IGF-1 from the brain reported herein may be the reason for the lesser effectiveness of this growth factor.
Conclusion
Further analyses of the normal flow of CSF and distribution of its entrained materials are needed for the comprehensive understanding of this system and its therapeutic potential. Such studies may shed light on more appropriate sites for intracerebral injections for accessing specific regions and pathways within the brain via this system.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TNN performed all the intracerebral injections, blood sampling and QAR quantification studies and drafted the manuscript. PP carried out the QAR analyses. MG participated in the dissection of brains and plasma precipitation and measurements of radioactivity. PDG participated in the design of the study and funded the study. CSP performed the emergence function analysis. JDF conceived the study, participated in its design and coordination and helped draft the manuscript. All authors read and approved the final manuscript.
Acknowledgements
The authors thank Wei Hong Zhao for her skilled technical assistance. This work was supported by NIH grant 1RO1NS34839. An abstract of the study was presented in the 28th Annual Conference of the Society for Neuroscience.
==== Refs
de Pablo F de la Rosa EJ The developing CNS: a scenario for the action of proinsulin, insulin and insulin-like growth factors Trends Neurosci 1995 18 143 150 7754526 10.1016/0166-2236(95)93892-2
Baskin DS Wilcox BJ Figelwicz DP Dorsa DM Insulin and insulin-like growth factors in the CNS Trends Neurosci 1988 2 107 111 2465607 10.1016/0166-2236(88)90155-5
Bondy C Werner H Roberts CT JrLeRoith D Cellular pattern of type-1 insulin-like growth factor gene expression during maturation of the rat brain: comparison with insulin-like growth factors I and II Neuroscience 1992 46 909 923 1311816 10.1016/0306-4522(92)90193-6
Aldred AR CM B Schreiber G The cerebral expression of plasma protein genes in different species Comp Biochem Physiol B 1995 111 1 15 7749630 10.1016/0305-0491(94)00229-N
Glick RP Unterman TG Radioimmunoassay of insulin-like growth factors I and II in the cerebrospinal fluid of patients with pituitary and other central nervous system disorders Neurosurgery 1995 36 556 564 7753356
Dore S Kar S Quirion R Rediscovering an old friend, IGF-1: potential use in the treatment of neurodegenerative diseases Trends Neurosci 1997 20 326 331 9246719 10.1016/S0166-2236(96)01036-3
Guan J Williams C Gunning M Mallard C Gluckman P The effects of IGF-1 treatment after hypoxic-ischemic brain injury in adult rats J Cereb Blood Flow Metab 1993 13 609 616 8314914
Bergstedt K Wieloch T Changes in insulin-like growth factor 1 receptor density after transient cerebral ischemia in the rat. Lack of protection against ischemic brain damage following injection of insulin-like growth factor J Cereb Blood Flow Metab 1993 13 895 898 8360296
Williams C Guan J Miller O Beilharz E McNeill H Sirimanne E Gluckman PD The role of the growth factors IGF-1 and TGF-1 after hypoxic-ischemic brain injury Ann N Y Acad Sci 1995 765
Guan J Williams CE Skinner SJM Mallard EC Gluckman PD The effects of insulin-like growth factor (IGF)-1, IGF-2 and des-IGF-1 on neuronal loss after hypoxic-ischemic brain injury in adult rats: evidence for a role for IGF binding proteins Endocrinology 1996 137 893 898 8603600 10.1210/en.137.3.893
Johnston BM Mallard EC Williams CE Gluckman PD Insulin-like growth factor-1 is a potent neuronal rescue agent after hypoxic-ischemic injury in fetal lambs J Clin Invest 1996 97 300 308 8567948
Guan J Gunn AJ Sirimanne ES Tuffin J Gunning MI Clark R Gluckman PD The window of opportunity for neuronal rescue with insulin-like growth factor-1 after hypoxia-ischemia in rats is critically modulated by cerebral temperature during recovery J Cereb Blood Flow Metab 2000 20 513 519 10724116
Ghersi-Egea JF Gorevic PD Ghiso J Frangione B Patlak CS Fenstermacher JD Fate of cerebrospinal fluid-borne amyloid beta-peptide: rapid clearance into blood and appreciable accumulation by cerebral arteries J Neurochem 1996 67 880 883 8764620
Ghersi-Egea JF Finnegan W Chen JL Fenstermacher JD Rapid distribution of intraventricularly administered sucrose into cerebrospinal fluid cisterns via subarachnoid velae in rat Neuroscience 1996 75 1271 1288 8938759 10.1016/0306-4522(96)00281-3
Maness LM Banks WA Zadina JE Kastin AJ Periventricular penetration and disappearance of ICV Tyr-MIF-1, DAMGO, Tyrosine and albumin Peptides 1996 17 247 250 8801529 10.1016/0196-9781(95)02135-3
Nicholson C Sykova E Extracellular space structure revealed by diffusion analysis Trends Neurosci 1998 21 207 215 9610885 10.1016/S0166-2236(98)01261-2
Patlak CS Fenstermacher JD Measurements of dog blood-brain transfer constants by ventriculocisternal perfusion Am J Physiol 1975 229 877 884 1190330
Fenstermacher J Kaye T Drug "diffusion" within the brain Ann N Y Acad Sci 1988 531 29 39 3382143
Patlak CS Pettigrew KD A method to obtain infusion schedules for prescribed blood concentration time courses J Appl Physiol 1976 40 458 463 819414
Namba H Lucignani G Nehlig A Patlak C Pettigrew K Kennedy C Sokoloff L Effects of insulin on hexose transport across the blood-brain barrier in normoglycemia Am J Physiol 1987 252 E299 E303 3548427
Otsuka T Wei L Bereczki D Acuff V Patlak C Fenstermacher J Pentobarbital produces dissimilar changes in glucose influx and utilization in brain Am J Physiol 1991 261 R265 275 1877685
Paxinos G Watson C The rat brain in stereotaxix coordinates 1982 New York: Academic Press
Shibata M Yamada S Ram Kumar S Calero M Bading J Frangione B Holtzman DM Miller CA Strickland DK Ghiso J Zlokovic BV Clearance of Alzheimer's amyloid 1–40 peptide from brain by LDL receptor-related protein at the blood-brain barrier J Clin Invest 2000 106 1489 1499 11120756
Pardridge WM Drug delivery to the brain J Cereb Blood Flow Metab 1997 17 713 731 9270488 10.1097/00004647-199707000-00001
Bereczki D Wei L Acuff V Gruber K Tajima A Patlak C Fenstermacher J Technique-dependent variations in cerebral microvessel blood volumes and hematocrits in the rat J Appl Physiol 1992 73 918 924 1400056
Cashion MF Banks WA Kastin AJ Sequestration of centrally administered insulin by the brain: effects of starvation, aluminum and TNF Hormones Behav 1996 30 280 286 10.1006/hbeh.1996.0034
Moos T Morgan EH Kinetics and distribution of [59 Fe-125I] transferrin injected into the ventricular system of the rat Brain Res 1998 790 115 128 9593852 10.1016/S0006-8993(98)00055-9
Werther GA Abate M Hogg A Cheesman H Oldfield B Hards P Hudson B Power B Freed K Herington AC Localization of insulin-like growth factor -1 mRNA in rat brain by in situ hybridization: relationship to IGF-1 receptors Mol Endocrinol 1990 4 773 778 2177145
Christenson SD Lake KD Ooboshi H Faraci FM Davidson BL Heistad DD Adenovirus-mediated gene transfer in vivo to cerebral blood vessels and periventricular tissue in mice Stroke 1998 29 1411 1416 9660397
Heistad DD What's new in cerebral circulation? Microcirculation 2001 8 365 375 11781810 10.1038/sj/mn/7800109
Muraszko K Sung C Walbridge S Greenfield L Dedrick RL Oldfield EH Youle RJ Pharmacokinetics and toxicology of immunotoxins administered into the subarachnoid space in nonhuman primates and rodents Cancer Res 1993 53 3752 3757 8339287
Bergman I Burckart GJ Pohl CR Venkataraman R Barmada NA Griffin JA Cheung N-K Pharmacokinetics of IgG and ImG anti-ganglioside antibodies in rats and monkeys after intrathecal administration J Pharmacol Exp Ther 1998 284 111 115 9435168
Boulton M Young A Hay J Armstrong D Flessner M Schwartz M Johnston M Drainage of CSF through lymphatic pathways and arachnoid villi in sheep: measurement of 125I-albumin clearance Neuropathol Appl Neurobiol 1996 22 325 333 8875467
Boulton M Flessner M Armstrong D Hay J Johnston M Lymphatic drainage of the CNS: effects of lymphatic diversion/ligation on CSF protein transport to plasma Am J Physiol Regulatory Integrative Comp Physiol 1997 272 R1613 R1619
Peretto P Merighi A Fasolo A Bonfanti L The subependymal layer in rodents: A site of structural plasticity and cell migration in the adult mammalian brain Brain Res Bull 1999 49 221 243 10424843 10.1016/S0361-9230(99)00037-4
Alvarez-Buylla A Garcia-Verdugo JM Neurogenesis in adult subventricular zone J Neurosci 2002 22 629 634 11826091
Guan J Skinner SJM Beilharz EJ Hua KM The movement of IGF-1 into the brain parenchyma after hypoxic-ischemic injury NeuroReport 1996 7 632 636 8730846
Guan J Beilharz EJ Skinner SJM Williams CE Gluckman PD Intracerebral transportation and cellular localization of insulin-like growth factor-1 following central administration to rats with hypoxic-ischemic brain injury Brain Res 2000 853 163 173 10640614 10.1016/S0006-8993(99)02030-2
Bobo RH Laske DW Akbasak A Morrison PF Dedrick RL Oldfield EH Convection-enhanced delivery of macromolecules in the brain Proc Natl Acad Sci U S A 1994 91 2076 2080 8134351
Broaddus WC Prabhu SS Gillies GT Neal J Conrad WS Chen Z-J Fillmore HL Young HF Distribution and stability of antisense phosphorothioate oligonucleotides in rodent brain following direct intraparenchymal controlled-rate infusion J Neurosurg 1998 88 734 742 9525721
Yamamoto H LJ M Enzymatic conversion of IGF-1 to des(1–3)IGF-1 in rat serum and tissues: a further site of growth hormone regulation of IGF-1 action J Endocrinol 1995 146 141 148 7561610
Guan J Waldvogel HJ Faull RLM Gluckman PD Williams CE The effects of the n-terminal tripeptide of insulin-like growth factor-1, glycine-proline-glutamate in different regions following hypoxic-ischemic brain injury in adult rats Neuroscience 1999 89 649 659 10199602 10.1016/S0306-4522(98)00338-8
Guan J Krishnamurthy R Waldvogel HJ Faull RLM Clark R Gluckman PD N-terminal tripeptide of IGF-1 (GPE) prevents the loss of TH positive neurons after 6-OHDA induced nigral lesion in rats Brain Res 2000 859 286 292 10719076 10.1016/S0006-8993(00)01988-0
|
16045806
|
PMC1190198
|
CC BY
|
2021-01-04 16:37:38
|
no
|
Cerebrospinal Fluid Res. 2005 Jul 26; 2:5
|
utf-8
|
Cerebrospinal Fluid Res
| 2,005 |
10.1186/1743-8454-2-5
|
oa_comm
|
==== Front
Clin Mol AllergyClinical and molecular allergy : CMA1476-7961BioMed Central London 1476-7961-3-91598969010.1186/1476-7961-3-9ReviewThe basophil activation test by flow cytometry: recent developments in clinical studies, standardization and emerging perspectives Boumiza Radhia [email protected] Anne-Lise [email protected] Guillaume [email protected] Immunology Laboratory, Lyon-Sud University Hospital, Lyon, France2 Immunology Laboratory, Hôpital Neurologique, Lyon, France2005 30 6 2005 3 9 9 4 5 2005 30 6 2005 Copyright © 2005 Boumiza et al; licensee BioMed Central Ltd.2005Boumiza et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The diagnosis of immediate allergy is mainly based upon an evocative clinical history, positive skin tests (gold standard) and, if available, detection of specific IgE. In some complicated cases, functional in vitro tests are necessary. The general concept of those tests is to mimic in vitro the contact between allergens and circulating basophils. The first approach to basophil functional responses was the histamine release test but this has remained controversial due to insufficient sensitivity and specificity. During recent years an increasing number of studies have demonstrated that flow cytometry is a reliable tool for monitoring basophil activation upon allergen challenge by detecting surface expression of degranulation/activation markers (CD63 or CD203c). This article reviews the recent improvements to the basophil activation test made possible by flow cytometry, focusing on the use of anti-CRTH2/DP2 antibodies for basophil recognition. On the basis of a new triple staining protocol, the basophil activation test has become a standardized tool for in vitro diagnosis of immediate allergy. It is also suitable for pharmacological studies on non-purified human basophils. Multicenter studies are now required for its clinical assessment in large patient populations and to define the cut-off values for clinical decision-making.
allergybasophilsflow cytometryCD63CD203CCRTH2
==== Body
Introduction
Anaphylaxis consists of an immediate IgE-dependent reaction in response to allergens. Clinical symptoms are caused by an initial systemic histamine release by mast cells and basophils that may lead to shock with laryngeal edema, lower-airway obstruction and hypotension. The most frequent allergens involved in immediate allergy are found in peanuts, fish, bee and wasp venoms, drugs and latex [1,2]. The identification of responsible allergens remains a key step for practicing allergen avoidance and specific immunotherapy. The diagnosis is mainly based upon an evocative clinical history (including temporal association between symptoms and allergen exposure), positive skin tests, which remain the gold standard in this context and, if available, detection of specific IgE [3]. In most patients, these features allow both diagnosis and identification of the offending allergen. Nevertheless, skin testing is contraindicated in some patients with histories of life-threatening anaphylaxis, and discrepant results may be found between clinical assessment of the disease and biological results, especially for drug allergy. In these cases, functional in vitro tests are necessary. The general concept of those tests is to mimic in vitro the contact between allergens and the cells responsible for symptoms (i.e,. those possessing the ability to release histamine). Until recently, basophils were neglected and only considered to be circulating forms of mast cells of minor importance. Furthermore, basophils represent in peripheral blood less than 0.5 percent of total leukocytes, making their purification difficult in clinical laboratories. This lack of satisfactory in vitro protocols has clearly hampered research on basophils for many years [4]. Nevertheless, there is considerable recent evidence that basophils are clinically relevant. Indeed, they are now considered as equivalent to tissue mast cells cells since they play, by themselves, a pivotal role in the immediate allergic reaction [5-8]. Consequently, functional in vitro tests for allergic reactions are focused on circulating basophils. The first approach to basophil functional studies was the histamine release test. However, its clinical benefit has remained controversial due to insufficient sensitivity and specificity [3,9,10]. That is why several groups took advantage of flow cytometry to develop new tools for monitoring basophil activation upon allergen challenge by detecting surface expression of degranulation markers [11-13].
Principle of basophil activation test by flow cytometry
As flow cytometry is a valuable tool for the analysis of many different cell types and can be used to identify specific populations of cells, even when present in low numbers, it seemed to be suitable for the study of allergen-induced basophil degranulation. Identification of cells was initially based both on CD45 expression, a common leukocyte antigen, and on the presence of IgE on the cell surface, since basophils express the high affinity receptor for IgE (FcεRI) [9,14]. In this gated population, cell activation upon allergen challenge was assessed by the expression of CD63 on the membrane [15,16]. CD63 is anchored in the basophilic granule membrane (which contains histamine) and its exposure to the outside of the cells reflects cell degranulation due to fusion between granules and plasma membranes (figure 1). Thus, CD63 expression has been proposed as a reliable means to monitor basophil activation [11-13]. Briefly, whole blood was incubated at 37°C with allergens for 15 minutes. The reaction was stopped on ice, followed by a 30-min staining with antibodies (figure 1). Finally, samples were lysed to eliminate red cells. Basophils expressing both CD45 and surface IgE were then examined for their CD63 expression. The threshold for positivity was determined with the use of a negative control (i.e., whole blood and vehicle without allergen). Results were considered positive when at least 2 sequential allergen dilutions induced greater than than 10% increases in CD63-positive basophils above control values. This kind of protocol has been validated for common allergens by several groups and has shown convincing results [17-24]. The technique has proven to be accessible, rapid (results in less than 1 hour) and requires small amount of blood (< 5 mL, even for assessing several allergens in the same experiment). In our hands, in allergy to muscle relaxants, the results were quite interesting, since we found the sensitivity of the CD63 test was similar to that for specific IgE detection and higher than the one for histamine release test [25]. This confirmed the value of performing the CD63 test rather than histamine release, which is furthermore costly in terms of both reagents and laboratory technician time. In accord with previous studies focusing on different allergens [17-21], this method showed excellent specificity. However, with respect to drug allergy, the main indication for this kind of test, three independent studies reported similar sensitivities ranging between 50 and 64 %, which is not sufficient for clinical usefulness [12,25,26]. In fact, this first approach relied on two important characteristics of basophils which were problematic: recognition through the expression of IgE on their surface (which is known to be highly variable from one patient to another) and the monitoring of their activation by detecting CD63 (which is also expressed to some extent by other activated leukocytes and by activated platelets that may adhere to basophils). This may explain why, when applied to drug allergy, these tests have remained somewhat disappointing in terms of sensitivities [12,25,26]. Consequently, we concluded that an activation marker that is more specific and/or sensitive than CD63 would be desirable.
Figure 1 Principle of the basophil activation test by flow cytometry (triple staining). Basophils are identified on the basis of CD45 expression (fluorescence 3 / Phyco-Cyanine 5) and the presence of IgE or CRTH2/DP2 on their surface (fluorescence 1 / Fluorescein isothiocyanate). Resting basophils do not express CD63 (anchored in the basophilic granule) and weakly express CD203c. The cross-linking of two FcεRI (induced by an allergen or anti-IgE antibodies) provokes the histamine release (and as a consequence the CD63 expression) and the upregulation of CD203c. The rise in CD63 or CD203c expression (measured by fluorescence 2 / Phycoerythrin) before and after allergen challenge reflects thus the basophil activation / degranulation in response to an allergen.
CD203c as a specific marker of activated basophils
CD203c corresponds to a surface antigen expressed on human basophils recently recognized by the monoclonal antibody 97A6 [27]. This antigen, belonging to the type II transmembrane protein family, is a multifunctional ecto-enzyme called ectonucleotide pyrophosphatase phophodiesterase 3 (E-NPP3) [28] that catalyzes the cleavage of a number of molecules including deoxynucleotides and nucleotide sugars [29]. In addition, E-NPP3 contains a somatomedin B-like domain and a cell adhesive motif, but their potential functions remain totally unknown with respect to basophil physiology. Among leukocytes CD203c appears to be selectively expressed on the basophil/mastocytes lineage [27]. To date, no other cells from human peripheral blood have been reported to express this marker. Its expression on basophils is rapidly upregulated after stimulation with the appropriate allergen in patients sensitized to acarids or hymenoptera or after crosslinking of FcεRI with anti-IgE antibodies [28,30]. This suggests that CD203c up-regulation is more or less specific to the crosslinking of FcεRI (figure 2). Hence, as CD203c is rapidly upregulated after allergen challenge, it has been proposed as a new tool for allergy diagnosis [30-33]. We compared basophil activation tests using either CD63 or CD203c in the diagnosis of latex allergy [34] and found that the sensitivity was considerably higher with CD203c (75% compared to 50% with CD63). The improved sensitivity may be due to two factors. First, the recognition of basophils is better with CD203c. Indeed, the identification of basophils using prior protocols relied on a single IgE-labeling, although it is known that FcεRI expression can vary considerably on cell surfaces from one patient to another [35]. This may explain why in some cases basophils were difficult or impossible to identify. The second reason for the improved sensitivity with CD203c is due to its higher expression in activated basophils compared to CD63 in our experiments. In sensitized patients, basophils increased their CD203c levels up to 350 % above control values in response to allergens whereas the increase in CD63 was below 100 %. Similar results were obtained when expressing the results as the percentages of basophils that were CD203c- or CD63-positive. Even with the highest concentration of latex, the mean percentage of CD63-positive basophils was below 20 % while that of CD203c-positive basophils was 48 %, allowing a clear distinction between resting and activated basophils [34]. In conclusion, both easier gating and higher range of activation in response to allergen may contribute to an improvement in the basophil activation test when using CD203c rather than CD63. However, as very few studies concomitantly compared CD203c and CD63, this point remains to be confirmed by additional works dealing with various allergens. Bühring and colleagues in a recent report proposed to use both markers in the same test to increase sensitivity [32]. It is supported by recent evidence showing that CD63 and CD203c overexpression depend on different stimulatory pathways [36,37]. It is to note that some novel basophil-activation markers (CD13, CD107a, CD164) have been very recently identified [37]. They have to be further investigated in clinical studies either by their own or in combination with CD63 or CD203c.
Figure 2 CD203c expression in whole blood before and after basophil activation. Ungated leukocytes are shown as a biparametric representation on the basis of side scatter characteristics (SSC, y-axis) and CD203c (x-axis). Left histogram depicts resting cells, basophils express low levels of CD203c (some of them are not distinguishable from lymphocytes and monocytes). Right histogram depicts cells after anti-IgE challenge, activated basophils are easily recognized on the basis of their high CD203c expression.
CRTH2/DP2 as a new marker for basophil recognition
Finally, the last drawback of the previously described protocols remained the use of an anti-IgE reagent to identify basophils. Because of its selective expression on cells associated with Th2 responses (Th2 lymphocytes, eosinophils and basophils), CRTH2 (chemoattractant receptor-homologous molecule expressed on Th2 cells)/DP2 has been proposed and validated as the most reliable tool for the detection of circulating human Th2 cells [38,39]. CRTH2 is also termed DP2 since it corresponds to the second receptor of prostaglandin D2 [40,41]. As CRTH2 is highly expressed on basophils, we hypothesized that it could improve the basophil activation test by facilitating basophil recognition. Consequently, we developed a new three-colour flow cytometric protocol (PE-CD203c / FITC-CRTH2 / PC5-CD3) for monitoring allergen-induced basophil activation. First results were encouraging: CRTH2 staining allowed CRTH2-expressing cells (eosinophils, basophils and Th2 lymphocytes) to easily be distinguished from other cells in samples of whole blood (figure 3). On the basis of light scattering, eosinophils were easily excluded from the analysis (figure 3). Basophils could then readily be distinguished from Th2 lymphocytes on the basis of CD3, staining, as this marker is not present on basophils (figure 3). Finally, on this gated population of basophils (low light scatterings, CRTH2+ and CD3-), modulation of CD203c after allergen challenge was monitored as described in the former protocol (figure 4). To validate this protocol, 18 subjects were included in a preliminary study [42]. Patients were allergic to either latex (k82) or Dermatophagoïdes pteronyssinus (d1), had a suggestive clinical history, positive skin test and/or specific IgE ≥ class III. Healthy donors, from our laboratory, were not known to be allergic and presented total IgE < 100 kU/L. In terms of clinical interpretation, sensitivity and specificity were 88% and 100%, respectively [40]. CRTH2 staining was an excellent means to identify basophils and we confirmed our earlier observations of a wide range of CD203c expression in response to allergen in tehse cells. In terms of basophil recovery, we compared our CRTH2-staining protocol with 2 others protocols using either anti-IgE or anti-CD123 (IL-3 receptor). In all patients and healthy individuals, we found more basophils (up to 50 % in certain patients) with the CRTH2-staining protocol, illustrating its superiority with respect to basophil recovery. To conclude, the easy recognition of basophils and the reliable assessment of their activation make this protocol the most reliable tool for investigating basophil activation by flow cytometry. It may constitute a critical step for the interlab standardization of this kind of test. Lastly, since CRTH2 is also a marker of Th2 cells and eosinophils, it may become a promising tool for flow cytometry, providing a direct overview of cells involved in "Th2 diseases" such as allergy.
Figure 3 Identification of CRTH2 expressing cells by flow cytometry. Left histogram : ungated leukocytes biparametric representation on the basis of side scatter characteristics (SSC, Y axis) and FITC-CRTH2 (X axis). Two CRTH2 expressing cell populations are easily distinguishable: the one with high light scatterings corresponds to the eosinophil population; the second one (gating region: A) comprises Th2 lymphocytes and basophils. Right histogram: cells from the gating region (A) expressed on the basis of PE-CD203c (X axis) and PC5-CD3 (Y axis) characteristics. Th2 lymphocytes were readily separated from basophils based on their positive CD3 expression while activated basophils express high levels of CD203c without expressing CD3.
Figure 4 Representative increased expression of CD203c after allergen challenge in a patient allergic to Dermatophagoïdes pteronyssinus (d1). Gated CRTH2-positive basophils (after excluding Th2 lymphocytes as described in figure 3) are presented on the basis of CD203c-CRTH2 staining: before stimulation (negative control, upper left dot-plot), after anti-IgE challenge (positive control, upper right) and after allergen challenge at 3 different concentrations (dose-effect response, lower dot-plots). Activated basophils: percentage of basophils expressing CD203c.
Perspectives in pharmacological studies
Until recently, due to the very low number of circulating basophils in humans, pharmacological studies on these cells were difficult to perform. This required large amount of blood and / or lengthy purification procedures that may induce nonspecific activation. By the use of flow cytometry, the effects of different compounds on basophils may be examined in unfractionated human blood cells. Recently, we have been able to demonstrate that among various eicosanoids, prostaglandin D2 was by far the most potent activator of basophils, inducing CD203c and CD11b elevation [37]. This response was mediated by the DP2 receptor / CRTH2 as it was shared by selective agonists of this receptor. As previously observed in eosinophils [39], the interaction of prostaglandin D2 with the DP1 receptor limited the activation of basophils by this prostaglandin. This suggested that the balance between DP1 and DP2 receptors may be crucial in determining the magnitude of basophil responses during allergic processes since prostaglandin D2 is known to be involved in allergic diseases and asthma. Using a similar approach, Heinemann et al. [43] examined the effects of various chemokines on human basophils and demonstrated a different pattern of chemokine receptor usage than those described for eosinophils and monocytes.
These studies illustrate that it is now possible to perform pharmacological and drug screening studies by flow cytometry. This approach could be very useful in assessing the possible risks of inducing anaphylactoid or pseudo-anaphylactoid reactions when developing new molecules. To this end, one important task for the future will be to extend these kinds of protocols to animal models although, to our knowledge, there is no available information on CD203c in animals and monoclonal antibodies directed against human CD203c do not cross-react with other species [32].
Conclusion
After several improvements, the basophil activation test (using either CD203c or CD63 as activation marker) has become a robust and reliable test for in vitro investigations of immediate allergy, complementary to other existing in vitro tests. It is suitable for experimental and pharmacological studies as well as allergy diagnosis in clinical practice. There is now a crucial need for inter-laboratory standardization in clinical decision-making. Each allergen has to be assessed one by one to determine its optimal concentration (i.e., inducing maximal activation in vitro) as well as the definition of the threshold for positivity (using ROC analysis) since the use of an arbitrary cut-off value is likely not suitable for all allergens. The present challenge is to take advantage of the availability of improved methods to perform multicenter studies using a standardized protocol.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RB participated as investigator and is the main author of the article.
ALD participated in drafting the manuscript.
GM was project leader and participated in the design of the different studies and drafting the manuscript.
Acknowledgements
We thank the technical staff of the flow cytometry unit – Immunology lab (J. Baudot, C. Fernandez, MA. Guinand, MC. Gutowski) at the Lyon-Sud University Hospital, Pr. J. Bienvenu (Immunology lab, Lyon-Sud University Hospital) for supporting our work on basophil activation over the years, G. Bouvier and C. Canino (Immunotech, Marseille, France) for kindly providing anti-CRTH2 antibodies.
==== Refs
Kay AB Allergy and allergic diseases (part 1) N Engl J Med 2001 344 30 7 11136958 10.1056/NEJM200101043440106
Kay AB Allergy and allergic diseases (part 2) N Engl J Med 2001 344 109 13 11150362 10.1056/NEJM200101113440206
Hamilton RG Adkinson NF In vitro assays for the diagnosis of IgE-mediated disorders J Allergy Clin Immunol 2004 114 213 225 15316492 10.1016/j.jaci.2004.06.046
Falcone FH Haas H Gibbs BF The human basophil: a new appreciation of its role in immune responses Blood 2000 96 4028 38 11110670
Gauchat JF Henchoz S Mazzei G Aubry JP Brunner T Blasey H Life P Talabot D Flores-Romo L Thompson J Induction of human IgE synthesis in B cells by mast cells and basophils Nature 1993 365 340 43 7690905 10.1038/365340a0
Kasaian MT Clay MJ Happ MP Garman RD Hirani S Luqman M IL-4 production by allergen-stimulated primary cultures: Identification of basophils as the major IL-4 producing cell type Int Immunol 1996 8 1287 97 8918698
Devouassoux G Forster B Scott LM Metcalfe DD Prussin C Frequency and characterization of antigen-specific IL-4 and IL-13 producing basophils and Tcells in peripheral blood of healthy and asmathic subjects J Allergy Clin Immunol 1999 104 811 19 10518826
Galli SJ Mast cells and basophils Curr Opin Hematol 2000 7 32 9 10608502 10.1097/00062752-200001000-00007
Gane P Pecquet C Lamblin P Abuaf N Leynadier F Rouger P Flow cytometric evaluation of human basophils Cytometry 1993 14 344 8 8472612
Demoly P Lebel B Messaad D Sahla H Rongier M Daures JP Godard P Bousquet J Predictive capacity of histamine release for the diagnosis of drug allergy Allergy 1999 54 500 6 10380783 10.1034/j.1398-9995.1999.00020.x
Sainte-Laudy J Sabbah A Vallon C Guerin JC Analysis of anti-IgE and allergen induced human basophil activation by flow Cytometry. Comparison with histamine release Inflamm Res 1998 47 401 8 9831325 10.1007/s000110050351
Abuaf N Rajoely B Ghazouani E Levy DA Pecquet C Chabane H Leynadier F Validation of a flow cytometric assay detecting in vitro basophil activation for the diagnosis of muscle relaxant allergy J Allergy Clin Immunol 1999 104 411 8 10452764
Monneret G Gutowski MC Bienvenu J Detection of allergen-induced basophil activation by expression of CD63 antigen using flow cytometric method Clin Exp Immunol 1999 115 393 6 10193408 10.1046/j.1365-2249.1999.00841.x
Gane P Pecquet C Crespeau H Lambin P Leynadier F Rouger P Flow cytometric monitoring of allergen induced basophil activation Cytometry 1995 19 361 365 7796701 10.1002/cyto.990190411
Knol EF Mul FP Jansen H Calafat J Roos D Monitoring human basophil activation via CD63 monoclonal antibody 435 J Allergy Clin Immunol 1991 88 328 38 1716273
Fureder W Agis H Sperr WR Lechner K Valent P The surface membrane antigen phenotype of human blood basophils Allergy 1994 49 861 5 7709996
Pâris-Köhler A Demoly P Persi L Lebel B Bousquet J Arnoux B In vitro diagnosis of cypress pollen allergy by using cytofluorimetric analysis of basophils (Basotest) J Allergy Clin Immunol 2000 105 339 45 10669856 10.1016/S0091-6749(00)90085-X
Sainte-Laudy J Sabbah A Drouet M Lauret MG Loiry M Diagnosis of venom allergy by flow cytometry. Correlation with clinical history, skin tests, specific IgE, Histamine and Leukotriene C4 release Clin Exp Allergy 2000 30 1166 71 10931125 10.1046/j.1365-2222.2000.00863.x
Sanz ML Sanchez G Gamboa PM Vila L Uasuf C Chazot M Dieguez I De Weck AL Allergen-induced basophil activation: CD63 cell expression deteced by flow cytometry in patients allergic to Dermatophagoides pteronyssinus and Lolium perenne Clin Exp Allergy 2001 31 1007 13 11467990 10.1046/j.1365-2222.2001.01122.x
Saporta M Kamei S Persi L Bousquet J Arnoux B Basophil activation during pollen season in patients monosensitized to grass pollen Allergy 2001 56 442 5 11350309 10.1034/j.1398-9995.2001.056005442.x
Ebo DG Hagendorens MM Bridts CH Schuerwegh AJ De Clerck LS Stevens WJ In vitro allergy diagnosis: should we follow the flow? Clin Exp Allergy 2004 34 332 339 15005724 10.1111/j.1365-2222.2004.01891.x
Torres MJ Padial A Mayorga C Fernandez T Sanchez-Sabate E Cornejo-Garcia JA Antunez C Blanca M The diagnostic interpretation of basophil activation test in immediate allergic reactions to betalactams Clin Exp Allergy 2004 34 1768 75 15544603 10.1111/j.1365-2222.2004.02110.x
Erdmann SM Sachs B Kwiecien R Moll-Slodowy S Sauer I Merk HF The basophil activation test in wasp venom allergy: sensitivity, specificity and monitoring specific immunotherapy Allergy 2004 59 1102 9 15355470 10.1111/j.1398-9995.2004.00624.x
Ebo DG Hagendorens MM Bridts CH Schuerwegh AJ De Clerck LS Stevens WJ Flow cytometric analysis of in vitro activated basophils, specific IgE and skin tests in the diagnosis of pollen-associated food allergy Cytometry 2005 64 28 33 15668987 10.1002/cyto.b.20042
Monneret G Benoit Y Debard AL Gutowski MC Topenot I Bienvenu J Monitoring of Basophil activation using CD63 and CCR3 in allergy to muscle relaxant drugs Clin Immunol 2002 102 192 9 11846462 10.1006/clim.2001.5156
Sanz ML Gamboa PM Antepara I Uasuf C Vila L Garcia-Aviles C Chazot M De Weck AL Flow cytometric basophil activation test by detection of CD63 expression in patients with immediate-type reactions to betalactam antibiotics Clin Exp Allergy 2002 32 277 86 11929494 10.1046/j.1365-2222.2002.01305.x
Buhring HJ Simmons PJ Pudney M Muller R Jarrossay D van Agthoven A Willheim M Brugger W Valent P Kanz L The monoclonal antibody 97A6 defines a novel surface antigen expressed on human basophils and their multipotent progenitors Blood 1999 94 2343 56 10498606
Buhring HJ Seiffert M Giesert C Marxer A Kanz L Valent P Sano K The basophil activation marker defined by antibody 97A6 is identical to the ectonucleotide pyrophosphatase/phosphodiesterase 3 Blood 2001 97 3303 5 11342463 10.1182/blood.V97.10.3303
Bollen M Gijbers R Ceulemans H Stalmans W Stefan C Nucleotide pyrophosphatase/phophodiesterases on the move Crit Rev Biochem Mol Biol 2000 35 393 432 11202013
Platz IJ Binder M Marxer A Lischka G Valent P Bühring HJ Hymenoptera-venom-induced upregulation of the basophil activation marker ecto-nucleotide pyrophosphatase/phosphodieterase 3 in sensitized individuals Int Arch Allergy Immunol 2001 126 335 42 11815741 10.1159/000049531
Hauswirth AW Natter S Ghannadan M Majlesi Y Schernthaner GH Sperr WR Buhring HJ Valenta R Valent P Recombinant allergens promote expression of CD203c on basophils in sensitized individuals J Allergy Clin Immunol 2002 110 102 9 12110828 10.1067/mai.2002.125257
Bühring HJ Streble A Valent P The basophil-specific ectoenzyme E-NPP3 (CD203c) as a marker for cell activation and allergy diagnosis Int Arch Allergy Immunol 2004 133 317 329 15031605 10.1159/000077351
Kahlert H Cromwell O Fiebig H Measurement of basophil-activating capacity of grass pollen allergens, allergoids and hypoallergenic recombinant derivatives by flow cytometry using anti-CD203c Clin Exp Allergy 2003 33 1266 72 12956749 10.1046/j.1365-2222.2003.01756.x
Boumiza R Monneret G Forissier MF Savoye J Gutowski MC Powell WS Bienvenu J Marked improvement of the basophil activation test by detecting CD203c instead of CD63 Clin Exp Allergy 2003 33 259 265 12580920 10.1046/j.1365-2222.2003.01594.x
Kinet JP The high-affinity IgE receptor (FcεRI): From physiology to pathology Ann Rev Immunol 1999 17 931 72 10358778 10.1146/annurev.immunol.17.1.931
Hennersdorf F Florian S Jakob A Baumgartner K Sonneck K Nordheim A Biedermann T Valent P Buhring HJ Identification of CD13, CD107a, and CD164 as novel basophil-activation markers and dissection of two response patterns in time kinetics of ige-dependent upregulation Cell Res 2005 15 325 35 15916720
Monneret G Boumiza R Gravel S Cossette C Bienvenu J Rokach J Powell WS Effects of prostaglandin D2 and 5-lipoxygenase products on the expression of CD203c and CD11b by basophils J Pharmacol Exp Ther 2005 312 627 34 15388786 10.1124/jpet.104.074823
Cosmi L Annunziato F Galli MIG Maggi RME Nagata K Romagnani S CRTH2 is the most reliable marker for the detection of circulating human type 2 Th and type 2 T cytotoxic cells in health and disease Eur J Immunol 2000 30 2972 2979 11069080 10.1002/1521-4141(200010)30:10<2972::AID-IMMU2972>3.0.CO;2-#
Nagata K Tanaka K Ogawa K Kemmotsu K Imai T Yoshie O Abe H Tada K Nakamura M Sugamura K Takano S Selective expression of a novel surface molecule by human Th2 cells in vivo J Immunol 1999 162 1278 1286 9973380
Hirai H Tanaka K Yoshie O Ogawa K Kenmotsu K Takamori Y Ichimasa M Sugamura K Nakamura M Takano S Nagata K Prostaglandin D2 selectively induces chemotaxis in T helper type 2 cells, eosinophils, and basophils via seven-transmembrane receptor CRTH2 J Exp Med 2001 193 255 261 11208866 10.1084/jem.193.2.255
Monneret G Gravel S Diamond M Rokach J Powell WS Prostaglandin D2 is a potent chemoattractant for human eosinophils that acts via a novel DP receptor Blood 2001 98 1942 1948 11535533 10.1182/blood.V98.6.1942
Bouvier G Debard AL Gendt L Canino C Bienvenu J Monneret G A new three-colour flow cytometric combination (CD203C/CRTH2/CD3) for monitoring allergen-induced basophil activation Cytometry 2004 59A 134
Heinemann A Hartnell A Stubbs VE Murakami K Soler D LaRosa G Askenase PW Williams TJ Sabroe I Basophil responses to chemokines are regulated by both sequential and cooperative receptor signalling J Immunol 2000 165 7224 7233 11120855
|
15989690
|
PMC1190199
|
CC BY
|
2021-01-04 16:36:25
|
no
|
Clin Mol Allergy. 2005 Jun 30; 3:9
|
utf-8
|
Clin Mol Allergy
| 2,005 |
10.1186/1476-7961-3-9
|
oa_comm
|
==== Front
Clin Pract Epidemiol Ment HealthClinical Practice and Epidemiology in Mental Health : CP & EMH1745-0179BioMed Central 1745-0179-1-121607639510.1186/1745-0179-1-12Short reportThe prevalence of psychiatric disease in the significant others of patients with known mood and anxious disease Tavormina Giuseppe [email protected] Salvatore [email protected] Antonio [email protected] President of "Psychiatric Studies Centre" ("Cen.Stu.Psi."), piazza Portici, 11, 25050 Provaglio di Iseo (BS), Italy2 Consultant of Psychiatric Communities "Sol.Co.", Bergamo, Italy3 Mental Health Dept., Hospital of Castelfranco Veneto (TV), Italy2005 2 8 2005 1 12 12 23 5 2005 2 8 2005 Copyright ©2005 Tavormina et al; licensee BioMed Central Ltd.2005Tavormina et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Background -
Information about the Significant Others (S.O.) of 530 patients with mood and anxious spectrum disorders has been tabulated in this multicentre, retrospective, clinical observational study in order to learn the prevalence of the same mood and/or anxious spectrum diseases in the S.O. of the patients.
Methods -
The 530 outpatients (of age range from 18 to 70 years) with mood and anxious spectrum disorders have been treated by the authors, observed for a seven year period (from January 1995 until May 2003). The patients live in 16 different Italian provinces, but are predominantly from Lombardia and Veneto.
Mood disease (includes substance abuse) was present in 72% of the patients and anxious disease was present in 28% (DSM-IV diagnoses based upon clinical interviews).
The S.O. (various heterosexual long-term relationships) of each patient was interviewed for this study to establish a DSM-IV diagnosis of any psychiatric disease that might be present.
In cases in which the patient had no S.O. or in which information about the S.O. was unavailable, that information was collected. As data was collected, 10 item report was completed for each patient and the respective S.O.
Results -
Patients had an S.O. with a similar mental disease to their own in 41% of cases; only 16% of the patients chose their S.O. with no mental disease; 18% of the patients did not have any S.O. and in 26% of the cases the health of the S.O. was unknown.
Conclusion -
In this multicentre, retrospective, clinical observational study, the corresponding Significant Others of 530 patients with mood and anxious spectrum disorders presented with a high percentage of similar disease to the patients. These findings suggest that it may be appropriate to counsel our patients with these diseases to encourage their respective S.O. to undergo a psychiatric evaluation for possible treatable disease: the first objective of an S.O. is preventive care, secondarily the well-being of the partner may improve the treatment outcomes for the patient.
Furthermore, eventual studies could demonstrate whether the disease of S.O. precedes couple life (therefore a pre-existent cognitive functioning set out for the partner's choice) and whether it might stem from a difficult interpersonal relationship or chronic stress reaction to a life event.
significant othersmood spectrumanxious spectrumlife eventsstress
==== Body
Background
Information about the Significant Others (S.O.) of 530 patients with mood and anxious spectrum disorders has been tabulated in this multicentre, retrospective, clinical observational study in order to learn the prevalence of the same mood and/or anxious spectrum diseases in the S.O. of the patients.
Methods
The 530 outpatients (of age range from 18 to 70 years) with mood and anxious spectrum disorders have been treated by the authors, observed for a seven year period (from January 1995 until May 2003). [5]. The patients live in 16 different Italian provinces, but are predominantly from Lombardia and Veneto. [6].
Mood disease (including substance abuse) was present in 72% of the patients and anxious disease was present in 28% (DSM-IV diagnoses based upon clinical interviews). [1].
The S.O. (various heterosexual long-term relationships) of each patient was interviewed for this study to establish a DSM-IV diagnosis of any psychiatric disease that might be present. Psychiatric diagnoses were considered to be either "certain" or "probable" for the S.O. The opinions of "certain diagnosis" of S.O., "probable diagnosis" and "any mental disease" were derived from direct knowledge of the S.O. or from the family and personal anamnestic data taken from the patients. In cases in which the patient had no S.O. or in which information about the S.O. was unavailable, that information was collected as detailed below.
As data was collected the following 10 item report was completed for each patient and the respective S.O.:
1- Patient diagnosis: mood spectrum;
2- Patient diagnosis: anxious spectrum;
3- Certain diagnosis of S.O.: mood spectrum;
4- Certain diagnosis of S.O.: anxious spectrum;
5- Probable diagnosis of S.O.: mood spectrum;
6- Probable diagnosis of S.O.: anxious spectrum;
7- Total diagnoses of S.O. (certain + probable);
8- S.O. without any mental disease;
9- Unknown health of S.O.;
10- No S.O.
Results
Patients had an S.O. with a similar mental disease to their own in 41% (217/530 pt.) of cases (percentage data obtained from the item n° 7). Only 16% (85/530 pt.) of the patients chose their S.O. with no mental disease (percentage data obtained from the item n° 8).
18% (95/530 pt.) of the patients did not have any S.O. (item 10). In 26% (138/530 pt.) of the cases the health of the S.O. was unknown (item 9).
The sum of patients of items 9 and 10 together (44%) led us to recognize another relationship: if we considered only all the patients with known health of the S.O., we found that 72% of patients had an S.O. with a similar disease to their own.
The Table 1 summarizes all the items and the respective percentages for the total data.
Table 1 Summarise of all followed items with obtained percentage
1 – Patient diagnosis: mood spectrum 2 – Patient diagnosis: anxious spectrum
383 patients (72%) 147 patients (28%)
3 – Certain diagn. of S.O.: mood spectr. 4 – Certain diagn. of S.O.: anxious spectr.
80 (16%) 35 (6%)
5 – Probable diagn. of S.O.: mood spect. 6-Probable diagn. of S.O.: anxious spect.
62 (12%) 39 (7%)
7-Tot. diagn. of S.O. (certain + probable) 8 – Any mental disease of S.O.
216 (40%) 83 (16%)
9 – Unknown health of S.O. 10 – No Significant Others
138 (26%) 93 patients (18%)
Summarise of all followed items with obtained percentage
Conclusion
In this multicentre, retrospective, clinical observational study, the corresponding Significant Others of 530 patients with mood and anxious spectrum disorders presented with a high percentage of similar disease to the patients. These findings suggest that it may be appropriate to counsel our patients with these diseases to encourage their respective S.O. to undergo a psychiatric evaluation for possible treatable disease: the first objective of an S.O. is preventive care, secondarily the well-being of the partner may improve the treatment outcomes for the patient. [2,4].
Furthermore, the above mentioned results can also highlight a possible new direction for psychodynamic evaluation of couples. Eventual studies could demonstrate whether the disease of S.O. precedes couple life (therefore a pre-existent cognitive functioning set out for the partner's choice) and whether it might stem from a difficult interpersonal relationship or chronic stress reaction to a life event. [3].
==== Refs
American Psychiatric Association Diagnostic and Statistical Manual of Mental Disorders, 4 edition (DSM-IV) 1994 Washington: American Psychiatric Association
Barter JJ Talbott SW Primary prevention in psychiatry: state of the art 1986 Washington: American Psychiatric Press
Cooke DJ Hole D The aetiological importance of stressful life events Br J Psychiatry 1983 143 397 400 6626860
Mitchell ARK Psychiatrists in primary health care settings Br J Psychiatry 1985 147 371 79 4075024
Tavormina G Corea S Sirianni P Tavormina M Mood or anxious diseases of the significant others of patients with mood or anxious diseases World J Biol Psychiatry 2001 2 suppl 1 207
Tavormina G Corea S Citron A Mood or anxious diseases of the significant others of psichiatric patients European Psychiatry 2004 19 suppl 1 186
|
16076395
|
PMC1190200
|
CC BY
|
2021-01-04 18:01:01
|
no
|
Clin Pract Epidemiol Ment Health. 2005 Aug 2; 1:12
|
utf-8
|
Clin Pract Epidemiol Ment Health
| 2,005 |
10.1186/1745-0179-1-12
|
oa_comm
|
==== Front
Environ HealthEnvironmental Health1476-069XBioMed Central London 1476-069X-4-101592708510.1186/1476-069X-4-10ResearchLongitudinal assessment of PCBs and chlorinated pesticides in pregnant women from Western Canada Jarrell John [email protected] Siu [email protected] Russ [email protected] Howard [email protected] Department of Obstetrics and Gynecology, University of Calgary, 1430 29th ST NW, Calgary, AB T2N 2T9, Canada2 Department of Environmental Health, Harvard School of Public Health, Landmark Center, 3rd Floor East, 401 Park Drive, Boston, MA, 02215, USA2005 1 6 2005 4 10 10 6 1 2005 1 6 2005 Copyright © 2005 Jarrell et al; licensee BioMed Central Ltd.2005Jarrell et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Maternal exposures to organochlorines prior to pregnancy are considered a risk to neonatal welfare, specifically in relation to neurocognitive functions. There is growing interest in the evaluation of maternal blood testing as a marker for fetal exposure as well as the variable geographic distribution of these priority chemicals.
Methods
Three hundred and twenty-three women in the second trimester of pregnancy entered the study at a prenatal clinic providing genetic counselling information. Subjects who had an indication for genetic amniocentesis based on late maternal age were eligible to participate. Two hundred and thirty-eight completed an environmental questionnaire. A sample of amniotic fluid was taken for karyotype analysis in 323 women and blood samples during pregnancy (209), at birth (105) and from the umbilical cord (97) and breast milk (47) were also collected. These samples were tested for 29 PCB congeners and organochlorine pesticides.
Results
The concentrations of PCB 153 in these media were relatively low in relation to other studies. Σ PCBs measurements in samples taken during the second trimester of pregnancy, at birth and in the umbilical cord were strongly correlated. Specific measurements of PCB 153 and PCB 180 among those subjects with completed sampling of blood samples from mothers and cord samples were significantly correlated. The concentrations of PCBs and pesticides did not differ in relation to prior spontaneous abortion history. There were no organochlorines present in the amniotic fluid at the current level of quantification.
Conclusion
Pregnant women from the Western Canada region of Calgary, Alberta are exposed to relatively low concentrations of organochlorines. Measurement of maternal blood during the second trimester of pregnancy can reliably estimate the fetal exposure to PCBs. This estimate is reliable for Group 2 and 3 PCBs as well as PCB 153 and PCB 180. The amniotic fluid does not contain measurable concentrations of pesticides and PCBs under the conditions of the levels of quantification.
==== Body
Background
Polychlorinated biphenyls (PCBs) are ubiquitously present in the ecosystem and have been reported in the tissues of many animals and human population groups [1]. They are persistent and biomagnify in biota such that humans and predators at the top of the food chain experience the highest concentrations. Banned production of organochlorines in Canada and the United States in the 1970's has had an important effect in reducing exposure to these chemicals but because of the large residual quantities, release from storage and the limited rates of environmental degradation, they are still considered present in the environment. There is increasing interest in the geographic distributions of these chemicals and their potential impact on human health[2,3]
Maternal exposure to PCBs and organochlorine pesticides is an area of intense research. This is due to the potential long term effects of organochlorine exposure in the fetus and newborn[4]. Potentially severe adverse health effects during pregnancy have included preterm labor and intrauterine growth restriction in association with DDE, but the effects occurred at levels identified during the period 1959–1966[5]. These concentrations are substantially above exposure levels generally observed among current populations [6].
Postnatal adverse events have focused on the known neurotoxic effects of polychlorinated biphenyls [7]. Several studies have implicated exposure to these chemicals with impaired intellectual function in children [8-10]. Importantly the findings have demonstrated modest but significant effect sizes [10]. The potential mechanisms underlying these findings is unknown but recently dose-response relationships have been found between the concentrations of PCBs among members of the Oswego study and increases in response inhibition errors as well as reduced size of the splenium of the corpus callosum [11]. Another proposed mechanism of toxicity in the developing brain is the potential for altered thyroid function, a known contributor to impaired intelligence among newborns [12-14]. There is now evidence the actions of PCBs may reflect a direct effect of the chemical on the action of thyroid hormone in the absence of altered hormone concentrations [15]
The severity of the potential adverse central nervous system effects in children and the possibility that early intrauterine exposure to organochlorines is a determinant of impaired intelligence indicates a need for further exposure information. There is little information related to the longitudinal patterns of PCB levels in women during the pregnancy. In some cases there was no correlation between maternal and cord samples [16], although others have reported that there is a strong correlation [17]. The measurement of serum concentrations over the course of the pregnancy and delivery appears not to have been explored in detail previously although the ability to evaluate fetal exposure during pregnancy is an important objective, particularly because of the susceptibility of the developing nervous system.
This study was undertaken to define the concentrations of PCBs and organochlorine pesticides in amniotic fluid, maternal blood collected during the second trimester, and at the time of birth, cord blood and breast milk among a cohort of pregnant women from Calgary, Alberta, Canada. The inter-relationships of the concentrations in these compartments were of interest to determine if maternal blood levels measured during pregnancy could serve as a proxy for fetal exposure. Further, we were particularly interested in the concentrations of PCBs in the Calgary region compared to other studies that had evaluated neurocognitive development in local children. Finally, there was an interest in evaluating the concentrations of PCB 153 in light of its proposed use for comparing across environmental epidemiologic studies of PCB toxicity [18].
Methods
Pregnant women attending a prenatal counselling session at Foothills Hospital in the Calgary Health Region, Calgary, Alberta were approached to participate in this project. Approval of the project was obtained from the University of Calgary Ethics Committee. Three hundred and twenty-three subjects were enrolled after their eligibility to participate was determined. They were required to be seeking genetic counselling for the purpose of age-related indications and not to have another reason for such counselling. Subjects were also required to be 35 years of age at the time of entry
Of those agreeing to participate, two hundred and thirty-eight completed an environmental questionnaire. A sample of amniotic fluid was available from three hundred and twenty-three women. A blood sample at the same time as the amniocentesis during the second trimester of pregnancy was available from 209 women. A blood sample at the time of delivery was made available in 105 women. Cord blood samples were collected from 97 women and a sample of breast milk was collected from 47 women after birth and during the puerperium from home. Clinical information related to the pregnancy and delivery was collected at the three Calgary hospitals and three hospitals outside the Calgary health region. All subjects were given the special tubes and phlebotomy supplies to submit various blood and breast milk samples and contacted on a regular basis for the completion of the sampling. The nursing staff on each delivery suite was oriented to the project to gain maximal compliance with the study.
The blood samples were collected by the regional laboratory in specially prepared glass tubes, centrifuged and the serum transferred to the Centre for Toxicology, University of Calgary for analysis. A similar approach was used for amniotic fluid. It should be noted that subjects were not required to fast for the blood samples, either during pregnancy or at birth.
The samples were assayed at the Centre for Toxicology, University of Calgary. They underwent sample clean-up and extraction by solid phase extraction techniques. Details of the methods have been previously reported [19,20]The PCBs and most pesticides were measured with GC/negative chemical ionization spectrometry (GC/NCIMS). Some pesticides were measured with GC/electron ionization mass spectrometry (GC/EIMS). The Centre participated in the Intercomparison Program for Toxicological Analysis in Biological Materials, University of Erlangen-Nuremberg, and the performance had been good
In amniotic fluid the PCBs measured were as follows: PCB- 70, 74, 77, 87, 99, 101, 105, 118, 128, 138, 151, 153, 156, 169, 170, 180, 183, 187, 191, 194, 205, 206, 208, 209. The pesticides measured were: Aldrin, p, p'-DDE, p, p'-DDT, dieldrin, endosulfan I and II, endrin, heptachlor, hexachlorobenzene, hexachlorocyclohexanes, α-,β-,γ-, hexachloroethane, methoxychlor, mirex, pentachlorobenzene, 1,2,3-trichlorobenzene, 1,2,4-trichlorobenzene, 1,2,3,4-tetrahlorobenzene. These analytes were also measured in serum and breast milk with some minor changes
The limit of quantification (LOQ) varied among analytes as well as matrices. For example, the LOQ for most PCBs was 0.01 ng/ml in all matrices. The LOQ for most other industrial chemicals was 0.05 ng/ml, however there were exceptional situations. For example, the LOQ for p, p'-DDT was 0.5 ng/ml. In order to evaluate the measurements of Σ PCBs and Group 2 and 3 PCBs, 1/2 of the level of detection was used in analysis. This may result is some bias in the analysis as only PCB 153, 138, 180 and 170 had detection rates >90% in maternal blood during pregnancy. There were similar patterns of detection rates for the PCBs 153, 138 and 103 in maternal blood at birth and PCBs 153 in cord blood and PCB 118, 153, 138, 187, 183, 156, 180, 170 and 194 in breast milk. The presence of bias in the use of this technique is appreciated. Other approaches such as the use of designation of non-detected samples as zero values was not used because the values are called non-ignorable data that may also result bias if the samples are simply excluded. Alternative imputation analysis approaches may provide more accurate methods for pregnancy related PCBs but will need to be confirmed in comparison to other studies [21].
The lipid content of serum was calculated by the sum of cholesterol, triglycerides and phospholipids. These analytes were measured by enzymatic methods using a chemistry analyzer. The lipid content of milk was measured by a gravimetric method.
Statistical analysis
The data from the clinical charts, analysis of chemical results and the environmental questionnaires were assembled in SPSS. Only normally distributed data were evaluated using parametric analysis (paired t-test and Pearson Correlation). Parametric analysis was undertaken wherever possible. The individual analytes were not normally distributed and were normalized by logarithmic function and adjusted to account for non-detection of samples. All PCB concentrations were highly skewed. Samples were corrected for non-detection (0.5*level of quantification), lipid adjusted and log transformed for a normal distribution. Because the relationships of maternal blood to cord blood are key in this study and the fetal blood is lower in lipid content, PCBs levels were evaluated as ng/ml and as ng/g lipid. Specific analysis was done on PCB 153 because of the relatively high concentrations observed. Samples also were aggregated by measuring Σ PCBs as well as Σ PCBs in Group 2 and Group 3 of those proposed by Wolff et al. [22]. For this study, Group 2 included PCBs 74, 118, 156, 138 and 170. Group 3 included PCBs 99, 153, 180 and 183. Data concerning the concentrations of HCB and DDE were treated in a similar fashion for non-detection and normalization.
To explore changes in the concentrations in pregnancy, paired t-tests were undertaken between all possible groupings of samples on the normalized data. Although there were sufficient pairings to evaluate differences in the pair groups, it is important to note these comparisons do not exactly reflect the same cohort over the four periods of study because of missing data. In order to evaluate the correlation of the results, and bearing in mind the desire to study the maternal blood during pregnancy and at birth as a potential marker of fetal exposure, the non-lipid adjusted samples were normalized for comparison purposes. The non-detection samples were adjusted to reflect 0.5 the level of quantification.
To evaluate the potential association of exposures to spontaneous abortion, the concentrations were assessed in relation to clinical history by one way analysis of variance.
Results
Demographics
The subjects were all from the Calgary Health Region, located in southern Alberta, Canada during the period from an entry in 2001 to 2003 when the last delivery occurred. A total of 323 subjects were approached and 315 subjects completed the consent form and formally entered into the study. From this group, 308 provided samples of amniotic fluid for karyotype studies as well as the measurement of the organochlorines. The balance of subjects decided not to have the amniocentesis undertaken based upon information from the Genetics Counselling Clinic. There were 209 blood samples collected during the second trimester around the time of the amniocentesis. There were 105 blood samples analyzed from the mothers at the time of giving birth. There were 97 samples of cord blood and 47 samples of breast milk collected. The breast milk samples were collected at varying times in the postpartum period.
There were 203 pairs of amniotic fluid and maternal blood during the time of amniocentesis and 85 pairs of samples that were obtained from maternal blood at birth and cord samples. There were 23 subjects in which all samples were collected in all tissues compartments.
The demographic and reproductive histories of the subjects are presented in Table 1. Consistent with the inclusion criteria, the age of the subjects was 39.0+/- 0.1 (Mean+/- SEM). During the previous year, 15.1% of the subjects identified themselves as smokers and 8% continued to smoke during the pregnancy. Most of these smokers reported smoking less than 20 cigarettes per day (see Additional file: 1).
A specific diet was reported by 66 subjects including weight loss (13.4%), sodium free (3.4%), fat free (8.4%), diabetic (0.4%), weight gain (0.4%) and vegetarian (1.7%). Eighty five individuals reported a significant weight loss on the preceding year and the average amount was reported to be 7.3 pounds.
Prior medical history was significant for the regular use of prescription medication (27%), Anesthetic exposure (18%), and infertility (12%). There was an infrequent rate of prior cancer (2.1%), skin disease (2.9 %) and liver disease (1.7%).
The most common chemical exposure in the previous year was to paint (50.5%) and solvents (32.8%). There was a reported exposure of this cohort to pesticides in 22%, herbicides 22%, fungicides 5.9% and dry cleaning chemicals 6.5%.
Of the cohort, there were only 36 primigravidas. Sixteen percent reported one induced abortion and 6% had had 2 or more. There were 21% of the subjects with one, 6.3% with two and 6% with three or more spontaneous abortions. Two percent reported an ectopic pregnancy.
Of the pregnancies under review, there were 291 births, of which there was a complication rate of pregnancy of pregnancy induced hypertension in 8.2%, premature labor in 3.4%, gestational diabetes in 5.2% a birth weight less than 2500 g in 4.8% and a Caesarean section rate of 27%.
The sex ratio of the cohort of infants was 1.13 (male/female). Karyotype indicated two cases of Klinefelter syndrome and 2 cases of trisomy that proceeded to delivery. Other anomalies resulted in pregnancy termination.
Concentrations of the Organochlorine Chemicals
Amniotic Fluid
None of the PCBs or pesticides was measurable in the amniotic fluid. It should be noted the amniotic fluid measurement was done after centrifugation and the amniocytes removed for karyotyping.
Total PCBs and Pesticides
The concentrations of all available samples of PCB 153, Σ PCBs, Group 2 and Group 3 PCBs and the pesticides DDE and HCB (lipid adjusted) are presented in Additional file: 2. Statistical analysis was performed for Σ PCB samples using a paired t-test of normalized data. In relation to ΣPCBs, samples collected from women during pregnancy did not differ from those collected at birth (p > 0.05) but the levels in maternal blood were higher than samples from the cord and breast milk (p < 0.001). Breast milk was found to be higher than cord blood ΣPCBs (p = 0.008) all tests after lipid-adjustment.
A similar pattern was identified for Group 2 and Group 3 PCBs. There were no differences between blood collected during pregnancy and the samples taken at birth (n.s.). There were higher concentrations in blood taken during pregnancy and at birth over cord blood (p < 0.01) and breast milk (p < 0.01). Breast milk was higher than cord blood for Group 2 (p < 0.05) but not Group 3 PCBs.
PCB 153 concentrations were not different between maternal blood during pregnancy and at birth although they were all significantly higher than breast milk.
This increase in the concentrations of PCBs in the cord blood samples was potentially a consequence of the relatively low concentrations of lipid in the fetal circulation. The levels of non-lipid adjusted levels of PCBs are shown in Additional file: 2. The lipid concentrations of the tissue compartments were as follows (mean +/- S dev.): maternal blood during pregnancy – 0.71+/-0.13, at birth – 0.74+/-0.16, cord blood- 0.21+/-0.06 and breast milk – 2.83+/-1.33 g/100 ml plasma.
The pattern of PCBs among non-lipid adjusted groups is shown in Additional file: 2. There was no difference in the maternal samples during pregnancy compared to the cord blood samples. There were significantly higher levels of both Groups in maternal blood during pregnancy and at birth compared to cord blood, consistent with the Σ PCB level pattern (p < 0.001)
The concentrations of pesticides are presented in Additional file: 2. There were no differences in lipid adjusted HCB concentrations of maternal blood during pregnancy and at birth when compared to cord blood (n.s.) There were significantly lower concentrations of lipid adjusted HCB in breast milk when compared to all other groups (p < 0.001). The same pattern of result was observed for lipid adjusted DDE samples.
Without lipid adjustment there was a significant difference in HCB in maternal blood during pregnancy and at birth compared to cord blood (p < 0.001) (see Additional file: 2). HCB during pregnancy and at birth were significantly less than breast milk samples without lipid adjustment (p < 0.001). The same pattern was observed for non-lipid adjusted DDE samples.
PCB Correlations
The relationships of the Σ PCBs are closely correlated in all of the tissues as indicated in Table 3. There are several notable relationships with very high correlation coefficients. Non-lipid adjusted, normalized Σ PCBs during pregnancy were found to highly correlate with levels at birth (r = 0.770, p < 0.01) and in cord blood (r = 0.498, p < 0.01) but not with breast milk (n.s.). Σ PCBs taken at birth also correlated well with cord blood (p < 0.001) but not with breast milk PCBs.
Table 3 Correlation of ΣPCBs In Human Tissues
ΣPCB DP ΣPCB AB ΣPCB CB ΣPCB BM
ΣPCB DP r 1.000 0.759 0.357 0.504
p 0.000 0.001 0.001
n 97.000 91.000 40.000
ΣPCB AB r 1.000 0.600 0.325
p 0.000 0.075
n 87.000 31.000
ΣPCB CB r 1.000 -0.012
p 0.954
n 27.000
There were significant correlations between Group 2 and 3 PCBs (Additional file: 3). The most significant correlations were between groups but within tissue compartments. Group 2 PCBs of maternal blood during pregnancy correlated most strongly with Group 3 PCBs of maternal blood during pregnancy (r = 0.755, p < 0.01) but also strongly correlated with Group 2 PCBs at birth (r = 0.628, p < 0.01) and Group 3 PCBs at birth (r = 0.717, p < 0.001); Group 3 PCBs of maternal blood during pregnancy also correlated with Group 2 and 3 at birth and Group 2 and 3 cord blood (all p < 0.001). The Group 2 PCBs at birth correlated well with Group 2 and 3 cord blood cord (r = 0.603, p < 0.001 and r = 0.783, p < 0.001 respectively); Group 3 PCBs at birth also correlated with Group 2 and 3 cord blood (all p < 0.001). There was no correlation between breast milk and the other values.
To explore more rigorously the relationships of PCBs between compartments PCB 153 and PCB 180 were evaluated among all subjects with completed collection of maternal and fetal blood samples. These PCBs were selected because of the relatively high rate of detection of analytes. Cord blood PCB 153 was significantly correlated with maternal blood during pregnancy (r = 0.691, p < 0.001) and maternal blood at birth (r = 0.905, p < 0.001). Cord blood PCB 180 was also significantly correlated with maternal blood during pregnancy (r = 0.511, p < 0.001) and maternal blood at birth (r = 0.830, p < 0.001).
The correlations between the tissue compartments and HCB and DDE are presented in Additional file: 4. Specifically, there were significant correlations for both HCB and DDE between both the samples during pregnancy and at birth when compared to cord blood (all p < 0.001).
Clinical Variables and Concentrations of PCBs and Pesticides
There were no differences in the concentrations of PCBs or pesticides among women reporting one or two prior spontaneous abortions compared to those individuals who had had no spontaneous abortion. It should be noted this was not the primary objective of the study and the clinical outcomes are part of secondary analysis. There were no significant differences in concentrations of PCBs and pesticides among the subjects reporting medical conditions with the exception that those individuals reporting infertility had significantly higher concentrations of HCB. There were no differences in these concentrations among women reporting exposure to herbicides, pesticides, fungicides or insecticides compared to those reporting no such exposure in the previous year. Similarly, there was no significant association of a particular diet (weight loss, diabetic, salt free, low fat) with the measurement of PCB and pesticides in this cohort of subjects. Although there was a dose related increase in the mean concentrations of Σ PCBs in maternal blood during pregnancy in relation to fish consumption, the relationship was not significant.
Discussion
The results of this study are reassuring and support the observations that indicate levels of PCBs are relatively low in western Canada compared to other studies in previous years among pregnant women in different countries or regions of Canada. Specifically, using PCB 153 as a marker as recommended [23], it would appear the women in Calgary are at the lower range of values among studies that evaluated neurocognitive outcomes of children. The meta analysis of ten studies indicated a range of the 25th to the 75th percentiles of PCB 153 to range from approximately 20 to 800 ng/g lipid [18]. In this study, the arithmetic mean concentrations of PCB 153 were 21.16, 25.71 and 18.22 ng/g lipid in the maternal blood during pregnancy, at birth and cord blood respectively, indicating a relatively low exposure level in relation to the previous reports.
There is evidence that the maternal serum during the second trimester of pregnancy can be considered a biomarker for the exposure of the fetus in utero as determined by cord blood levels and the significant correlations among these tissue compartments. These observations have been made previously at higher PCB concentrations [3]
The concentrations of this study are consistent with other reports of the concentrations of PCB 153, HCB and DDE in the cord samples. Specifically, there is close agreement with the findings in southern Quebec, Canada and New Bedford, USA for all three chemicals measured on a wet weight basis [24]. Additionally the combination of PCB 138, 153 and 180 in the New Bedford study was approximately twice the concentrations of the same PCBs in the current study (mean, median and standard deviation: 0.84, 0.6, 1.14 ng/ml respectively), and agreed more closely with the Σ PCBs which contained PCBs 74, 118, 138, 153, 187, 183, 156, 180, 170. These differences may reflect a greater exposure in New Bedford than in southern Alberta. The differences in cord concentrations from other reports of higher levels may be a reflection of the declining concentrations in recent years since the chemicals were banned [25-28]. Decreases in tissue concentrations have been reported in Northern Canada between 1994 and 2001 [29]. The lower concentrations may also reflect the findings of a population study of healthy women attending a genetics counselling service that was unselected for dietary intake of foods containing PCBs [30].
Absence of these chemicals in the amniotic fluid at the level of quantification was observed but not unanticipated because of the low lipid concentration, an assessment of the amniocytes may have isolated analytes [31, 32]. In this study the cells were not available owing the need for karyotyping.
There were a number of important findings from this cohort that again indicate the relatively low concentrations did not exercise a measurable biological effect of the prior health of the women. For example, there were no differences in the concentrations of PCBs or pesticides among women reporting prior spontaneous abortion. The sample size may be too small to detect a significant increase. The higher concentrations of HCB during pregnancy among those reporting infertility may simply reflect the lack of prior pregnancy-related loss of body burden and is not interpreted as a causal function of infertility. There were no significant differences in the concentrations of PCBs and pesticides among the subjects reporting current or previous serious medical conditions or exposure to pesticides, fungicides, herbicides or other common chemicals such as dry cleaning fluid in the previous year. Similarly, there was no significant association of diet with the measurement of Σ PCB and pesticides in this cohort of subjects. Although there was an increase in the mean concentrations of Σ PCBs in relation to daily fish consumption, the relationship of diet was not significant.
Conclusion
In summary, it would appear that pregnant subjects from Calgary Alberta are exposed to PCBs and pesticides as are most individuals but the extent of exposure appears to be low in relation to other published studies from other areas of Canada or other countries. There is no evidence of exposure to pesticides or PCBs in the amniotic fluid under the current measurement limitations of quantification. Measurement of maternal blood during the second trimester of pregnancy reliably estimates the maternal blood levels at birth as well as fetal exposure to PCBs and organochlorines as measured by cord blood at these relatively low levels. The evaluation of subjects who had complete blood sampling indicated specifically that the cord blood PCB 153 and PCB 180 are reliably predicted by a measure of maternal blood during the second trimester of pregnancy.
List of Abbreviations
DP: maternal blood taken during the first trimester of pregnancy
AB: Maternal blood taken at the time of birth
C Blood taken from the cord at birth
BM: Breast milk
DDE: Dichlorodiphenyldichloroethylene
HCB: Hexachlorobenzene
DDT: Dichlorodiphenyltrichloroethane
PCBs: Polychlorinated biphenyls
PCB 70 2,3',4',5-Tetrachlorobiphenyl
PCB 74 2,4,4',5-Tetrachlorobiphenyl
PCB 77 3,3',4,4'-Tetrachlorobiphenyl
PCB 87 2,2'3,4,5'-Pentachlorobiphenyl
PCB 99 2,2',4,4'5-Pentachlorobiphenyl
PCB 101 2,2'4,5,5'-Pentachlorobiphenyl
PCB 105 2,3,3',4,4'-Pentachlorobiphenyl
PCB 118 2,3',4,4',5-Pentachlorobiphenyl
PCB 128 2,2',3,3',4,4'-Hexachlorobiphenyl
PCB 138 2,2',3,4,4',5'-Hexachlorobiphenyl
PCB 151 2,2',3,5,5'6-Hexachlorobiphenyl
PCB 153 2,2',4,4',5,5'-Hexachlorobiphenyl
PCB 156 2,3,3'4,4',5-Hexachlorobiphenyl
PCB 169 3,3'4,4',5,5'-Hexachlorobiphenyl
PCB 170 2,2',3,3',4,4',5-Heptachlorobiphenyl
PCB 180 2,2',3,4,4',5,5'-Heptachlorbiphenyl
PCB 183 2,2'3,4,4'5',6-Heptachlorbiphenyl
PCB 187 2,2',3,4',5,5',6-Heptachlorbiphenyl
PCB 191 2,3,3',4,4',5',6-Heptachlorbiphenyl
PCB 194 2,2',3,3',4,4'5,5'-Octachlorbiphenyl
PCB 205 2,3,3',4,4',5,5',6-Octachlorobiphenyl
PCB 206 2,2',3,3',4,4',5,5',6-Nonachlorbiphenyl
PCB 208 2,2',3,3',4,5,5',6,6'-Nonachlorobiphenyl
PCB 209 Decachlorobiphenyl
Competing interests
The author(s) declare that they have no competing interests.
Authors' Contributions
JJ was original co-investigator and undertook patient recruitment, clinical data acquisition and data analysis in relation to the analysis of human tissue samples, and drafted the article. SC was principal investigator of the original proposal, undertook sample measurement and analysis and participated in revising the draft for consideration; RH participated in data analysis and revisions of the article; HH participated in data analysis and revisions of the article. All authors read and approved the final manuscript.
Supplementary Material
Additional file 1
A “.doc file” describing the characteristics of the women enrolled in the study.
Click here for file
Additional file 2
A “.doc file” that describes concentrations of the organochlorines isolated from
the women during the study period.
Click here for file
Additional file 3
A “.doc file” describes the strength of the comparisons of the levels of certain
groups of PCBs in the women’s tissues.
Click here for file
Additional file 4
A “.doc file that describes the strength of the comparisons of the levels of
certain pesticides in women’s tissues.
Click here for file
Acknowledgements
All authors contributed equally to this work. The authors wish to express appreciation for the funding received from Toxic Substances Research Initiative, Health Canada and Environment Canada. The assistance of Cheryl Swaby and Margaret Sevcik in the completion of this work is appreciated.
==== Refs
Aguilar A Borrell A Reijnders PJ Geographical and temporal variation in levels of organochlorine contaminants in marine mammals Mar Environ Res 2002 53 425 452 12054104 10.1016/S0141-1136(01)00128-3
Baibergenova A Kudyakov R Zdeb M Carpenter DO Low birth weight and residential proximity to PCB-contaminated waste sites Environ Health Perspect 2003 111 1352 1357 12896858
Ayotte P Muckle G Jacobson JL Jacobson SW Dewailly E Assessment of pre- and postnatal exposure to polychlorinated biphenyls: lessons from the Inuit Cohort Study Environ Health Perspect 2003 111 1253 1258 12842782
Taylor PR Lawrence CE Hwang HL Paulson AS Polychlorinated biphenyls: influence on birthweight and gestation Am J Public Health 1984 74 1153 1154 6433730
Longnecker MP Klebanoff MA Zhou H Brock JW Association between maternal serum concentration of the DDT metabolite DDE and preterm and small-for-gestational-age babies at birth Lancet 2001 358 110 114 11463412 10.1016/S0140-6736(01)05329-6
Longnecker MP Klebanoff MA Zhou H Brock J Associaton between maternal serum concentration of the DDT metabolite DDE and preterm and small-for-gestational-age babies at birth Lancet 2001 358 110 114 11463412 10.1016/S0140-6736(01)05329-6
Rogan WJ Gladen BC Hung KL Koong SL Shih LY Taylor JS Wu YC Yang D Ragan NB Hsu CC Congenital poisoning by polychlorinated biphenyls and their contaminants in Taiwan Science 1988 241 334 336 3133768
Rogan WJ Gladen BC Study of human lactation for effects of environmental contaminants: the North Carolina Breast Milk and Formula Project and some other ideas Environ Health Perspect 1985 60 215 221 3928347
Jacobson JL Jacobson SW Intellectual impairment in children exposed to polychlorinated biphenyls in utero N Engl J Med 1996 335 783 789 8703183 10.1056/NEJM199609123351104
Darvill T Lonky E Reihman J Stewart P Pagano J Prenatal exposure to PCBs and infant performance on the fagan test of infant intelligence Neurotoxicology 2000 21 1029 1038 11233749
Stewart P Fitzgerald S Reihman J Gump B Lonky E Darvill T Pagano J Hauser P Prenatal PCB exposure, the corpus callosum, and response inhibition Environ Health Perspect 2003 111 1670 1677 14527849
Zoeller RT Dowling AL Vas AA Developmental exposure to polychlorinated biphenyls exerts thyroid hormone-like effects on the expression of RC3/neurogranin and myelin basic protein messenger ribonucleic acids in the developing rat brain Endocrinology 2000 141 181 189 10614638 10.1210/en.141.1.181
Zoeller TR Dowling AL Herzig CT Iannacone EA Gauger KJ Bansal R Thyroid hormone, brain development, and the environment Environ Health Perspect 2002 110 Suppl 3 355 361 12060829
Langer P Kocan A Tajtakova M Petrik J Chovancova J Drobna B Jursa S Pavuk M Koska J Trnovec T Sebokova E Klimes I Possible effects of polychlorinated biphenyls and organochlorinated pesticides on the thyroid after long-term exposure to heavy environmental pollution J Occup Environ Med 2003 45 526 532 12762077
Gauger KJ Kato Y Haraguchi K Lehmler HJ Robertson LW Bansal R Zoeller RT Polychlorinated Biphenyls (PCBs) Exert Thyroid Hormone-like Effects in the Fetal Rat Brain but Do Not Bind to Thyroid Hormone Receptors Environ Health Perspect 2004 112 516 523 15064154
Akiyama K Ohi G Fujitani K Yagyu H Ogino M Polychlorinated biphenyl residues in maternal and cord blood in Tokyo metropolitan area Bull Environ Contam Toxicol 1975 14 588 592 812580 10.1007/BF01683376
Covaci A Jorens P Jacquemyn Y Schepens P Distribution of PCBs and organochlorine pesticides in umbilical cord and maternal serum Sci Total Environ 2002 298 45 53 12449328 10.1016/S0048-9697(02)00167-5
Longnecker MP Wolff MS Gladen BC Brock JW Grandjean P Jacobson JL Korrick SA Rogan WJ Weisglas-Kuperus N Hertz-Picciotto I Ayotte P Stewart P Winneke G Charles MJ Jacobson SW Dewailly E Boersma ER Altshul LM Heinzow B Pagano JJ Jensen AA Comparison of polychlorinated biphenyl levels across studies of human neurodevelopment Environ Health Perspect 2003 111 65 70 12515680
Dmitrovic J Chan SC Chan SHY Analysis of pesticides and PCB congeners in serum by GC/ms with SPE sample cleanup Toxicol Lett 2002 134 253 258 12191885 10.1016/S0378-4274(02)00173-X
Dmitrovic J Chan S Analysis of PCB congeners in humaan milk by GC/NCIMS after sample cleanup by solid phase extraction J Chromatogr 2002 778 147 158 10.1016/S0378-4347(01)00447-9
Lubin JH Colt JS Camann D Davis S Cerhan JR Severson RK Bernstein L Hartage P Epidemiologic evaluation of measuremet data in the presence of detection limits Environ Health Perspect 2004
Wolff MS Camann D Gammon M Stellman SD Proposed congener groupings for epidemiological studies Environ Health Perspect 1997 105 13 14 9074863
Korrick SA Altshul LM Tolbert PE Burse VW Needham LL Monson RR Measurement of PCBs, DDE, and hexachlorobenzene in cord blood from infants born in towns adjacent to a PCB-contaminated waste site J Expo Anal Environ Epidemiol 2000 10 743 754 11138666
Janousek V Krift J Malbohan M Cibula D Lukas W Lammers W Huisman M Boersma ER van der Paauw CJ Volegaar EF Winneke G Schmidt E Steingruber HJ Cord blood levels of potentially neurotoxic pollutants (polychlorinated biphenyls, lead and cadmium) in areas of Prague, (Czech Republic) and Katowice (Poland). Comparison with reference values in the Netherlands Cent Eur J Public Health 1994 2 73 76 7697024
Koopman-Esseboom C Huisman M Weisglass-Kuperus N van der Paauw CG Tuinstra JGM Boersma ER Sauer PJJ PCBs and dioxn levels in plasma and human milk of 418 Dutch women and their infants. Predictive value of PCB congener level in maternal plasma for feta and infant exposur to PCBs and dioxin Chemosphere 1994 28 1721 1732 10.1016/0045-6535(94)90428-6
Bush B Snow J Koblintz R Polychlorobiphenyl (PCB) congeners, p,p'-DDE and hexachlorobenzene in maternal and fetal cord blood from mothers in Upstate New York Arch Environ Contam Toxicol 1984 13 517 527 6435547 10.1007/BF01056331
Hagmar L Becher G Heikkila A Frankman O Dyremark E Schutz A Ahlborg UG Dybing E Consumption of fatty fish from the Baltic Sea and PCB in whole venous blood, plasma and cord blood from delivering women in the Aland/Turku archipelago J Toxicol Environ Health A 1998 53 581 591 9572157 10.1080/009841098159042
Dallaire F Dewailly E Laliberte C Muckle G Ayotte P Temporal trends of organochlorine concentrations in umbilical cord blood of newborns from the lower north shore of the St. Lawrence river (Quebec, Canada) Environ Health Perspect 2002 110 835 838 12153768
Chan HM A database for environmental contaminants in traditional foods in northern and Arctic Canada: development and applications Food Addit Contam 1998 15 127 134 9602917
Stenback F Ojala A Histochemical properties of amniotic fluid cells Acta Cytol 1969 13 389 394 4242043
Messeri G Curiel P Caldini AL Lipid content of amniotic fluid cells Clin Chim Acta 1980 100 201 207 7353308 10.1016/0009-8981(80)90267-3
|
15927085
|
PMC1190201
|
CC BY
|
2021-01-04 16:36:32
|
no
|
Environ Health. 2005 Jun 1; 4:10
|
utf-8
|
Environ Health
| 2,005 |
10.1186/1476-069X-4-10
|
oa_comm
|
==== Front
Environ HealthEnvironmental Health1476-069XBioMed Central London 1476-069X-4-131604279810.1186/1476-069X-4-13ResearchRisk of high blood pressure in salt workers working near salt milling plants: A cross-sectional and interventional study Haldiya Kripa Ram [email protected] Murli Lal [email protected] Raman [email protected] Habibulla N [email protected] Deputy Director Senior Grade, Desert Medicine Research Centre (ICMR), Jodhpur, 342005, India2 Deputy Director, Desert Medicine Research Centre (ICMR), Jodhpur, 342005, India3 Deputy Director, Desert Medicine Research Centre (ICMR), Jodhpur, 342005, India4 Director, National Institute of Occupational Health (ICMR), Meghani Nagar, Ahmedabad, 380816, India2005 25 7 2005 4 13 13 24 5 2005 25 7 2005 Copyright © 2005 Haldiya et al; licensee BioMed Central Ltd.2005Haldiya et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Workers working close to salt milling plants may inhale salt particles floating in the air, leading to a rise in plasma sodium, which, in turn, may increase the blood pressure and the risk of hypertension.
Methods
To test the above hypothesis, occupational health check-up camps were organized near salt manufacturing units and all workers were invited for a free health examination. The workers who worked with dry salt in the vicinity of salt milling plants were defined as "non-brine workers," while those working in brine pans located far away from milling plants were defined as "brine workers." Blood pressure (BP) was measured during each clinical examination. In all, 474 non-brine workers and 284 brine workers were studied.
Results
Mean systolic blood pressure of non-brine workers (122.1 ± 13.3 mm Hg) was significantly higher than that of brine workers (118.8 ± 12.8 mm Hg, p < 0.01). Mean diastolic blood pressure of non-brine workers (71.5 ± 10.4 mm Hg) was significantly higher than that of brine workers (69.7 ± 9.4 mm Hg, p = 0.02). The prevalence of hypertension was significantly higher in non-brine workers (12.2%) than in brine workers (7.0%, p = 0.02). Nineteen salt workers were monitored while they used face masks and spectacles, for six days. Systolic, as well as diastolic, blood pressure of these workers began declining on the third day and continued to decline on the fourth day, but remained stationary up to the sixth day. The concentration of salt particles in the breathing zone of these workers was 376 mg/m3 air.
Conclusion
Inhalation of salt particles in non-brine workers may be an occupational cause of increased blood pressure.
==== Body
Introduction
There is an abundance of scientific evidence demonstrating a direct relation between salt intake and blood pressure (BP) [1]. Many animal studies [2], large population-based studies [1,3-6], epidemiological studies [7-9], meta-analyses of clinical trials [10-12], and randomized controlled trials [13,14] have shown that BP is directly related to salt intake. People's occupations also have varying impact on their BP [15-19]. Salt workers involved in the process of manufacturing, milling, and packing of salt are exposed to salt via their environment. Since most salt milling plants in India are not fully enclosed, salt particles float in the air in the vicinity of the workers. These workers may therefore inhale considerable amounts of salt during working hours. These salt particles may be inhaled and therefore absorbed in the airway surface epithelium [20-24] or the lungs [25]. These same fine particles are also able to translocate from the lungs into the circulatory system [26]. Inhaled salt particles may be carried via a continuous upward mucocilliary current on the airway surface to throat, where they can be swallowed. This is likely to increase the plasma sodium level, which in turn may increase the BP [27] and the risk of hypertension in the exposed workers. However, this problem may be completely preventable. This hypothesis was tested through a cross-sectional and experimental study involving salt workers; the results are presented and discussed below.
Methods
A cross-sectional study was conducted among salt workers of the Sambhar, Nawa, and Rajas salt-manufacturing sites of Rajasthan, which are at about 150 km. from Jaipur, the capital of Rajasthan. Occupational health check-up camps were held at these three sites, under the Project on Prevention and Control of Occupational Health Hazards Among the Salt Workers, sponsored by the Ministry of Health, Government of India. This project was approved by the Scientific Advisory Committee of the National Institute of Occupational Health, Ahmedabad, India. The procedures followed were in accordance with the Helsinki Declaration. The camps were organized at Sambhar, Nawa, and Phalodi, in collaboration with owners of salt manufacturing units and the Department of Salt, Government of India. Each camp lasted 5 days. All the workers from nearby salt manufacturing units were invited for a free health examination. Workers who were absent on the dates of the health camp were not included in the study.
The aim of the study was explained to the subjects. Their age (in years), sex, detailed occupational history (including exact nature of job and duration of working in salt industry) were recorded on schedules especially designed for occupational health examinations.
After obtaining the informed consent, the clinical examination was carried out by one of the authors, who did not measure the blood pressure. After the subject had rested for five minutes in a supine position, the blood pressure was measured in the right arm using digital blood pressure equipment (Omron T-4). The cuff size was 25 cm × 13 cm. Three readings were taken by the trained field investigators, under supervision of another author. The first two readings were to familiarize the subjects with the process and the third reading was recorded for analysis. Prior to the camps, the field investigators were trained by the authors for fifteen days in measuring blood pressure. Body weight and height were measured by another trained field investigator. Height was measured in centimeters, using an anthropometric rod, while the subject stood erect on a flat platform.
Eight-hundred-and-ninety-one salt workers attended the camps, and three blood pressure measurements were taken in 875 workers. The workers who were involved in crushing, grinding, milling, packing, and loading salt, and who did not work with brine, were defined as non-brine workers. These workers worked in the vicinity of salt milling plants. Workers who worked with brine pans for the purpose of crystal reshuffling and raw salt heaping were defined as brine workers; their site of work was far away from the salt milling plants. The workers who worked as non-brine workers for some time and also worked as brine workers on some other days were excluded from analysis. Workers who were involved in only administrative and other related activities were also excluded from the analysis.
Hypertension was defined as systolic blood pressure more than 139 mmHg and/or diastolic blood pressure 90 mmHg or above. Body mass index was calculated as [Weight in Kg/(Height in meters)2]. Systolic and diastolic blood pressure was compared in the brine workers and non-brine workers. Student's t-test and Chi square test were used to determine the statistical significance of the differences.
Since mean systolic BP, mean diastolic BP, and prevalence of hypertension were found to be significantly higher in non-brine workers as compared to brine workers, an intervention study was carried out to test the hypothesis that exposure of non-brine workers to salt particles floating in environment may contribute to rise in their blood pressure. For this purpose, thirty-three non-brine workers, working at or close to salt milling plants, who volunteered to participate in the study, were registered. We explained the study hypothesis and provided them with face masks and spectacles with plain glasses. The masks were dust guards made of poly vinyl chloride, containing a disposable filter cartridge of nitrocellulose. In one of our earlier studies, we found that these masks could filter 82.8% dust particles of size 10 μm or less [28]. The workers were trained and motivated to use them properly while working, and were observed and followed for six consecutive days. During this period, their resting blood pressure was measured in the supine position, before starting work in the morning. Only nineteen of them regularly attended the worksite and used the face mask and eyeglasses for all six consecutive days, while others were present on some days, but absent on others. Workers were requested to provide urine samples before starting the intervention study and after completing the intervention for more than 3 days. These samples were collected twice a day – once in the morning before starting work and then in the evening, after completion of working hours. Only eight subjects out of 33 workers provided both (morning and evening) urine samples before intervention and only six workers (not the same) provided both urine samples after intervention. These samples were then analyzed for sodium and potassium levels using an AVL electrode electrolyte analyzer (AVL Medical Instruments, Schaffhausen, Switzerland). Additionally, the concentration of salt particles in the air in the environment of the work site was measured by using a respirable dust sampler (Environtech). The dust sampler was placed at two sites namely Sambhar Salts and at Nawa for six days. The particles of 10 μm or more were collected at the bottom of the cyclone of the sampler and those smaller than 10 μm were deposited on the filter paper of the sampler. Volume of total air entering the sampler and weight of particles collected were used to calculate the average concentration of both types of dust in the environment.
Results
Out of 758 salt workers studied, 474 (62.5%) workers were non-brine workers, while 284 (37.5%) were brine workers. The characteristics of the study subjects are depicted in Table 1. These were comparable in brine and non-brine workers. Mean age of male brine workers was 31.8 ± 9.8 years, while male non-brine workers were comparatively a little younger (mean age 29.2 ± 10.0 years.). The mean age of female brine workers (35.1 ± 10.9 years) was not significantly different from that of female non-brine workers (36.5 ± 10.5 years). All workers were 15 years of age or older. The two groups did not show any significant difference in the prevalence of smoking, alcohol use, literacy, income, diet habits, and BMI. However, mean duration of working in the salt industry was lower in non-brine workers than brine workers.
Table 1 Characteristics of study subjects.
Characteristics Brine workers (n = 284) Non-brine workers (n = 474) p value
Age (Years)
Males 31.8 ± 9.8 (n = 238) 29.2 ± 10.0 (n = 398) <0.01*
Females 35.1 ± 10.9 (N = 46) 36.5 ± 10.5 (N = 76) 0.49†
Both Sexes 32.3 ± 10.0 30.4 ± 10.4 0.01*
Gender M/F (%) 83.8/16.2 84.0/16.0 0.97 ‡
Literacy (%) 35.2 43.5 0.03 ‡
Income (Rs. per anum) 17760.9 ± 12858.7 19684.5 ± 13761.4 0.06†
Smokers (%) 33.8 35.9 0.25 ‡
Alcohol users (%) 10.6 11.8 0.17 ‡
BMI Kg/m2 18.9 ± 2.2 18.7 ± 2.5 0.28†
Vegetarians (%) 62.3 67.7 0.15 ‡
Duration of working in salt industry (Years) 11.4 ± 7.2 8.7 ± 6.9 <0.01*
*Difference significant; Student's t-test
† Difference not significant; Student's t-test
‡ Difference not significant; Chi Square test
Mean systolic blood pressure of non-brine workers (122.1 ± 13.3 mmHg) was significantly higher than that of brine workers (118.8 ± 12.8 mm Hg)(p < 0.01). Z-test, as well as the Student's t-test (two-tailed), showed a highly significant difference in both sexes, separately (Table 2). Mean diastolic blood pressure of non-brine workers (71.5 ± 10.4 mm Hg) was significantly higher than that of brine workers (69.7 ± 9.4 mm Hg) (p = 0.01). This was also consistently higher in both sexes.
Table 2 Mean systolic and diastolic blood pressure of brine workers and non-brine workers.
Brine workers Non-brine workers p value
Average systolic BP
Males 119.9 ± 11.7 (n = 238) 122.8 ± 12.4 (n = 398) <0.01*
Females 113.2 ± 16.6 (n = 46) 118.3 ± 16.5 (n = 76) 0.01
Both sexes 18.8 ± 12.8 (n = 284) 122.1 ± 13.3 (n = 474) <0.01*
Average diastolic BP
Males 69.4 ± 9.6 (n = 238) 72.8 ± 10.2 (n = 398) 0.09
Females 71.1 ± 7.8 (n = 46) 75.2 ± 10.3 (n = 76) 0.02*
Both sexes 69.7 ± 9.4 (n = 284) 71.5 ± 10.4 (n = 474) 0.01*
*Difference significant; Z-test and Student's t-test (two-tailed)
Overall, the prevalence of hypertension in salt workers was 10.3%. It was significantly higher in non-brine workers (12.2%) than in brine workers (7.0%) (p = 0.02). The prevalence of hypertension was also consistently higher in non-brine workers than brine workers in different groups, according to age, sex, literacy, income, and body-mass index, duration of working in salt industry, smoking, alcohol use, tobacco chewing, and diet (Table 3).
Table 3 Prevalence of hypertension in brine workers and non-brine workers according to various characteristics.
Characteristics Brine Workers Non-brine Workers
No. Hypertensive cases No. Hypertensive cases
No. % No. %
Age <40 years 208 11 5.3 368 33 9.0
40+ years 76 9 11.8 106 25 23.6
Males 238 18 7.6 398 46 11.6
Females 46 2 4.3 76 12 15.8
Illiterate 184 16 8.7 268 38 14.2
Literate 100 4 4.0 206 20 9.7
Annual income Rs.<18000 179 10 5.6 256 33 12.9
>18000 105 10 9.5 218 25 11.5
BMI <18 Kg/m2 109 3 2.8 185 18 9.7
18+ Kg/m2 175 17 9.7 289 40 13.8
Duration of Work <10 Years 118 5 4.2 305 31 10.2
10+ Years 166 15 9.0 169 27 16.0
Smokers or ex-smokers 190 15 7.9 316 43 13.6
Non-smokers 94 5 5.3 158 15 9.5
Alcohol Users or ex-users 259 17 6.6 430 55 12.8
Non-users 25 3 12.0 44 3 6.8
Tobacco chewing Yes 217 14 6.5 325 40 12.3
No 67 6 9.0 149 18 12.1
Diet Vegetarian 177 15 8.5 321 41 12.8
Mixed 107 5 4.7 153 17 11.1
Total Prevalence 284 20 7.0 474 58 12.2
Results of experimental intervention
Table 4 shows the mean number of working hours, mean number of hours for which masks and glasses were used, and the mean morning blood pressure of nineteen workers who attended the worksite and used face mask and eyeglasses for all six days of intervention. Morning blood pressure was taken before starting their shift. The systolic, as well as diastolic, blood pressure of these workers began declining on the third day and continued to decline on forth day, but remained stationary, each day thereafter (Figure 1). Table 5 shows that the difference in blood pressure between day 1 and day 2 was not significant (for systolic BP p = 0.98 and for diastolic BP p = 0.95), but that between day 2 and day 3 (for systolic BP p = 0.03 and for diastolic BP p = 0.16), as well as between day 3 and day 4, was significant (for systolic BP p = 0.03 and for diastolic BP p < 0.01); again, the decline thereafter was not significant (for systolic BP p = 0.08 & 0.68 and for diastolic BP p = 0.55 & 0.65). Mean urinary sodium in morning samples before the intervention was 265.7 ± 250.8 mmol/L and decreased to 184.6 ± 46.3 mmol/L three days after the intervention. This decline was not statistically significant (p = 0.27). Mean urinary sodium in evening samples before the intervention was 310.8 ± 304.2 mmol/L, as compared to 180.5 ± 41.2 mmol/L three days after the intervention. This decline was also not statistically significant (p = 0.31). Mean concentration of salt particles of a size less than 10 μm (PM 10) was 15 mg/m3 and that of larger particles was 361 mg/m3 air in the breathing zone of these workers, during these six days.
Table 4 Mean working hours, period of use of protective devices and morning blood pressure of workers on the days of intervention (n = 19).
Day of intervention Mean no. of hours Worked Mean no. of hours masks used Mean no. of hours glasses used Mean Systolic Blood Pressure (mm Hg) Mean Diastolic Blood Pressure (mm Hg)
Day 1 6.2 ± 0.5 3.9 ± 1.0 4.8 ± 0.8 127.8 ± 11.1 80.7 ± 8.8
Day 2 10.0 ± 1.4 5.5 ± 1.5 6.5 ± 1.6 127.8 ± 11.8 80.6 ± 12.8
Day 3 9.7 ± 1.8 4.6 ± 1.3 5.2 ± 1.9 123.4 ± 10.3 76.4 ± 8.6
Day 4 7.9 ± 0.5 4.4 ± 1.9 4.8 ± 1.3 117.5 ± 9.9 62.6 ± 7.8
Day 5 9.3 ± 0.9 4.2 ± 1.1 4.5 ± 0.9 113.8 ± 7.0 63.8 ± 8.0
Day 6 9.1 ± 1.3 4.8 ± 1.4 5.2 ± 1.5 114.6 ± 6.5 63.0 ± 5.5
Total 8.7 ± 1.7 4.6 ± 1.5 5.1 ± 1.5
Figure 1 Blood Pressure of non-brine workers declined by use of masks and glasses.
Table 5 Comparison of mean systolic and diastolic blood pressures of the workers on consecutive days of intervention.
Day of intervention Mean Systolic Blood Pressure (mm Hg) p value Mean Diastolic Blood Pressure (mm Hg) p value
Day 1 127.8 ± 11.1 0.98 80.7 ± 8.8 0.95
Day 2 127.8 ± 11.8 80.6 ± 12.8
Day 2 127.8 ± 11.8 0.04* 80.6 ± 12.8 0.16
Day 3 123.4 ± 10.3 76.4 ± 8.6
Day 3 123.4 ± 10.3 0.03* 76.4 ± 8.6 0.001*
Day 4 117.5 ± 9.9 62.6 ± 7.8
Day 4 117.5 ± 9.9 0.08 62.6 ± 7.8 0.55
Day 5 113.8 ± 7.0 63.8 ± 8.0
Day 5 113.8 ± 7.0 0.68 63.8 ± 8.0 0.65
Day 6 114.6 ± 6.5 63.0 ± 5.5
* Difference significant; Student's t-test (two-tailed)
Discussion
In the present study, systolic, as well as diastolic, BP and prevalence of hypertension were found to be higher in non-brine salt workers, who were occupationally exposed to sodium chloride particles in the air of the breathing zone. This is a new observation, though it is in line with the hypothesis that, after being inhaled, salt may be absorbed from respiratory tract [20-24] or the mucocilliary current may transport it to pharynx, where it is swallowed and can then be absorbed from the gastrointestinal tract. Consequent increases in plasma sodium may be responsible for increases in the BP [27]. Differences in urinary sodium, an indicator of sodium intake, and plasma sodium are associated with BP differences of clinical and public health relevance [29]. The exact mechanisms whereby raised plasma sodium increases the BP are not clear. Existing concepts focus on the tendency for an increase in extracellular fluid volume (ECV), but raised plasma sodium increases a transfer of fluid from the intracellular to the extracellular space, and stimulates the thirst center. Accordingly, the rise in plasma sodium is responsible for the tendency for an increase in ECV. Although the change in ECV may have a pressure effect, the associated rise in plasma sodium itself may also cause the BP to rise [27]. Systolic and diastolic BP and the prevalence of hypertension of the non-brine (exposed) workers were compared with the brine salt workers, who were not exposed to salt particles in air. BP is affected by multiple factors, including age, nature of job, socioeconomic status, living standard, nutritional status, smoking habits, and alcohol consumption. Both groups of studied workers did not differ on these parameters (Table 1). However, mean age and mean duration of working in the salt industry (exposure) were lower in non-brine workers, compared to brine workers, but these can be causes of lower BP, rather than of higher BP. The prevalence of hypertension was consistently higher in different subgroups of non-brine workers (Table 3), and this consistency further strengthens the above observation. It can, therefore, safely be concluded that BP and prevalence of hypertension of non-brine workers were higher than brine workers.
To further confirm the hypothesis about probable mechanism involved, an experimental intervention was carried out. The decline in BP while using face masks and spectacles during work again strengthens this hypothesis. The urinary sodium levels also declined after use of masks and glasses for three days, though the decline was not statistically significant (probably because of smaller sample size).
A limitation of the study is that serum sodium levels of the workers involved in the intervention study could not be measured. The total concentration of salt particles in the air was 376 mg/m3. Considering the average tidal volume of 800–1000 ml/breath and respiratory rate of 18–25/min while working, the average worker could inhale 2.60 to 4.51 gm sodium chloride over the course of an eight hour shift. Average use of the mask was 52.9% of working hours, which could have prevented inhalation of about 1.37 to 2.39 gm of salt per day. The exact mechanism by which the decline of 1.37 to 2.39 g of salt intake per day could significantly reduce the BP is not clear. Thus, the results of this intervention study do not fully support the hypothesis that the cause of higher BP and higher prevalence of hypertension in non-brine workers is inhalation of salt particles from the environment. Eye glasses were provided to protect their eyes from salt particles and the study design did not allow us to find out whether these contributed to lowering of BP, though the salt particle sticking on to conjunctiva may also pass along with tears through the naso-lacrimal duct to respiratory tract and further on to the gastrointestinal tract. The psychological effect of using some of the intervention devices expected to reduce BP can also not be ruled out in this study. Further studies on salt workers are needed to elucidate our findings.
List of Abbreviations
BP : Blood Pressure
ECV : Extra Cellular fluid Volume
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
KRH contributed in conception and design, acquisition of data, analysis and interpretation of data and drafting the paper; MLM contributed in acquisition of data, statistical analysis and interpretation of data and drafting the paper; RS contributed in acquisition of data and drafting the paper; and HNS contributed in conception and design and critical evaluation of the data and drafting of the paper. All authors read and approved the final manuscript.
Acknowledgements
Ministry of Health and Family Welfare, Govt. of India had financed the project entitled Prevention and Control of Occupational Health Hazards among the Salt Workers working in Desert areas of Gujarat and Western Rajasthan. Authors are grateful to the Ministry as all data presented in this paper were collected under this project. Authors are also grateful to staff of Salt Department, Government of India who helped in this study. We are thankful to the staff of Hindustan Salts Ltd. in Jaipur, India for their help with this work.
==== Refs
Stamler J The INTERSALT Study: background, methods, findings Am J Clin Nutr 1997 65 626S 642S 9022559
Denton D Weisinger R Mundy NI Wickings EJ Dixson A Moisson P The effect of increased salt intake on blood pressure of chimpanzees Nat Med 1995 1 1009 1016 7489355 10.1038/nm1095-1009
Law MR Frost CD Wald NJ By how much does dietary salt reduction lower blood pressure? I. Analysis of observational data among populations BMJ 1991 302 811 815 2025703
Elliott P Observational studies of salt and blood pressure Hypertension 1991 17 13 18
William MV Frank MS Jamy A Lawrence JA George AB Denise G Effects of Diet and Sodium Intake on Blood Pressure: Subgroup Analysis of the DASH-Sodium Trial Ann Intern Med 2001 135 1019 1028 11747380
Obarzanek E Proschan MA Vollmer WM Moore TJ Sacks FM Appel LJ Individual blood pressure responses to changes in salt intake: results from the DASH-Sodium trial Hypertension 2003 42 459 67 12953018 10.1161/01.HYP.0000091267.39066.72
He J Tell GS Tang YC Mo PS He GQ Relation of electrolytes to blood pressure in men Hypertension 1991 17 378 385 1999367
Frost CD Law MR Wald NJ By how much does dietary salt reduction lower blood pressure? II. Analysis of observational data within populations BMJ 1991 302 815 819 2025704
Intersalt Cooperative Research Group INTERSALT: an international study of electrolyte excretion and blood pressure: results for 24 hour urinary sodium and potassium excretion BMJ 1988 297 319 328 3416162
Cutler JA Follmann D Allender PS Randomized trials of sodium reduction: an overview Am J Clin Nutr 1997 65 643S 651S 9022560
Midgley JP Matthew AG Greenwood CM Logan AG Effect of reduced dietary sodium on blood pressure: a meta-analysis of randomized controlled trials JAMA 1996 275 1590 1597 8622251 10.1001/jama.275.20.1590
Graudal NA Galloe AM Garred P Effects of sodium restriction on blood pressure, renin, aldosterone, catecholamines, cholesterols, and triglyceride: a meta-analysis JAMA 1998 279 1383 1389 9582047 10.1001/jama.279.17.1383
MacGregor GA Sagnella GA Markandu ND Singer DRJ Cappuccio FP Double-blind study of three sodium intakes and long-term effects of sodium restriction in essential hypertension Lancet 1989 2 1244 1247 2573761 10.1016/S0140-6736(89)91852-7
The Trials of Hypertension Prevention Collaborative Research Group Effects of weight loss and sodium reduction intervention on blood pressure and hypertension incidence in overweight people with highnormal blood pressure: the Trials of Hypertension Prevention, Phase II Arch Intern Med 1997 157 657 667 9080920 10.1001/archinte.157.6.657
Chiplonkar SA Agte VV Tarwadi KV Paknikar KM Diwate UP Micronutrient deficiencies as predisposing factors for hypertension in lacto-vegetarian Indian adults J Am Coll Nutr 2004 23 239 247 15190049
Anand MP Prevalence of hypertension amongst Mumbai executives J Assoc Physicians India 2000 48 1200 1201 11280229
Waskiewicz J Zaborski L State of circulatory system in equipment operators working in harbour Bull Inst Marit Trop Med Gdynia 1998 49 87 96 10431650
Tiwai RR Pathak MC Zodpey SP Babar VY Hypertension among cotton textile workers Indian J Public Health 2003 47 34 36 14723294
Hazarika NC Biswas D Narain K Kalita HC Mahanta J Hypertension and its risk factors in tea garden workers of Assam Natl Med J India 2002 15 63 68 12044117
Knowles M Murray G Shallal J Askin F Ranga V Gatzy J Bioelectric properties and ion flow across excised human bronchi J Appl Physiol 1984 56 868 877 6144659
Kilburn KH A hypothesis for pulmonary clearance and its implications Am Rev Respir Dis 1968 98 449 463 5691682
Boucher RC Human airway ion transport (Part 1) Am J Respir Crit Care Med 1994 150 271 281 8025763
Boucher RC Human airway ion transport (Part 2) Am J Respir Crit Care Med 1994 150 581 593 8049852
Knowles MR Boucher RC Mucus clearance as a primary innate defense mechanism for mammalian airways J Clin Invest 2002 109 571 577 11877463 10.1172/JCI200215217
Aswania O Chrystyn H Relative lung and systemic bioavailability of sodium cromoglycate inhaled products using urinary drug excretion post inhalation Biopharm Drug Dispos 2002 23 159 163 12015790 10.1002/bdd.308
Nemmar A Hoylaerts MF Hoet PH Nemery B Possible mechanisms of the cardiovascular effects of inhaled particles: systemic translocation and prothrombotic effects Toxicol Lett 2004 149 243 253 15093270 10.1016/j.toxlet.2003.12.061
de Wardener HE He FJ MacGregor GA Plasma sodium and hypertension Kidney Int 2004 66 2454 2266 15569339 10.1111/j.1523-1755.2004.66018.x
Desert Medicine Research Centre Annual Report 1996 Jodhpur, India
Khaw KT Bingham S Welch A Luben R O'Brien E Wareham N Blood pressure and urinary sodium in men and women: the Norfolk Cohort of the European Prospective Investigation into Cancer (EPIC-Norfolk) Am J Clin Nutr 2004 80 1397 1403 15531692
|
16042798
|
PMC1190202
|
CC BY
|
2021-01-04 16:36:32
|
no
|
Environ Health. 2005 Jul 25; 4:13
|
utf-8
|
Environ Health
| 2,005 |
10.1186/1476-069X-4-13
|
oa_comm
|
==== Front
Environ HealthEnvironmental Health1476-069XBioMed Central London 1476-069X-4-141608349710.1186/1476-069X-4-14ResearchCocaine in surface waters: a new evidence-based tool to monitor community drug abuse Zuccato Ettore [email protected] Chiara [email protected] Sara [email protected] Davide [email protected] Renzo [email protected] Silvia [email protected] Roberto [email protected] Department of Environmental Health Sciences, Mario Negri Institute for Pharmacological Research, Via Eritrea 62, 20157 Milan, Italy2 Department of Biotechnology and Molecular Sciences, University of Insubria, Via Dunant 3, 21100 Varese, Italy2005 5 8 2005 4 14 14 5 5 2005 5 8 2005 Copyright © 2005 Zuccato et al; licensee BioMed Central Ltd.2005Zuccato et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Cocaine use seems to be increasing in some urban areas worldwide, but it is not straightforward to determine the real extent of this phenomenon. Trends in drug abuse are currently estimated indirectly, mainly by large-scale social, medical, and crime statistics that may be biased or too generic. We thus tested a more direct approach based on 'field' evidence of cocaine use by the general population.
Methods
Cocaine and its main urinary metabolite (benzoylecgonine, BE) were measured by mass spectrometry in water samples collected from the River Po and urban waste water treatment plants of medium-size Italian cities. Drug concentration, water flow rate, and population at each site were used to estimate local cocaine consumption.
Results
We showed that cocaine and BE are present, and measurable, in surface waters of populated areas. The largest Italian river, the Po, with a five-million people catchment basin, steadily carried the equivalent of about 4 kg cocaine per day. This would imply an average daily use of at least 27 ± 5 doses (100 mg each) for every 1000 young adults, an estimate that greatly exceeds official national figures. Data from waste water treatment plants serving medium-size Italian cities were consistent with this figure.
Conclusion
This paper shows for the first time that an illicit drug, cocaine, is present in the aquatic environment, namely untreated urban waste water and a major river. We used environmental cocaine levels for estimating collective consumption of the drug, an approach with the unique potential ability to monitor local drug abuse trends in real time, while preserving the anonymity of individuals. The method tested here – in principle extendable to other drugs of abuse – might be further refined to become a standardized, objective tool for monitoring drug abuse.
==== Body
Background
The use of cocaine, one of the most potent and addictive illicit drugs, appears to be increasing in some countries [1-3]. International drug agencies suggest that this should be closely monitored, in particular among young people in urban areas [3]. The trends and magnitude of drug abuse are currently estimated indirectly from general statistics mainly based on population surveys, consumer interviews, medical records, and crime statistics [3,4]. These general indicators, however, may not realistically estimate the phenomenon at the regional level, where specific socio-economic and cultural patterns can strongly influence drug abuse habits and trends. Moreover, since self-reporting of socially censured behavior is likely to be unreliable, the figures obtained by interviewing known or potential users may be underestimates. New methods are therefore needed not only to provide more realistic estimates of illicit drug consumption, but also to promptly detect changes in abuse trends in local populations. Such methods would therefore help social scientists and the authorities to respond to changing habits with appropriate preventive countermeasures, in "real time".
Several studies, including our own, have reported that therapeutic and veterinary drugs excreted by humans and animals end up in the aquatic environment through the sewage system [5-9]. We collected evidence that environmental levels of largely used therapeutic drugs approximately reflect the total amounts consumed by the local population, as calculated from prescription figures [10-12]. Thus, when factors such as the drug's pharmacokinetics and metabolism and the environmental fate of excretion products are appropriately taken into account, the environmental loads (amounts entering the environment over time) of a drug and/or its major metabolites can become indicators of the drug's consumption by the local population. The idea of possibly using "non-intrusive drug monitoring at sewage treatment facilities" "to determine collective drug usage parameters at the community level" was proposed by Daughton [13] in 2001 but, to our knowledge, has never been implemented.
In the present study we tested whether the above approach could be used to estimate the community consumption figures for a common drug of abuse, namely cocaine. As for therapeutic drugs, in fact, excretion products of cocaine consumed in a given area could in principle be trackable in local waste water (WW) and the receiving surface waters (SW). These environmental compartments can be viewed in fact as a sort of transient "depository" for any sufficiently stable compound excreted by the local population. Thus, finding an excretion product of cocaine in WW and SW could be used to help estimate local consumption. Moreover, if monitored regularly, changing drug concentrations in WW or SW could reflect changes in drug use in real time.
In humans, only a small percentage of a cocaine dose is excreted in urine as the parent drug, while a large amount is excreted as benzoylecgonine (BE) [14,15]. BE is in fact the metabolite often measured in urine to obtain evidence of cocaine use in forensic medicine. Therefore, we searched for and measured both cocaine and BE in aquatic environmental samples, but used concentrations of BE to calculate cocaine consumption more accurately (see Methods).
The method used here may obviously – in this first rather unrefined field application – have some intrinsic limitations (discussed below) in the accuracy of collective consumption estimates. Nevertheless, we felt it was worthwhile to test whether or not this evidence-based approach offered a significant improvement over existing indirect methods. After having identified cocaine and BE in the aquatic environment, our main goal was initially to verify how our consumption estimates compared with official figures. We expected our field data on cocaine consumption to give estimates within the range of the official estimates, or perhaps lower, but certainly not higher. In fact, the type of consistent evidence collected from environmental samples and the assumption made for our calculations (see Methods) might possibly lead to under-estimates but hardly over-estimates of the true cocaine consumption. A certain (still unknown) fraction of cocaine excretion products, entering the sewage system from a myriad of scattered inlet points, could in fact be lost and/or degraded before reaching the common sampling site, thus resulting in under-estimates of the true consumption figures. And, again, if we consider that cocaine metabolites sampled from WW and SW cannot reasonably come from sources other than human excretion (apart from sporadic but highly unlikely cases of cocaine disposal into the sewage system or rivers), and that their concentration in flowing waters cannot reflect accumulation, we must once more conclude that we could not have over-estimated true values. With these caveats, we therefore tested this approach on the Italian territory (Figure 1), and compared our findings with official figures for cocaine use in Italy, obtained from surveys of the general population [2].
Figure 1 Sampling sites for cocaine measurement. Map of Italy showing the River Po basin with the site of sampling, and the locations of the urban waste water treatment plants.
Methods
Chemicals and Materials
The reference standards (99% purity) of cocaine and BE were from MacFarlan-Smith Ltd (Edinburgh, UK), and LGC Promochem s.r.l. (Milan, Italy), respectively. The internal standard (IS), salbutamol-D3 (99.1% D) was from CDN Isotopes (Quebec, Canada). Standards were dissolved in methanol at 1 mg/ml and subsequently diluted to 10 ng/μl. Purity of the solutions was checked before each analytical run by HPLC-MS-MS. All solutions were stored at -20°C in the dark. The cartridge used for solid phase extraction was a 3-ml disposable OASIS MCX (60 mg, Waters Corp., Milford, MA).
Sample collection
Composite water samples (pools of five 500-ml samples collected every 30 min) were collected on four different days from the River Po at Mezzano, Pavia (Figure 1). At this sampling site (average flow rate for the period, 743 m3sec-1) the basin's population equivalent is 5.4 × 106. The flow rates were kindly provided by the Ufficio Mareografico ed Idrografico del Po. Water samples were also taken from influent WW at four treatment plants (WWTPs) serving medium-size Italian cities (Cagliari, Latina, Cuneo, and Varese; location shown in Figure 1). Flow rates of the WWTPs were 1.0, 0.36, 0.22 and 0.46 m3sec-1, and population equivalents 270, 140, 45 and 110 × 103, respectively. For each plant, a 24-h, two-liter composite sample was obtained by pooling water collected every 20 min by an automatic sampling device. Water samples were stored at 4°C, and processed within 3 days, to minimize possible degradation.
Solid phase extraction
Cocaine and BE were measured by adapting our method for pharmaceuticals in river water [10]. Water samples (500 ml) were filtered on a glass micro-fiber filter and spiked with 10 ng of the IS. The pH was then adjusted to 2.0 with 37% HCl. Oasis MCX cartridges were conditioned before use by washing with 6 ml methanol, 3 ml MilliQ water and 3 ml water acidified to pH 2. Samples were then passed through the cartridges under vacuum, at a flow rate of 20 ml/min. Cartridges were vacuum-dried for 5 min and eluted with 2 ml methanol, and 2 ml 2% ammonia solution in methanol. The eluates were pooled and dried under an air stream.
Liquid chromatographic separation
Before analysis, samples were re-dissolved in 100 μL acetic acid 0.01% in water (pH 3.5), then centrifuged and transferred into glass vials. Aliquots of 10 μl were injected using an auto sampler. The HPLC system consisted of two Series 200 pumps and a Series 200 auto sampler (Perkin-Elmer, Norwalk, CT). A Luna C8 column 50 mm × 2 mm i.d., 3 μm particle size (Phenomenex, Torrance, CA, USA) was used for the chromatographic separation. The elution started with 100% of eluent A (formic acid 0.1% in water, pH 2) followed by a 10-min linear gradient to 100% of eluent B (acetonitrile), 2-min isocratic elution and a 2-min linear gradient to 100% of eluent A, which was maintained for 6 minutes to equilibrate the column. During the analysis the flow rate was 200 μl/min and the column was kept at room temperature.
Mass spectrometry (MS-MS)
An API 3000 triple quadrupole mass spectrometer (Applied Biosystems – Sciex, Thornhill, Ontario, Canada) was used for quantitative determinations. Analyses were done in ESI positive mode, with a spray voltage of 5.4 kV, orifice skimmer voltages that varied from 30 to 54 V and ring electrode voltages from 180 to 280 V. Data acquisition was performed with multiple reaction monitoring (MRM) of selected fragmentation products of the protonated pseudo-molecular ions (m/z 290 -> 105 and 290 -> 168 for BE, 304 -> 105 and 304 -> 182 for cocaine, 243 -> 151 and 243 -> 169 for the IS). Five-point calibration curves were generated for each compound by injecting 10 μl of 0.01% acetic acid solutions containing known amounts (0–1 ng/μl) of BE and cocaine, and the IS (0.1 ng/μl). Calibration curves run with each batch of samples showed excellent linearity (r2 > 0.998). Instrumental blanks (standard solution with IS only), showed no traces of interfering compounds. Procedural blanks and recoveries were performed using mineral water. Blanks showed no detectable cocaine and BE. Recoveries were >90% for both compounds. The limits of detection were respectively 0.06 and 0.12 ng/liter for BE and cocaine (calculated as the concentration giving a signal-to-noise ratio of 3 in recovery tests). The identity of cocaine and BE, and the absence of interfering compounds, were further confirmed by MS/MS qualitative analyses performed on a LCQ DecaXP Plus (Thermo Electron, Waltham, MA) ion trap mass spectrometer. In this case, chromatographic conditions were the same previously described, while the mass analysis was made by acquiring ESI-MS and MS/MS spectra, corresponding to the pseudo-molecular ions of BE and COC. The relative amounts of the fragment ions of the substances were in accordance (+/-20%) with those of reference standards.
Calculations and assumptions
Given that about half a cocaine dose is excreted in urine as BE, and only a small fraction as the unchanged drug, we used the concentrations of BE in WW or SW to estimate the amounts of cocaine consumed locally. Concentrations of cocaine were useful to verify that the BE/cocaine ratio was stable and in the expected range, thus giving confidence about their source being human consumption. If an unlikely accidental or intentional disposal of a significant amount of cocaine were to occur at any of these sites, the normal BE/cocaine ratio (see Results) would be transiently and markedly altered in favor of cocaine. BE loads (g/day) at each sampling site – calculated from the BE concentration in water (ng/liter) and water flow rate (m3/sec) – were used to estimate the loads of parent cocaine, multiplying by a factor of 2.33. This takes into account the BE/cocaine molar mass ratio (0.954) and the average molar fraction (45%) of a cocaine dose that is excreted as BE, according to different studies [14,15]. Cocaine loads were then related to the local population equivalents (i.e. the number of people served by a WWTP or living in the river's catchment basin), using data from the Italian 14th General Population and Housing Census (2001) [16]. The estimated consumption (g per day per 1000 people) at each site was referred both to the general population and to young adults (15–34 y), since the latter group reportedly includes almost all consumers [2]. The data were also expressed as the number of doses per day per 1000 people, assuming 100 mg as an average dose [1] (the equivalent of four 25-mg "lines" of cocaine).
Results
Cocaine and BE were found in all WW and SW samples tested. Concentrations at the various sampling sites are shown in Table 1. As expected, the parent drug levels were much lower than the metabolite, their ratio in WW samples (0.15 ± 0.03, mean ± SD) being in accordance with the known metabolic fate of cocaine in humans. In the River Po, the cocaine/BE ratio was stable over time but lower than expected (0.05 ± 0.02), suggesting a different pattern of degradation and/or partition for cocaine and BE in WWTP and environmental media.
Table 1 Levels and loads of cocaine and its metabolite (benzoylecgonine, BE) in the River Po and WWTPs.
Levelsa Loads
Cocaine (ng/liter) BE (ng/liter) Cocaine equivalentsb (g/day)
River Po 1.2 ± 0.2c 25 ± 5c 3800 ± 720c
WWTPsd
Cagliari 83 640 130
Cuneo 76 420 30
Latina 120 750 33
Varese 42 390 36
aCocaine and BE were analyzed by HPLC-MS/MS
bCocaine loads were estimated from BE concentrations in the waters (see Methods)
cMean ± SD
dWaste water treatment plant locations
On four different occasions, at the same sampling site, the River Po was found to steadily carry almost 4 kg of cocaine equivalents per day (Table 1). This suggests a total of about 40,000 doses per day, or about seven doses for every 1000 people living in the river's basin. However, considering only young adults, the estimated use reaches 27 doses per day per 1000 people (Table 2). In agreement with these findings, cocaine loads determined at WWTPs gave drug consumption estimates of about 2–7 doses per 1000 people, or 9–26 doses per day per 1000 young adults (Table 2).
Table 2 Local use of cocaine in the River Po basin and medium-size Italian cities, as estimated from BE levels in waters
Estimated local cocaine use
per 1000 people per 1000 young adultsa
g/day no. dosesb/day g/day no. dosesb/day
River Po 0.70 ± 0.13c 7.0 ± 1.3c 2.7 ± 0.5c 27 ± 5c
WWTPsd
Cagliari 0.47 4.7 1.7 17
Cuneo 0.21 2.1 0.9 9
Latina 0.73 7.4 2.6 26
Varese 0.32 3.2 1.4 14
0.44 ± 0.23c 4.4 ± 2.3c 1.7 ± 0.7c 17 ± 7c
a15–34 yr old
b1 dose = 100 mg
cMean ± SD
dWaste water treatment plant locations
Discussion
The method we have initially tested here with cocaine, as a possible new tool to monitor collective consumption of illicit drugs, gave reproducible estimates from the WWTP, confirmed on a larger scale by the River Po data. However, if this method were to be used in general for continuous monitoring of local drug use, it would be preferable to use repeated sampling in a given, well characterized setting. Sampling WW for drug analysis before it enters a TP would avoid changes in drug concentrations due to removal or degradation that might occur within the TPs. Levels of an illicit drug in a major river of a heavily populated area could still serve to help evaluate consumption on a larger than local scale, but only in those cases where a drug is fairly stable in the aquatic environment.
Clearly, the method implemented here needs to be refined and validated, and adapted for other drugs of abuse before it can become a general tool for monitoring drug abuse. The main aspects to be thoroughly validated involve the chemical and biological stability of the drug's main excretion products and its partition in sewage. Of less concern, in our opinion, is some inaccuracy in estimating consumption that may derive from assumptions related to pharmacokinetics and metabolism. In fact, if consumption is back-calculated from the levels of excreted products using an average drug-to-metabolite fractional conversion factor from multiple thorough studies, the accuracy of the estimate would be only marginally affected by this parameter.
Our data suggest that actual cocaine consumption may be much greater than estimated by current methods. This is a striking result, considering that – as discussed above – the method employed and the assumptions made could only lead to underestimated consumption figures. There is in fact no reasonable mechanism by which cocaine excretion products could accumulate in flowing surface waters, and we found steady concentrations in the River Po over time. Moreover, having chosen in our study to monitor an abundant metabolite in addition to the parent drug, any increase in cocaine levels due to illicit disposal rather than human use would be promptly disclosed by a transient increased cocaine/metabolite ratio.
Official statistics [2] for the year 2001 indicate that in Italy about 1.1% of young adults (15–34 yrs old) admit having used cocaine "at least once in the preceding month", but the actual dosages and frequency of use are not known. Therefore, it is hard to estimate the amount of cocaine that is consumed by the population. If we consider that in the River Po basin there are about 1.4 million young adults, the official figures in this area would translate into at least 15,000 cocaine use events per month. We however found evidence of about 40,000 doses per day, a vastly larger estimate. The economic impact of trafficking such a large amount of cocaine would be staggering. The large amount of cocaine (at least 1500 kilograms) that our findings suggest are consumed per year in the River Po basin would amount, in fact, to about $150 million in street value (based on an average US street value of $100 per gram [17,18]).
The above estimates – obtained from the heavily populated basin of the largest Italian river – were confirmed by similar values found in a completely different setting, i.e. in urban WW of medium-sized cities, chosen in widespread geographical locations to estimate local cocaine consumption on a small scale. The fair correspondence of SW and WW findings, despite the different settings and assumptions, suggests that our approach is reliable, and our estimates realistic. The rather narrow variation of estimated consumption among WWTPs may reflect different local habits, as the urban areas chosen have some socio-cultural differences.
Conclusion
Surveys of the general population are useful to describe patterns of drug abuse, but they are very expensive, and certainly too lengthy to detect changing trends promptly [4]. Continuous monitoring of illicit drug consumption would be very important for assessing the actual extent of this phenomenon, and detecting changes in trends. A more realistic picture of local use patterns for the most common illicit drugs would also be needed to identify priority problems and plan selective countermeasures. The evidence-based approach first tested here, which is in principle adaptable to other illicit drugs, could be refined and further validated to become a general, rapid method to help estimate drug abuse at the local level. This approach [13], with its unique ability to monitor changing habits in real time, could be helpful to social scientists and authorities for continuously updated appraisal of drug abuse.
Abbreviations
WW – waste waters
SW – surface water
WWTP – waste water treatment plant
BE – benzoylecgonine
MS – mass spectrometry
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
EZ designed the study and wrote the paper. CC analyzed the data and wrote the paper. SC collected and analyzed the samples. DC designed the study and analyzed the data. RB developed the analytical method and supervised the analyses. SS collected and analyzed the samples. RF had the original idea, and critically reviewed the results and the manuscript. All authors read and approved the final manuscript.
Acknowledgements
Prof. Davide Calamari regrettably died before the paper was completed. We dedicate this paper to his memory.
The University and Scientific Research Ministry (MIUR) funded this study (project no. 2002098317, 2002). Silvia Schiarea was the recipient of a "COFIN 2002" fellowship from University of Insubria.
==== Refs
United Nations Office of Drug and Crime World Drug Report 2004. Volume 2. Statistics
Ministero del Lavoro e delle Politiche Sociali Italy drug situation 2001. Report to the EMCDDA by the Reitox National Focal Point
European Monitoring Centre for Drugs and Drug Addiction The state of the drug problem in the European Union and Norway. Annual Report 2003 2004 Lisbon: EMCDDA
European Monitoring Centre for Drugs and Drug Addiction Handbook for surveys on drug use among the general population EMCDDA project CT99EP08B 2002 Lisbon: EMCDDA
Ternes TA Occurrence of drugs in German sewage treatment plants and rivers Water Research 1998 32 3245 3260 10.1016/S0043-1354(98)00099-2
Zuccato E Calamari D Natangelo M Fanelli R Presence of therapeutic drugs in the environment Lancet 2000 355 1789 1790 10832833 10.1016/S0140-6736(00)02270-4
Heberer T Occurrence, fate, and removal of pharmaceutical residues in the aquatic environment: a review of recent research data Toxicol Lett 2002 131 5 17 11988354 10.1016/S0378-4274(02)00041-3
Kolpin D Furlong ET Meyer MT Thurman EM Zaugg SD Barber LB Buxton HT Pharmaceuticals, hormones and other organic wastewater contaminants in U.S. streams, 1999–2000: a national reconnaissance Environ Sci Technol 2002 36 1202 1211 11944670 10.1021/es011055j
Kummerer K Drugs in the environment: emission of drugs, diagnostic aids and disinfectants into wastewater by hospitals in relation to other sources Chemosphere 2001 45 957 969 11695619 10.1016/S0045-6535(01)00144-8
Calamari D Zuccato E Castiglioni S Bagnati R Fanelli R Strategic survey of therapeutic drugs in the rivers Po and Lambro in northern Italy Environ Sci Technol 2003 37 1241 1248 10.1021/es020158e
Castiglioni S Fanelli R Calamari D Bagnati R Zuccato E Methodological approaches for studying pharmaceuticals in the environment by comparing predicted and measured concentrations in River Po, Italy Regul Toxicol Pharmacol 2004 39 25 32 14746777 10.1016/j.yrtph.2003.10.002
Heberer T Feldmann D Contribution of effluents from hospitals and private households to the total loads of diclofenac and carbamazepine in municipal sewage effluents-modeling versus measurements J Hazard Mater 2005 122 211 218 15967276 10.1016/j.jhazmat.2005.03.007
Daughton CG Daughton CG, Jones-Lepp T Illicit Drugs in Municipal Sewage: Proposed New Non-Intrusive Tool to Heighten Public Awareness of Societal Use of Illicit/Abused Drugs and Their Potential for Ecological Consequences Pharmaceuticals and Personal Care Products in the Environment: Scientific and Regulatory Issue Symposium Series 791 2001 Washington, DC: American Chemical Society 348 364
Baselt RC Disposition of toxic drugs and chemicals in man 1982 2 Davis (CA): Biomedical Publications
Ambre J The urinary excretion of cocaine and metabolites in humans: a kinetic analysis of published data J Anal Toxicol 1985 9 241 245 4079334
ISTAT 14° Censimento Generale della Popolazione e delle Abitazioni 2001
Drug Policy Information Clearinghouse (Office of National Drug Control Policy) Fact Sheet 2003
United States Department of State, Foreign Press Center Drug Control: International Policy and Approaches CRS Issue Brief for Congress May 14, 2004
|
16083497
|
PMC1190203
|
CC BY
|
2021-01-04 16:36:32
|
no
|
Environ Health. 2005 Aug 5; 4:14
|
utf-8
|
Environ Health
| 2,005 |
10.1186/1476-069X-4-14
|
oa_comm
|
==== Front
Environ HealthEnvironmental Health1476-069XBioMed Central London 1476-069X-4-161609295910.1186/1476-069X-4-16ResearchAssessing poisoning risks related to storage of household hazardous materials: using a focus group to improve a survey questionnaire Kaufman Martin M [email protected] Susan [email protected] David [email protected] Department of Earth and Resource Science, University of Michigan-Flint, 516 Murchie, Flint, Michigan 48502-1950, USA2 Children's Hospital of Michigan, 4160 John R, Detroit, Michigan 48201-2196, USA3 Office of Research, University of Michigan Flint, 530 French Hall, Flint, Michigan 48502-1950, USA2005 10 8 2005 4 16 16 24 3 2005 10 8 2005 Copyright © 2005 Kaufman et al; licensee BioMed Central Ltd.2005Kaufman et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
In fall of 2004, the authors began an investigation to characterize the correlations between the storage of Household Hazardous Materials and the associated health risks, particularly to children. The study area selected was Genesee County, Michigan, near Flint, with data to be collected by a phone survey of residents and through the acquisition of county hospital records containing procedure codes indicating treatment for poison emergencies, and review of poison control center data.
Methods
A focus group was used to identify key topics and relationships within these data for improving the phone survey questionnaire and its analysis.
Results
The focus group was successful in identifying the key issues with respect to all the data collection objectives, resulting in a significantly shorter and more topically focused survey questionnaire. Execution time of the phone survey decreased from 30 to 12 minutes, and useful relationships between the data were revealed, e.g., the linkage between reading food labels and reading labels on containers containing potentially harmful substances.
Conclusion
Focus groups and their preparatory planning can help reveal data interrelationships before larger surveys are undertaken. Even where time and budget constraints prevent the ability to conduct a series of focus groups, one successful focus group session can improve survey performance and reduce costs.
==== Body
Background
In the United States today, the average household stores 3–10 gallons of hazardous materials [1]. Inadvertent exposure to these household hazardous materials (HHMs), along with their improper use and disposal can create health risks. The most recent annual report of the American Association of Poison Control Centers identifies over 2.3 million exposures in 2003. Of these 2.3 million cases, over 50 percent were children under 6 years of age [2]. The Institute of Medicine reported that poisoning is a larger and more important public health hazard than previously realized. They describe 30,800 poisoning-related deaths in 2001, which makes poisoning the second leading cause of injury-related death in the United States. During that same year, there were 282,012 hospitalizations for treatment of poisoning. The estimated annual cost of poisoning was $8.5 billion in 1989, which is equivalent to $12.6 billion in 2003 dollars [3]. For all ages nationwide, the substances most commonly involved in human exposures are analgesics – common painkillers – followed by cleaning substances and cosmetic products. Nationally for children under 6, cosmetics are the most common toxic exposure, followed by cleaning substances and analgesics [2]. Although improvements have occurred in the prevention and treatment of poison emergencies [4-6], the poisoning incidence rate for young children (< 6 yrs.) is still alarming, with the actual rate being close to 4 million annually, since the proportion of incidents reported to Poison Control Centers is estimated to be as low as 26 percent [6].
In fall of 2004, the authors began an investigation to characterize the correlations between the storage of HHMs and the associated health risks, particularly to children. The study area selected was Genesee County, Michigan, near Flint, with data to be collected by a phone survey of residents and through the acquisition of county hospital records containing procedure codes indicating treatment for poison emergencies, and review of poison control center data. Analysis of the survey data would be used to help characterize the storage characteristics of toxic substances within households, and the regional hospital and poison control center call data would be used to help determine if these storage patterns were related to increased visits to hospital emergency rooms. Ultimately, this information would be used to develop intervention strategies for reducing the risks from HHMs to children in Genesee County.
The initial phone survey developed by the research team had 159 questions and in pre-trial testing took 30 minutes to complete. Given project budget constraints, and the strong likelihood respondents would not want to commit 30 minutes of their time to a survey, the research team saw the use of a focus group as an opportunity to shorten – and if possible – improve the survey instrument.
In the public health arena, focus groups have been used to assist research in a wide variety of applications, such as nursing education [7], contraception alternatives [8], and injury prevention among adolescents [9]. Basche [10] defines focus groups as "a qualitative research technique used to obtain data about feelings and opinions of small groups of participants about a given problem, experience, service, or other phenomenon", and they are characterized as an exploratory research method [11]. Exploratory studies typically serve three purposes: (1) to satisfy the researcher's curiosity and desire for better understanding, (2) to test the feasibility of undertaking a more careful study, and (3) to develop the methods to be employed in a more careful study [11].
This paper describes how one focus group held during fall 2004 at the University of Michigan-Flint was used to help refine a phone survey instrument. The topic of the session was "Reducing the Risks to Children from Household Hazardous Materials". Implications for the entire project are also discussed.
Methods
The methods used in this research were guided by 4 factors: (1) the objectives of the focus group; (2) development of an analytical framework for refining the survey with the focus group results; (3) recruiting the best mix of participants to help achieve the objectives of the focus group; and, (4) developing a favorable environment to ensure the success of the focus group session.
Focus Group Objectives
Focus groups work particularly well for determining the feelings, attitudes, and manner of thinking of a study population [12]. These characteristics of focus group methodology lend themselves well to the general purpose of improving the content and reducing the length of telephone surveys. The primary purpose of this focus group was to gauge the reaction of the participants to the key terms and concepts used by the researchers and to determine the level of sophistication of participants in differentiating across broad conceptual lines. Specifically, there was concern regarding the ability of members of the study population to differentiate across the concept of "toxicity" as well as the level of "awareness" of the population to the problem. In addition, the terms used in geo-spatial thinking on the part of the population were unknown. By testing the proposed survey questions before a small focus group it was felt that a significant insight could be gained that would allow the researchers to reduce redundancy by focusing on only those terms understood by the population and to rewrite questions in such a way that potential respondents would be able to understand and answer the questions. A brief overview of the development of the phone survey will help provide context for the implementation of the focus group.
The phone survey was designed by the authors – who are an environmental scientist, toxicologist, and research consultant, respectively. A primary objective of this survey was to characterize the storage characteristics of toxic substances within households. Specifically, this meant quantifying the distribution of certain HHMs throughout the different rooms of a home, as well as their storage elevations within each room. The HHMs selected for analysis were the substances involved with the most frequent poison exposures to young children (< 6 years old) as determined by Toxic Exposure Surveillance System (TESS) data for the study area (ibuprofen, acetaminophen, bleach, diaper rash products, acrylic nail products, mouthwash cosmetics, birth control pills, silica gel, vitamins with iron, peroxide, Hg thermometers, prescription medications). For instance, the survey results would yield the relative percentages of ibuprofen stored in the bathroom and bedroom, as well as a general breakdown of its storage elevations (low, below 4 feet; or high, over 4 feet) within each room. It would also be possible to ascertain whether households with young children were storing HHMs at higher elevations. This information would become useful when designing educational materials during the intervention stage of the project.
As noted, the initial phone survey had 159 questions; with the majority involving this 9-question sequence for each of the 13 HHMs listed above: (1) "Do you have this product?"; (2) "How often it was used?"; (3) "Where it was kept?"; (4) "Was the product moved to a new container?"; (5) "Was that container a food container?"; (6) "Has anyone in the house swallowed this substance?"; (7) "How did you respond?"; then, if the respondent did not call the PCC: The interviewer would ask: (8) "Did you consider calling the Poison Control Center?", and, (9) "Was there any particular reason you didn't call the Center?" Although it was not likely questions 2–9 would get asked for every substance, there were potentially 117 (13 * 9) questions possible from this sequence.
The remainder of the survey consisted of: 6 questions which asked respondents to rank certain actions as to whether they constituted a poison emergency (this was the same group of questions used to begin the focus group); 5 questions related to who would be called in the case of a poison emergency (which are redundant with questions 8 and 9 in the sequence listed above); 11 questions which asked how long ago some of the toxic substances listed above had been purchased; 4 questions about the presence and respondents' use of local programs for recycling/collecting hazardous waste; and the 16 remaining questions were divided among ascertaining the respondents' demographic characteristics, whether young children (< 6 yrs.) were present in the household, and the respondents' use and storage of weekly pill planners.
Development of the Analytical Framework
From this mass of 159 questions, the research team identified 6 general areas of content: (1) assessing the public's general knowledge of what events would constitute a poison emergency; (2) obtaining feedback about the specific products within the home which posed the most risks – especially to children; (3) how people received toxicity information about household hazardous materials; (4) respondents' perception and knowledge of the Poison Control Centers; (5) citizens' involvement with activities such as HHM recycling/pickup programs to reduce the presence of HHMs in their homes; and, (6) storage issues, including where people stored purses, their use and storage of weekly pill planners, and whether containers containing HHMs were changed.
These six content areas became the framework which was used to provide the structure for the questions posed during the focus group, and for the analysis of the participant responses used to change the phone survey. Questions for the focus group were generated within each of the 6 content areas using these guidelines: (1) they would be simple, and without compound construction; (2) the initial questions posed in content area #1 would be designed as "ice breakers" – they would be very straightforward, and act as a transition to the discussion about content group #2; (3) potential linkages between the content groups would be explored by follow-up questions. For example, since education is an important component of each content area, follow-up questions about education would be included throughout the session.
The analysis process using the focus group output was performed systematically. For each of the six content areas, the responses from the focus group were matched to questions of corresponding content in the existing phone survey. Then, the survey questions were evaluated based on these three criteria: (1) were there redundancies present; (2) did the question yield good information; and, (3) did opportunities exist to develop new avenues for analysis.
Participant Recruitment
While recognizing that focus group methodology is a non-random approach not intended for the purpose of generalization or inference [12], it must also be recognized that a critical requirement for achieving successful exploratory research is "representativeness", which occurs when a sample has the same distribution of characteristics as the population from which it was selected [11]. In this research, the target population consisted of households containing small children. Thus, when recruiting participants for the focus group, most of the effort was directed toward parents of small children and grandparents, since children frequently visit their residences. Since a phone survey was to be conducted throughout the county, there was a need for the ethnic composition of the focus group participants to be representative of the ethnic composition of the broader region. Flyers were placed at the daycare facility on the University of Michigan-Flint's (UM-F) campus, and e-mail notices were sent to university faculty and staff. Since UM-F is a commuter campus, there is a high percentage of non-traditional students (e.g., the average age of the students is over 25). In addition, the university is a major employer in the county, so the constituency of the daycare center provides a good representation of the region's demographics.
After the first attempt during the late summer of 2004 failed to produce enough participants, a second effort mounted in September 2004 using the same recruiting methods yielded 13 participants. A total of 15 participants were initially recruited with the hope of 10 participating. One reasonable explanation for the success of the second effort in September is related to the large numbers of people on vacation in late August. Specific language was used in the publicity campaigns to recruit a mix of parents with young children (under 6 years) and grandparents. The recruitment of grandparents was based on the reasoning that young children often visit their grandparent's homes; by obtaining the characteristics of HHMs within these households it would also provide relevant information for reducing risks to children.
The resulting gender/age/ethnic composition consisted of 11 women and 2 men; with 4 grandparents (3 women, 1 man) and 9 parents of young children (8 women, 1 man). Among the women there were 2 African-Americans, one Hispanic, and 8 whites, while both male participants were white. These characteristics provided a good representation of the ethnicity of the study area, along with a good distribution of young mothers and grandparents.
Developing a Favorable Environment
Another important factor contributing to a successful focus group is the need to conduct the discussion in a conducive environment [12]. To enable this environment, a central location and convenient time were needed; the location selected was a room at the university with a 4:00 P.M. start time. The university is centrally located in Genesee County, and the start time corresponded to the time many participants picked up their children from the daycare at the university. To help participants through the scheduled two-hour session, pizza and a $50 stipend were also provided.
There were extensive discussions between the project team and staff from the University's Research Office with broad focus group experience regarding the use of video and tape recorders during the session. Unlike focus groups designed to assist the researcher in theory building through the development of rich qualitative data where audio and video recording are essential, focus groups conducted for the purpose of restructuring surveys can be more informal. When the primary purpose of the focus group is to explicate the terms in which participants think note taking and flip charting may be adequate means of recording.
To minimize the potential for bias in the analysis, the decision was made to have two sets of notes transcribed and compared after the session was completed. One person would record notes on a flip chart, and the moderator would also takes notes. The moderator felt that taking notes would help him develop additional questions spontaneously, yet not detract from his effectiveness during the session. University research personnel concurred, noting that while video and audio taping would likely not be intrusive, there were no firm rules for focus group data recording. Another reason for taking this approach was to create pauses at different times during the session to provide breaks. These pauses would allow the moderator to stay on track, and enable participants to confirm exactly what was said – which might not have been discernable from the video or tape recordings after the session was finished.
Results
A two-hour focus group about reducing the risks to children from household hazardous materials was conducted on September 30, 2004. As participants entered the session, they were requested to voluntarily supply demographic information (parent, grandparent, ethnicity, gender), and all participants provided this information. The session formally opened with the moderator welcoming participants and thanking them for their participation. Notes were taken by personnel from the university's Office of Research and the moderator. No props were used during the session.
Table 1 summarizes the analysis of the phone survey content areas using the evaluation criteria. The high number of "Y" cells in the table indicates that the focus group output provided an effective means for improving the phone survey. A "N/A" entry indicates there were no questions in the phone survey within this content area. Row by row descriptions of how the criteria were used with the focus group responses and applied to the content areas within the phone survey are provided.
Table 1 Content area evaluation of the phone survey (N = no, Y = yes, N/A = not applicable)
Content area Redundancy Present Good Information New Analysis Opportunities
Poison Emergencies Y N Y
Products posing most risks Y N Y
Acquisition of HHM info N N/A Y
Perception of the PCCs Y Y Y
HHM Recycling N Y Y
Storage Issues N Y Y
General Knowledge of Poison Emergencies
As an "ice breaker" at the beginning of the session, participants were given a list of 6 questions which asked them to indicate on a scale of 1 to 5 whether exposure to certain substances were "not a poison emergency" (the low end of the scale) or "a very serious poison emergency" (the high end of the scale). The focus group responses indicated good awareness among the participants about the toxicity of swallowing bleach, paint, toadstools, and prescription medications. Swallowing mouthwash and spilling gasoline on someone's shoes were not deemed poison emergencies.
The ability of the participants in 5 out of 6 cases to correctly distinguish between what is and is not a poison emergency (swallowing mouthwash is a poison emergency) indicated there was a good general understanding of the difference between toxic and non-toxic substances. This result led the research team to consider a broader measurement of "awareness" as a better means to understand the public's baseline knowledge in this area. Awareness was conceived as a composite construct consisting of these components: whether a person would call the PCC first in case of a poison emergency; the location within the home for the PCC phone number;, the number of sources used to obtain information about HHMs; and the variety of HHMs (e.g., oil, batteries) taken to a hazardous materials collection center. This last item was included because it indicates a proactive attitude towards a potential hazard.
As indicated across the first row of Table 1, these questions were now redundant (column 1) since other questions fulfilled their role. Nor did they provide good information (column 2), since other questions were deemed more effective for the purpose of assessing existing knowledge of poison emergencies. The responses from the focus group did lead to the development of a new index for measuring awareness (column 3). As a result of this analysis, these six questions were deleted from the phone survey and replaced by four questions related to the composite construct of awareness described above.
Products Posing the Most Risks
The ice breaking questions were effective in focusing the participants' attention on specific toxic substances, and thus provided an easy transition into a discussion of the products posing the most risk in the home. The unanimous opinion of the participants was that cleaning products were the most toxic and posed the most risk to children because of their perfume-like smell, corrosiveness, and variety. Participants also noted the ingredients list on cleaning products frequently contained warnings about their safety. Vitamins containing iron supplements were mentioned as a threat, but knowledge of these products' toxicity was not well known among the participants.
In response to the follow-up question: "Are there potentially dangerous products which might be confused with non-toxic products – either through their labeling or their taste/smell", the answers were: TUMS® (look like candy); some high blood pressure medication (blue pills look like candy); bubble gum toothpaste; vitamins looking like gummy bears (these do not contain iron); and a newer product called Vitaballs® which looks like bubblegum.
At this juncture, one participant cited the issue of the "perception of parents to toxicity" as being particularly important. A discussion ensued about the levels of toxicity parents were willing to accommodate; e.g., some parents cannot stand cigarette smoke, and thus they are not willing to tolerate this exposure to themselves or their children. Other participants noted some parents were sensitive to food additives, and this sensitivity might translate into poison awareness as these people are more likely to read food labels and other product labels more carefully. Participants agreed there might be a correlation between nutrition consciousness and sensitivity to household hazardous materials (HHMs).
The sequencing of the questions in this content area focused the research team's attention on child safety. There would be increased child safety if access to HHMs was restricted by using locks and storing HHMs at higher elevations in the home. Thus, when the criterion of redundancy was applied, these responses provided the catalyst for consolidating the 9-question sequence used with the 13 toxic substances in the original survey (Table 1, row 2, column 1). The question sequence was: (1) "Do you have this product?"; (2) "How often it was used?"; (3) "Where it was kept?"; (4) "Was the product moved to a new container?"; (5) "Was that container a food container?"; (6) "Has anyone in the house swallowed this substance?"; (7) "How did you respond?"; then, if the respondent did not call the PCC: The interviewer would ask: (8) "Did you consider calling the Poison Control Center?", and, (9) "Was there any particular reason you didn't call the Center?" Although it was not likely questions 2–9 would get asked for every substance, there were potentially 117 (13 * 9) questions possible from this sequence.
As a result of using the framework, the 9-question sequence was reconfigured to contain only 2 questions: "Do you have this substance?" and, "Is this storage location secured by a lock or other device?" To facilitate this consolidation, the revised survey contained this interviewer instruction to respondents: Please indicate where you store the following items (if you don't have an item, please say: "don't have it"). To further cut costs and survey execution time, the substances queried were reduced from 13 to 10 (diaper rash products, acrylic nail products and mouthwash were the three products omitted), thus reducing the questions in the phone survey related to substance storage and access from 117 to 20.
Continuing across the second row, the original questions did not yield good information, since they did not explicitly gather information about storage elevations of HHMs within the home. With better information about storage locations available in the revision, it would now become possible to create an elevation code based upon whether each HHM was stored above or below 4 feet in elevation. Low elevations received a "1", whereas higher elevations received a "2". A composite elevation ratio would also be computable by normalizing the elevations (dividing the total actual elevation score of the items present by the highest potential elevation score). This opened up new avenues for analysis (column 3), as it now became possible to compare this ratio score between households with and without children.
The project team also added this question to the revised survey: "Do you read food labels" to explore the possible correlation between nutrition consciousness and sensitivity to household hazardous materials (HHMs). This item was omitted from the awareness index because the research team wanted to test the responses for correlation independently with the index to help provide confirmation for its validity as an indicator of awareness.
Acquisition of HHM Information
When participants were asked how they received toxicity information about HHMs, many responded "the Internet", with certain magazines, such as "Mothering" also noted as a source. Some people learned about HHMs and toxic materials from workplace training (food service); others noted they learned from their parents. In the follow-up discussion, most participants said they clearly understood certain substances such as oil, gas, and bleach were toxic. However, participants identified perfume, nail polish, hair products, cosmetics, fabric fresheners, toothpaste, sunscreen, ant/insect sprays, baby oils, hand cleaners with alcohol, and insect repellants with DEET as less threatening substances, with these products often not receiving special precautions in terms of their storage. Participants noted high areas (top shelves) were used for their storage of oil, gas, and bleach.
Participants responded to the next question: "How can we help educate people about the less obvious products which still pose threats within the home?" with: "pediatricians could be more active in their dissemination of information about harmful substances", and by "consistent labels" being placed on these products – a baby's face with a line drawn through it was suggested.
Complacency was seen as an obstacle to parents' education – since many products were used so often – it was difficult to protect children all the time. Participants suggested it might be beneficial to teach children "not to touch anything that is not food". The participants also felt is was more important to educate the parents about HHMs rather than children, since the parents could then teach their children. However, children should also receive HHM education in the schools via videos and guest speakers, as they often do on fire safety.
As shown in Table 1, 3rd row, column 1, there were no redundancies in the original survey since there were no existing questions on this topic. There would also be no way to tell if the information was good. As a result of the discussion in the focus group, we added this open-ended question: "In the past, from what source(s) have you obtained information about potentially harmful materials in the home?". Continuing across the 3rd row, the results would create new analysis opportunities by counting the total sources mentioned, and then including this total as part of the index of awareness.
Perception of the Poison Control Centers
Discussion was then directed to the perception and knowledge of the Poison Control Centers (PCC). All respondents answered the PCC was not the first place they would call in a poison emergency. After polling the participants, the response ranking was (in order of most to least): pediatrician, emergency room, looking at the label, calling a friend, with the obstacles to calling the PCC being: the phone number was inaccessible; lack of confidence in the "round the clock" and "instantaneous" availability of the PCC services; and possible language barriers. For raising the awareness of PCCs, respondents unanimously supported a suggestion that PCC kits be given out at the time the mother and child are discharged from the birthing center. Midwives could also be given the PCC information packets.
These two questions: "Did you consider calling the Poison Control Center?", and, "Was there any particular reason you didn't call the Center?", were part of the 9-question sequence which was restructured. This accounts for the "Y" in the redundancy criterion (row 3, column 1). The remaining questions pertaining to the PCCs in the existing survey were: In the case of a poison emergency, who would you call first for help?" Who would you call second for help?" Who would you call third for help?", "Where do you keep the number for the Poison Control Center?", and, if the PCC was not mentioned as a place to call, "How would you locate the phone number for the Poison Control Center?".
Responses from the focus group indicated the PCCs were not first in their minds when a poison emergency occurred. This input helped to verify that the existing questions would help assess who people were calling in poison emergencies. Thus, they would yield good information. In terms of new analysis opportunities, two of the questions about PCCs did achieve this as they became part of the awareness index (whether the person would call the PCC first in case of a poison emergency, and the location within the home for the PCC phone number).
HHM Recycling/Pickup Program Involvement
Participants indicated they received information about HHM recycling from local newsletters (e.g., a local community's newsletter sent to all residents had information about paint recycling). Local TV stations also promoted the biannual hazardous recycling collections in Genesee County. In response to how expired medications were disposed of the participants replied: in the garbage disposal, flushed down the toilet, or thrown in the general garbage pail. No respondents considered taking expired medications back to the pharmacist, and participants were unaware of exchange programs to eliminate mercury thermometers.
The existing survey had four questions related to this content area. As a result of the focus group, these remained unchanged. As shown in row 5 of Table 1, these questions when evaluated were not redundant and yielded good information. The discussion within the focus group about expired medications and mercury thermometers led us to add these questions to the survey: "How do you dispose of expired medications?", and, "If the mercury thermometer broke (assuming the respondent had answered "yes" to having one), how would you dispose of it? No new analysis opportunities resulted from the focus group input in this content area.
Purse and Weekly Pill Planner Storage, Changing HHM containers
During discussion of these final items, participants stated they stored purses "anywhere" – on tables, on the couch, and hung from the doorknob. When the moderator mentioned the term "weekly pill planners", which was the language used in the existing survey, this term was considered vague by the participants, who suggested "weekly pill dispenser". One respondent (not a senior) noted that childproofing is now available on weekly pill dispensers, but no other respondents (including the seniors) were aware of this capability.
Some participants noted they would change the containers for their pills when they went on vacation, or move liquids from their original spray bottles into old spray bottles (especially cleaning products). Sometimes concentrations of cleaning products were diluted, and this prompted their relocation to another container. Since some medications require refrigeration, one participant asked how these medications could be kept safe from probing children.
Referring to Table 1, 6th row, there were no redundancies in the original survey. The only existing questions on these topics were about weekly pill dispensers: "Does any member of the household use a weekly pill dispenser to dispense their medication?", and, "Do your children ever visit a household where a person uses a weekly pill dispenser (e.g., grandparents)?". We considered these questions adequate and able to yield good information. Based on the discussion in the focus group, these questions related to the transfer of cleaning items were added to the survey: "Have you ever transferred a household cleaning item to another container?", "What item was transferred?", "Was the container labeled?". We also added three similar questions about prescription medications. Finally, we added this question about purse storage: "Where are the purses of the household residents or guests typically placed?"
These new questions offered new analysis opportunities, such as the ability to correlate the relationship between the presence of children and the placement of purses within the home, e.g., were purses stored at higher elevations when children were present? Other avenues of analysis opened up by the focus group discussion were the ability to assess the prevalence of weekly pill dispensers within the study area and perform follow-up during the intervention phase with respect to their locking mechanisms. The discussion also raised the issue of deterring children from accessing refrigerated medications, which is a measure addressable in the intervention phase of the project. And finally, by correcting the terminology used for the pill dispensers, the focus group also helped to improve the survey.
This content area concluded the session. Participants were asked for any additional input, and when none was offered they were thanked for their efforts and again reminded of their capability to contact the project team for additional information about the research.
Discussion
The focus group was a primary catalyst for improving the phone survey. In effect, the focus group caused the research team to shift the emphasis of the survey away from a detailed substance by substance breakdown to one which considered broader spatial characteristics of HHMs. Specifically, the revised question sequence consisting of room location and elevation allowed investigators to create a spatial coding scheme for the characterization and analysis of toxic substances within the home. It was now possible to obtain a room-by-room accounting of toxic substance storage, and infer the risks to children associated with the substances' relative storage elevations. For example, items stored under the counter, on the floor, or in the drawer were coded as accessible to children; while items stored above the sink, in the medicine cabinet, or any storage location containing "high above" were coded as inaccessible to children.
This coding scheme enables a practical implementation of the "Virtual House" concept used by the U.S. Environmental Protection Agency (EPA) [13]. In the EPA's Virtual House, potentially toxic substances are depicted within each room of the home; this research provides actual substance locations and the associated risks to children. Moreover, the spatial characteristics of toxic substance storage can be compared to regional hospital data to help determine if these storage patterns are related to increased visits to hospital emergency rooms and non-hospitalized poisoning exposures. Ultimately, this information will be useful for developing intervention strategies for reducing the risks from HHMs to children in Genesee County.
Interdisciplinary Communication
Effective communication between the project team (with training in the natural sciences) and the social scientists from the Office of Research was essential to the success of the project. Underlying this imperative was the reality that the project team had no experience with focus groups (although they had some experience with survey design), while the social scientists knew little about the technical aspects of poison prevention. The communication between the two parties succeeded because of these two actions: (1) before implementation of the focus group, the project team provided the Office of Research personnel copies of the grant proposal, which contained a detailed explanation of the project. This helped acquaint them with the terminology used in poison prevention; (2) over the same period of time, project team members were assigned reading by the Office of Research personnel about focus groups. Meetings were then held to conduct "walkthroughs" of the anticipated focus group session. As noted in this paper, other discussions about taping the session were also held. Overall, communication between the two parties was ongoing and thorough.
Limitations of Using One Focus Group
There are many concerns about using only one focus group to refine a broader survey instrument. Three concerns come to the forefront: (1) is there adequate representation of the survey's target population?; (2) would there be enough data to help refine the survey?; and, (3) without replication, would the quality of the data suffice? Before conducting the focus group, we did not know the answers to any of these questions. What we concentrated on was recruiting the best mix of participants, and exploiting to good purpose the exploratory nature of the focus group instrument by asking simple open-ended questions in a comfortable environment. These tactics seemed to work.
Survey Consolidation
The substantial reduction of the questionnaire allowed investigators to insert some new open-ended questions. These questions allowed for the creation of measurement scales, and opened up more capabilities for analysis. The existing questions concerning demographic information, the presence of children, and the respondents' use and storage of weekly pill planners were retained, resulting in a final survey questionnaire containing 54 questions. This final questionnaire took 12 minutes to complete, compared to the 30 minutes needed for the pre-focus group questionnaire.
Project Implications
As demonstrated by the analysis of the existing survey, we learned that all responses from the focus group were relevant to this survey's refinement. In addition, the focus group responses provided substantial input to other aspects of the project. For example, the suggestion of giving PCC kits out at the time the mother and child are discharged from the birthing center may become one of the strategies attempted during the intervention phase of the project. Giving midwives the PCC information packets may also be attempted.
Analysis of current product labeling is another area where the feedback from the focus group will affect the intervention phase of the project. The responses to the question: "Are there potentially dangerous products which might be confused with non-toxic products – either through their labeling or their taste/smell" may lead to another potential intervention strategy. Here, an attempt may be made to work with local retailers to help identify these substances through the use of a special logo.
Conclusion
Focus groups are informal, but structured discussions between interested citizens and researchers designed to help the researchers identify key issues within their investigations. Focus groups provide a format of participant interaction that allows for self-disclosure beyond what can be achieved by other methods. This type of meeting format often helps to improve the quality of data collected. In this case, researchers from the University of Michigan-Flint and the Children's Hospital of Michigan solicited input on the characteristics of the harmful materials kept in/around people's homes which might pose risks to young children. The information obtained helped the researchers gain insight into several key issues related to household hazardous materials, consolidate a phone survey and improve the ability to analyze its results. Through follow-up work with citizens and local organizations, this information may also result in the design of more effective intervention strategies for reducing the risks posed to children by harmful substances in the home.
Systematic planning and effective communication must accompany any focus group implementation. Effective planning can help provide not only improved results for specific data collection instruments such as a survey questionnaire, but create general improvement within other project tasks. Effective communication between the project team and the research staff will educate both parties about their roles in the project and help ensure its success.
List of abbreviations
EPA : Environmental Protection Agency
HHM : Household Hazardous Materials
PCC : Poison Control Center
TESS : Toxic Exposure Surveillance System
UM-F : University of Michigan-Flint
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
MMK conceived the study, supervised all aspects of its implementation, and led the writing of the manuscript. SS assisted with the study and contributed to the design and analysis of all study components. DK designed the focus group session and assisted with the preparation of the methods section. All authors helped to interpret findings and review drafts of the manuscript.
Acknowledgements
This research was supported by a grant from the Ruth Mott Foundation. The authors would also like to thank Sally Harris and Sally Conley from the Office of Research at the University of Michigan-Flint for their assistance with the preparation of the focus group.
® TUMS is a trademark of GlaxoSmithKline
® Vitaballs is a trademark of Amerifit Nutrition Company
==== Refs
Shoemaker JM Vitale CY Healthy Homes Healthy Kids: Protecting Your Children from Everyday Environmental Hazards 1991 Washington, D.C.: Island Press
Watson WA Litowitz TL Klein-Schwartz W Rodgers GC Jr 2003 annual report of the American Association of Poison Control Centers Toxic Exposure Surveillance System Am J Emerg Med 2004 22 335 404 15490384 10.1016/j.ajem.2004.06.001
Institute of Medicine of the National Academies Forging a Poison Prevention and Control System 2004 Washington, D.C.: The National Academies Press
Woolf A Lewander W Filippone G Lovejoy F Prevention of childhood poisoning: efficacy of an educational program carried out in an emergency clinic Pediatrics 1987 80 359 363 2888072
Krenzelok EP The use of poison prevention and education strategies to enhance the awareness of the poison information center and to prevent accidental pediatric poisonings Clin Toxicol 1995 33 663 667
Shannon M Ingestion of toxic substances by children N Eng J Med 2000 342 186 191 10.1056/NEJM200001203420307
Morrison RS Peoples L Using focus group methodology in nursing J Contin Educ Nurs 1999 30 62 65 10382456
Otoide VO Oronsaye F Okonofua FE Why Nigerian adolescents seek abortion rather than contraception: evidence from focus-group discussions Int Fam Plan Perspect 2001 27 77 81
Murphy LMB Adolescent mothers' beliefs about parenting and injury prevention: results of a focus group J Pediatr Health Care 2001 15 194 199 11462127 10.1067/mph.2001.113658
Basche CE Focus groups interview: an underutilized research technique for improving theory and practice in health education Health Q 1987 14 411 418
Babbie E The Practice of Social Research 1992 6 Belmont, CA: Wadsworth Publishing Co
Kruger RA Focus Groups 1988 Newbury Park, CA: Sage Publications
Virtual House, United States Environmental Protection Agency
|
16092959
|
PMC1190204
|
CC BY
|
2021-01-04 16:36:32
|
no
|
Environ Health. 2005 Aug 10; 4:16
|
utf-8
|
Environ Health
| 2,005 |
10.1186/1476-069X-4-16
|
oa_comm
|
==== Front
Health Qual Life OutcomesHealth and Quality of Life Outcomes1477-7525BioMed Central London 1477-7525-3-471608350210.1186/1477-7525-3-47ReviewPsychosocial and socioeconomic burden of vasomotor symptoms in menopause: A comprehensive review Utian Wulf H [email protected] North American Menopause Society, 5900 Lander Brook Drive, Mayfield Heights, OH 44124, USA2005 5 8 2005 3 47 47 22 6 2005 5 8 2005 Copyright © 2005 Utian; licensee BioMed Central Ltd.2005Utian; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Many women experience vasomotor symptoms at or around the time of menopause. Hot flushes and night sweats are considered primary menopausal symptoms that may also be associated with sleep and mood disturbances, as well as decreased cognitive function. All of these symptoms may lead to social impairment and work-related difficulties that significantly decrease overall quality of life. Hot flushes have shown a great deal of variability in their frequency and severity in women. In some women, hot flushes persist for several months; in others, they may last for more than 10 years. Traditionally vasomotor symptoms were reported to begin 5 to 10 years prior to the cessation of the final menstrual cycle, corresponding with the initial decline in circulating gonadal hormones; however, night sweats in particular most often begin in perimenopause. The pathogenesis of hot flushes has not yet been fully elucidated, but the circuitry involving estrogen and neurotransmitters, norepinephrine and serotonin specifically, are hypothesized to play a major role in the altered homeostatic thermoregulatory mechanisms underlying these events.
Menopause-associated vasomotor symptoms are associated with significant direct and indirect costs. Overall costs of traditional pharmacotherapy or complementary and alternative medicine modalities, including over-the-counter treatments and dietary supplements, for managing menopause-related vasomotor symptoms are substantial and include initial and follow-up physician visits and telephone calls. Additional costs include laboratory testing, management of adverse events, loss of productivity at work, and personal and miscellaneous costs. Pharmacoeconomic analyses, including those that consider risks identified by the Women's Health Initiative, generally support the cost-effectiveness of hormonal therapy for menopause-associated vasomotor symptoms, which have been the mainstay for the management of these symptoms for more than 50 years. However, because many women now want to avoid hormone therapy, there is a need for additional targeted therapies, validated by results from controlled clinical trials that are safe, efficacious, cost-effective, and well tolerated by symptomatic menopausal women.
==== Body
Review
Menopause is characterized by physiologic and psychosocial changes in a woman's life. Menopause may be associated with vasomotor symptoms (VMS; hot flushes [also referred to as hot flashes] and night sweats), bone loss, urogenital atrophy, urinary tract infections and incontinence, increased cardiovascular risk, somatic symptoms, sexual dysfunction and decreased libido, and loss of skin elasticity. VMS, and the sleep and mood disturbances that often result from them, can have a significant negative impact on overall quality of life (QOL) for a substantial number of women. The impact of VMS has gained in importance as the lifespan of women has increased throughout the world since women can expect to spend a significant portion of their lives after menopause. This period should be a highly productive time for women, and maintaining functional ability and a good QOL is of utmost importance. Accordingly, it is important to understand the economic and QOL impacts of menopausal VMS as well as the most recent pharmacoeconomic analyses of different approaches to managing symptoms. Thus, the aims of this paper are to briefly review the epidemiology of VMS, describe what is known about the physiologic basis of these symptoms, and examine the global and health-related quality of life (HRQOL) effects of VMS in women, with a focus on psychosocial and economic impairments, and costs associated with treatments.
Epidemiology of VMS
Prevalence and risk factors
Recent US Census Bureau statistics indicate that approximately one third of women are older than 50 years of age [1]. It is estimated that 75% of women in this age group will experience hot flushes, a value supported by a recent longitudinal study of 454 women who were followed from premenopause to postmenopause [2]. Thus, in the United States alone, there are approximately 40 to 50 million women who experience hot flushes [1]. Worldwide, between 50% and 85% of women (approximately 360 million) older than 45 years of age experience hot flushes [3]. The prevalence of hot flushes varies widely across populations and is strongly influenced by culture and ethnicity. In the United States, the Study of Women's Health Across the Nation (SWAN) surveyed more than 16,000 women and found that the prevalence of hot flushes was highest among African Americans (46%), followed by Hispanics (34%), whites (31%), Chinese (21%), and Japanese (18%) [4]. In other parts of the world, rates of hot flushes vary widely as well, with the lowest prevalence observed in China (10%) and other Asian nations [5].
Many attempts have been made to identify demographic characteristics associated with a significantly increased risk of hot flushes. For many years, low body mass index (BMI) and race were considered significant predictors of VMS, with thin, white women believed to at the highest risk for hot flushes. More recent findings have suggested that high BMI and African American race are associated with a higher risk of VMS. This shift may be related to better sampling of the general population by major clinical trials because, traditionally, white middle-class women participated in clinical trials that often did not include women from other ethnic groups. The multiethnic SWAN not only demonstrated a link between an elevated BMI (≥27 kg/m2) and hot flushes [4], but an increased prevalence in African American women, as mentioned. Ongoing studies continue to investigate potential predictors of hot flushes. Smoking, maternal history, history of premenstrual complaints, elevated basal core body temperature, low physical activity, low socioeconomic status, and low levels of estrogen and high levels of luteinizing and follicle-stimulating hormones prior to the menopausal transition have all been associated with an increased risk of hot flushes [4,6-8].
Timing of hot flushes
The timing and frequency of hot flushes have been reviewed by several researchers [9]. SWAN demonstrated that hot flushes occur earlier than previously believed and may become less frequent and less intense as menopause progresses. SWAN data indicated that VMS were more frequently reported by women in late perimenopause with a relative risk for hot flushes at 1.0 during premenopause (the 1 or 2 years prior to menopause), 2.06 during early perimenopause (the early menopausal transition), 4.32 during late perimenopause (the late menopausal transition), and 2.81 during postmenopause [4]. The frequency of hot flushes varies but tends to remain consistent for an individual [10]. Many women have hot flushes on a daily basis, some as frequently as every hour, whereas others have VMS infrequently (ie, weekly or monthly) [10]. The majority of women experience hot flushes for 6 months to 2 years, with the highest number of women reporting symptoms during the first 2 postmenopausal years [9,10]. However, in another study, 26% of women reported having hot flushes for 6 to 10 years, whereas 10% have had VMS for more than 10 years [11].
Pathophysiology
The cause of hot flushes has yet to be determined because of the limited research focus in this therapeutic area. Hot flushes are believed to result from the brain's response to diminished hormones and hormonal fluctuations that occur during the menopausal transition. Ovarian hormones have been shown to influence thermoregulatory mechanisms that regulate temperature homeostasis in the hypothalamus. The neurotransmitters serotonin and norepinephrine play a role in modulating core body temperature, neurochemical messaging, and peripheral vasculature [12]. Kronenberg and colleagues were the first investigators to document cardiovascular, temperature, hormonal, and autonomic parameters with hot flushes and link them with thermoregulatory mechanisms [13], with more recent mechanistic information published by Freedman [14] and Deecher et al. [12].
Effects of associated VMS on QOL
Perceived QOL is difficult to measure and there is no universal agreement on how it should be quantified. Objective measurements of health status (often referred to as HRQOL) may not capture the patient's perception of overall life satisfaction. QOL can be defined as a reflection of an individual's belief about functioning and achievement. HRQOL may be viewed as the individual's perception regarding her physical, cognitive, and mental health as well as social situation [15]. Assessments of overall QOL for menopausal women must include consideration of somatic symptoms (hot flushes, night sweats, urogenital atrophy), psychological symptoms (depression, mood swings, irritability, anxiety), and life circumstances (function in the workplace). Thus, overall QOL may include four major factors: occupational, health-related, sexual, and emotional [15]. Consideration of HRQOL is also influenced by women's increased risk of multiple chronic diseases associated with menopause, including osteopenia, osteoporosis and related fractures, and cardiovascular disease [16].
VMS-related effects on QOL
VMS can have a significant negative impact on QOL in younger and older women, contributing to physical as well as psychosocial impairment (Table 1). Becoming flushed and sweating profusely in a social or work-related situation may cause extreme anxiety for many women and lead to social isolation [17].
Table 1 Vasomotor Symptoms and Related Psychosocial Impairment During the Menopausal Transition
Hot flushes
Night sweats
Sleep disturbances
Insomnia
Sleep apnea
Mood swings
Irritability
Sadness
Tension
Cognitive deficits
Poor concentration
Verbal memory problems
Social impairment
Disruption of family relationships
Social isolation
Work-related difficulties
Reduced productivity
Other Quality-of-life impairment
Embarrassment
Anxiety
Fatigue
Although it is generally accepted that VMS are troubling to many women and adversely impact their QOL, these effects are difficult to quantify because of the many factors that contribute to overall QOL satisfaction. For example, an objective of the Women's Health Initiative (WHI) was to determine whether hormonal therapy (HT) could reverse impaired HRQOL in 16,608 women aged 50 to 79 years. Results from this analysis indicated that estrogen plus progesterone did not yield any significant benefits in any of the HRQOL outcomes when compared with placebo over 3 years of follow-up [18]. However, this investigation did not examine global QOL in relation to VMS as an a priori outcome, and the women studied were not only excluded if the investigators felt they were having significant VMS that would interfere with long-term study involvement, but were older than those who typically have VMS and use HT. Another limitation was the use of a QOL questionnaire that may have been inadequate for determining the global impact of VMS. More recently, a new QOL scale has been designed specifically for the perimenopausal population [19]: The Utian Quality of Life (UQOL) Scale evaluates occupational, health, emotional, and sexual QOL. This 23-item assessment should increase the reliability of QOL measurement in perimenopausal and postmenopausal women.
The physiologic changes associated with menopause often result in increased anxiety and stress. These feelings may arise from sleep deprivation, mood swings, and unpredictable hot flushes. Before menopause, most women have a monthly hormonal rhythm. When the cycle becomes disrupted by erratic hormonal fluctuations, a woman's sense of well-being can be disturbed. These changes also tend to occur at a time when women are more likely to experience other life changes, including divorce, widowhood, children leaving home, concerns about aging parents, and other caregiving issues. Self-image is another important variable, and women with poor self-images have more flush-related distress [20]. In combination, these factors may each contribute to a reduced global QOL as well as decreased work productivity and difficulties with personal and social relationships.
A large number of studies have documented the negative impact of menopause on QOL. Ledesert and colleagues studied a cohort of 289 women aged 45 to 52 years who were no longer menstruating and reported lower HRQOL on several measures of the Nottingham Health Profile compared with values for menstruating women [21]. A population-based study evaluated the effects of various medical conditions on work impairment in more than 16,500 individuals [22]. Menopause was one of the factors associated with significant work limitations. The impact of VMS on HRQOL has also been studied by Bobula and colleagues, who evaluated 1,655 healthy, nonhysterectomized, postmenopausal women ranging in age from 40 to 65 years who were not receiving HT [23]. Their results indicated a significant correlation between moderate to severe hot flushes and decreased QOL. Women with moderate to severe hot flushes had significantly poorer scores than women with no hot flushes on four of the eight 36-item Short Form Health Survey (SF-36) subscales (vitality, bodily pain, social function, and role limitations-emotional), the mental composite score, and the Women's Health Questionnaire (WHQ), with pronounced differences for VMS, sleep problems, sexual behavior, and somatic symptoms. Women with moderate to severe hot flushes also had significantly poorer scores than women with mild hot flushes on two of the eight SF-36 subscales (vitality and social function) and five of the nine WHQ domains, including VMS, sleep problems, sexual behavior, somatic symptoms, and depressed mood. Finally, women with mild hot flushes had significantly worse scores than women with no hot flushes on two of the eight SF-36 subscales (vitality and role limitations-emotional) and on the WHQ for VMS, sleep problems, somatic symptoms, memory/concentration, and menstrual symptoms.
Impact of VMS on sleep, mood, and cognitive function
Despite the lack of agreement in the medical literature about the relationship between VMS on sleep quality, mood variability, and cognitive function, these symptoms are, in fact, primary complaints of menopausal women to their healthcare practitioners; as such, they are addressed in this review.
Sleep disturbances
The causes of menopause-related sleep disturbances are controversial. It has been suggested that problems with sleep may occur in older women independently of menopause. For example, nocturia increases with age and may disturb sleep [24]. Depression, stress, and other factors (eg, restless leg and other periodic limb movement syndromes) may also contribute to sleep disturbances in these patients [25].
Sleep disturbances also have been specifically related to hormonal changes that trigger hot flushes or night sweats, independent of age. A National Sleep Foundation poll of 1,000 women between the ages of 30 and 60 years found that 36% of perimenopausal, postmenopausal, and oophorectomized women experienced hot flushes during the night [26]. In this study, 44% of women who experienced VMS while sleeping were perimenopausal versus 28% of women who were postmenopausal. The poll also showed that menopausal and postmenopausal women slept less than premenopausal women. According to the National Sleep Foundation, women with night sweats experienced an average of three occurrences per week. These events disrupted sleep and led to daytime irritability [26].
Menopause-related VMS also may be associated with insomnia and disordered breathing at night. More perimenopausal and postmenopausal women than menstruating women have difficulty falling asleep, staying asleep, and achieving refreshing sleep [26]. Insomnia symptoms in women in the various stages of menopause include difficulty falling asleep (29%) and early awakening with an inability to fall back to sleep (21%). Respiratory abnormalities also may contribute to sleep disturbances in menopausal women. Results from a study of 589 women indicated that those in the menopausal transition were at a greater risk for complaints of sleep apnea and hypopnea than were younger women [27]. Postmenopausal women had a 2.6-to 3.5-fold greater rate of sleep-disordered breathing than their premenopausal counterparts [27]. Because sleep complaints are part of menopause-associated VMS, disordered breathing is often overlooked as a potential cause of menopause-associated sleep disturbance [27]. Although these and other results have suggested a correlation between the occurrence of hot flushes and sleep complaints in menopausal women, only a few studies have employed objective methods for sleep evaluation (eg, polysomnography, actigraphy, quantitative electroencephalographic analysis). Results from these assessments have indicated that hot flushes correlate with the occurrence of objectively demonstrable sleep disruption in at least some women [28].
Inadequate and unrefreshing sleep can have many consequences. Over time, disruption of sleep secondary to hot flushes and/or night sweats leads to chronic sleep deficits, significantly impaired alertness and mental acuity, carelessness, forgetfulness, and decreased work productivity. In some cases, night sweats can drench bedclothes and sheets, further disrupting sleep and necessitating a change of clothes and covers, which can also disturb the sleep of the individual's bed partner. Thus, lack of sleep, tiredness, and irritability can affect daytime productivity as well as familial and social relationships.
Mood
Menopause-associated changes in mood may result from a wide range of variables, including elevated sensitivity to environmental events secondary to decreased hormonal levels, changes in socioeconomic and/or marital status, culture, lifestyle factors, level of education, and history of depressive symptoms [29,30]. Longitudinal and cross-sectional studies carried out to date have not indicated a consistent relationship between the menopausal transition and increased risk of mood disorders [31,32]. However, results from the prospective population-based Melbourne Women's Midlife Health Project, which followed 438 women for 11 years and used the Center for Epidemiologic Studies Depression Scale to measure depression, indicated that depression scores were highest for women who were in the menopause transition stage (ie, had not reached their final menstrual period) or who had experienced surgical menopause. Current use of HT was associated with lower Center for Epidemiologic Studies Depression Scale scores (ie, less severe depressive symptoms) in this cohort [33]. These epidemiologic results are consistent with those from a small-scale clinical trial that demonstrated the significant benefit of short-term HT in perimenopausal women with depression [34]. It is important to note that many studies investigating mood and VMS have used depression scales. There are important distinctions between mood variability and major depression. It is incorrect to interpret the results from depression and simply extrapolate these findings to mood. Hopefully, new studies will address this issue and develop specific scales to delineate the impact of VMS on mood.
Cognitive decline
Memory impairment is directly related to hot flushes in women who have undergone oophorectomy, but natural menopause itself does not necessarily result in significant cognitive dysfunction [35]. During a hot flush, blood flow decreases in the hippocampus, possibly impairing memory and cognition [35]. It has been suggested that such reductions in blood flow may contribute to the decreased mental clarity and short-term verbal memory problems experienced by many perimenopausal and postmenopausal women [35]. The importance of estrogen in cognition has been demonstrated by Jacobs and colleagues, who measured cognitive function and verbal memory in 727 older women (average age, 74.2 years). Study results demonstrated that cognitive test scores and verbal memory were superior in the women who received HT compared with those who did not [36].
Costs associated with VMS and their treatment
In the year 2000, there were approximately 50 million women aged 45 to 60 years in the United States; as life expectancy increases, that number will increase [1]. Many of these women will use the health care system for premenopausal, perimenopausal, and postmenopausal needs. These requirements pose a high economic burden on the women themselves and on the health care system (Figure 1).
Figure 1 Flowchart of factors affecting management costs of menopause related vasomotor symptoms.
Determining the cost-effectiveness of treatments aimed at relieving VMS is complex. Analysis must consider the direct costs of treatment, costs associated with treatment-related adverse events, and the health care costs saved with effective therapy [37,38]. Cost-utility analysis of treatment for VMS also considers the impact of treatment on improvements in QOL associated with the alleviation of symptoms. This impact of treatment is expressed in quality-adjusted life years (QALYs) [39].
Since the publication of results from the WHI in 2002, therapy for menopausal symptoms has undergone a dramatic transition. A marked decline has been seen in prescriptions for oral estrogen and HT. Prescriptions for oral estrogens declined from 56.8 million in 2001 to 37.2 million in 2003. The respective values for oral HT were 24.0 and 10.3 million [40]. However, the pendulum has swung from fear to a greater understanding of the risks and benefits of HT [41].
Costs incurred for the management of menopause-related VMS include visits to physicians; follow-up visits and telephone calls for the management of medication-related side effects and changes in medication; self-prescribed over-the-counter remedies, including complementary alternative medications (CAMs); HT; laboratory tests; lost productivity at work; personal costs for hygiene-related supplies; energy costs for the increased use of air conditioning; and extra laundry requirements for clothing and bed sheets soiled with sweat. A recent Gallup poll of menopausal women showed that hot flushes and night sweats were the first-and second-ranked symptoms that prompted physician visits. Among the women surveyed, 70% complained of hot flushes, 68% complained of night sweats, 50% had mood changes and moodiness, and 49% experienced insomnia and sleeplessness [42]. The costs associated with many of these symptoms have not been quantified, but they undoubtedly result in a significant burden for the women who experience them.
Before seeking medical advice at the onset of VMS, women are likely to obtain information from their peers, family members, or the Internet. In many cases women resort to self-diagnosis and treatment. They may combine over-the-counter drugs with medications prescribed for other conditions (eg, analgesics for headache; anxiolytics and antidepressants for anxiety, tension, and mood changes; sedatives/hypnotics for insomnia). Most of these treatments fail to provide significant relief of VMS, and many women ultimately consult their physicians after these remedies fail. Physician visits for VMS include an initial visit to a primary care physician and, potentially, follow-up visits to a gynecologist or CAM specialist. Subsequent visits and telephone calls are often needed for medication adjustment, laboratory testing, and managing side effects. In addition, women may seek counseling from a psychologist or psychiatrist for mood changes, insomnia, and difficulties with family and social interactions. Women may also visit a neurologist for relief from headaches and/or help with cognitive deficits. All of these factors add to the economic burden of VMS.
Economic burden of therapy for symptom resolution
Hormonal therapy
The economic burden of VMS management remains high even with the use of HT. Costs associated with HT include one or two visits for diagnosis and medication prescription as well as follow-up visits and telephone calls to manage side effects and evaluate the efficacy of therapy. Serious, but rare, adverse events associated with HT can lead to exceptionally high acute and chronic costs [43]. Evaluation and management of more common transient adverse events, including vaginal and uterine bleeding, breast discomfort, and breast nodularity, can also add significantly to the overall cost of HT for menopause-related symptoms. Approximately one third of patients who use HT switch to another form of therapy or make medication adjustments because of adverse events or compliance problems, increasing the overall cost of therapy [43].
Despite the complexities associated with determining the cost-effectiveness of HT for the treatment of women with menopause, several pharmacoeconomic analyses support the use of such treatment. Results from an early evaluation by Weinstein suggested that HT was cost-effective in menopausal women with prior hysterectomy or osteoporosis, but not in asymptomatic women with an intact uterus. The evaluation included information about the risks of endometrial cancer, uterine bleeding, and gallbladder disease as well as the benefits associated with the relief of menopausal symptoms and prevention of osteoporosis and fractures [44].
Although long-term continuous-combined HT (0.625 mg/day of conjugated estrogens plus 2.5 mg of medroxyprogesterone [CEE/MPA]) has been associated with the potential for increased health risks in some women, therapy limited to ≤5 years is no doubt beneficial for decreasing VMS, somatic symptoms, and resultant sleep and mood disturbances in women who experience bothersome menopausal symptoms. Moreover, short-term, low-dose therapy may minimize the risk of adverse effects associated with longer-term HT [45]. Botteman and colleagues compared the cost-effectiveness of short-term CEE/MPA 0.625/2.5 mg against that of norethindrone acetate 1 mg and ethinyl estradiol 5 μg (NA/EE), another continuous-combined HT with a different side-effect profile, and no interventional therapy for the management of moderate to severe VMS [45]. Anticipated 1-year baseline costs for the management of VMS included drug acquisition costs ($357 for NA/EE vs $474 for CEE/MPA), initiation of therapy requiring two physician visits ($132), HT-related breakthrough bleeding requiring endometrial biopsy and associated visits and laboratory testing ($345), or spotting requiring a telephone call to the physician ($16), and two physician visits for VMS plus clonidine therapy ($162; Table 2) [45]. Results from this analysis indicated that NA/EE was less expensive and more effective than CEE/MPA in QALYs. The cost-effectiveness of NA/EE was greater for patients with severe VMS than those with mild symptoms. NA/EE also was more cost-effective than no treatment, unless the symptoms were so mild that the discomfort of spotting or bleeding (the most common significant adverse effects of short-term HT) offset the QOL improvements associated with HT [45]. It also should be noted that the health risks associated with shorter-term (1 to 5 years) HT for the treatment of VMS are not known. Another trial evaluating NA/EE against CEE/MPA 0.625/2.5 mg or no therapy in premenopausal and perimenopausal women indicated that NA/EE is cost-effective as first-and second-line therapy [46]. Results from this study showed that NA/EE increased costs and QALYs compared with CEE/MPA and no therapy. Despite the increase in direct costs over no therapy and CEE/MPA, the cost effectiveness of NA/EE compared to no therapy is based on a decrease in the cost of treating menopausal symptoms, vaginal bleeding, and hip fractures and assumes a substantial increase in compliance compared to CEE/MPA as a result of improved control of bleeding with NA/EE. Results from a study reported by Ohsfeldt et al. indicated that the 1-year cost of HT for treating VMS was approximately $300 greater than the cost associated with no treatment [47]. It is not clear how results from this Canadian analysis would apply to the United States since health care is much less expensive in that country.
Table 2 Baseline Costs for a Pharmacoeconomic Model of Vasomotor Symptoms
Drug acquisition costs
Norethindrone acetate/ethinyl estradiol $357
Conjugated estrogen/medroxyprogesterone $474
Therapy initiation
Two physician visits $132
Breakthrough bleeding at 3 months (or continued spotting at 6 months)
Endometrial biopsy $198
Pathology and laboratory fees $147
Telephone call to physician
Spotting at 3 months $16
Moderate to severe vasomotor symptoms
Two physician visits $132
90-day supply of clonidine $31
Adapted with permission from Botteman et al. [45].
As noted, the published results from the WHI have dramatically reduced the use of estrogen and HT by menopausal women. Results from the WHI prompted reanalysis of the cost-effectiveness of HT by investigators at the Stockholm School of Economics. Generalization of results from this analysis may be limited because Sweden has a national socialized medical system. However, the authors suggest that HT remains a cost-effective therapeutic strategy for women with menopausal VMS compared with no therapy [47].
SSRIs and SNRIs
Selective serotonin reuptake inhibitors (SSRIs) and serotonin-norepinephrine reuptake inhibitors (SNRIs) have received increased attention for the management of VMS in nondepressed menopausal women [48]. As yet, there have been no pharmacoeconomic analyses of any agents in these classes for this indication. Results from a recent systematic review of published economic evaluations of interventions for depression indicated that SSRIs and SNRIs are more cost-effective than older antidepressant medications (eg, tricyclic antidepressants) owing to their greater efficacy and decreased side-effect profile [49].
Other prescription medications
Other prescription medications approved for use in conditions not associated with menopause-related VMS have demonstrated varying degrees of efficacy [48]. The economic advantage for some of these medications is that they have been on the market for a number of years. Given the understanding that the VMS are the result of a dysfunction in thermoregulatory circuitry, new nonhormonal therapies that selectively target the serotonin and norepinephrine pathways, without the involvement of other pathways, seem likely to become the next generation of care for the management of VMS.
CAM treatments
Many symptomatic menopausal women are likely to treat themselves before consulting a medical practitioner, thinking that "natural" products are safer and the ingredients more pure than synthetic agents. Cost analysis was carried out for CAM treatments that women would commonly find through a basic search on the Internet using the terms complementary medicine and hot flash as search parameters. The most common CAM treatments that emerged were products containing individual and compounded formulas of herbs, isoflavones, and dietary supplements that promised to alleviate menopause-related hot flushes and night sweats, irritability, sleeplessness, mood swings, weight gain, headaches, insomnia, depression, menstrual irregularities, fatigue, and loss of sexual desire. These formulations also claim to promote mental clarity, increase energy levels, and improve physical performance [50-53]. The initial cost of a single product ranged from $19.95 to $58.00 per month (Table 3). A key limitation in the analysis of these products is that their clinical efficacy has generally not been documented by results from controlled clinical trials. It has also been noted that any benefits associated with herbal supplements may occur more slowly than those achieved with traditional medications [54]. Comparison of 1-year costs of CAM treatments versus HT indicated that three of seven alternative treatments were more expensive than traditional therapy (Table 3).
Table 3 Complementary and Alternative Medicine Costs Over 6 Months
Product Main Ingredients Cost ($US) at 1 Month Cost ($US) at 3 Months Cost ($US) at 6 Months
Femforte® Black cohosh, soy isoflavones, androstenedione, chaste berry 58.00 174.00 348.00
MACA750™ Organic maca root 19.95 59.85 119.70
Promensil™ Red clover 24.95 64.90 129.80
Remifemin® Black cohosh extract equivalent to 20 mg dried Cimicifuga rhizome 49.99 78.98 157.96
Sleep & Slim™ L-glutamine, L-lysine HCl, magnesium citrate, L-ornithine, glysine, L-arginine, collagen, vitamin B6 (pyroxidine HCl), L-carnitine, vitamin B3 (niacin), aloe vera, ascorbic acid, citric acid, sodium benzoate, potassium sorbate, carrageenan 49.95 149.85 299.70
Effisoy™ AglyMax: fermented soy germ extract 29.95 79.35 149.75
Hot Flash, Non-GMO Soy Geneistein-rich soy concentrate, black cohosh root extract, dong quai root extract, licorice root extract, vitex berry extract 45.98 137.94 275.88
Note: Three-and 6-month costs that are lower than 3 and 6 times the 1-month costs for certain items reflect bulk prices for the items.
Conclusion
Menopause-related VMS are very common and can be associated with a high patient and societal burden. These symptoms result in high direct and indirect costs and significantly reduced QOL. Current treatments for VMS include HT, prescription medications developed for other indications, and CAM treatments. Short-term HT has been shown to be cost-effective for the management of VMS, but the publicity given the WHI has substantially decreased the use of these treatments. The physiology underlying VMS is complex and not fully understood, but it is clear that alterations in noradrenergic and serotonergic mechanisms during hypothalamic thermoregulation are involved in their development. A significant unmet need remains for menopause-related VMS treatment options. Among women who are eligible for the treatment of menopause-related VMS, 80% do not seek treatment, receive inadequate counseling, or do not have access to local medical aid [39]. The development of therapies that specifically target VMS may provide high efficacy and reduced risk of serious and potentially costly adverse events, thus increasing the overall cost-effectiveness of therapy.
List of abbreviations
BMI Body mass index
CAM Complementary alternative medication
CEE/MPA Conjugated estrogens plus medroxyprogesterone acetate
HRQOL Health-related quality of life
HT Hormonal therapy
NA/EE Norethindrone acetate plus ethinyl estradiol
QALY Quality-adjusted life-years
QOL Quality of life
SF-36 36-Item Short Form Health Survey
SNRI Serotonin-norepinephrine reuptake inhibitor
SSRI Selective serotonin reuptake inhibitor
SWAN Study of Women's Health Across the Nation
UQOL Utian Quality of Life
VMS Vasomotor symptoms
WHI Women's Health Initiative
WHQ Women's Health Questionnaire
==== Refs
United States Census 2000. Census 2000 profile 2000 U.S.Department of Commerce. Economics and Statistics Administration. U.S.Census Bureau 2 5
Avis NE Crawford SL McKinlay SM Psychosocial, behavioral, and health factors related to menopause symptomatology Womens Health 1997 3 103 120 9332153
Mohyi D Tabassi K Simon J Differential diagnosis of hot flashes Maturitas 1997 27 203 214 9288692 10.1016/S0378-5122(97)83974-6
Gold EB Sternfeld B Kelsey JL Brown C Mouton C Reame N Salamone L Stellato R Relation of demographic and lifestyle factors to symptoms in a multi-racial/ethnic population of women 40-55 years of age Am J Epidemiol 2000 152 463 473 10981461 10.1093/aje/152.5.463
Ho SC Chan SG Yip YB Cheng A Yi Q Chan C Menopausal symptoms and symptom clustering in Chinese women Maturitas 1999 33 219 227 10656500 10.1016/S0378-5122(99)00056-0
Whiteman MK Staropoli CA Benedict JC Borgeest C Flaws JA Risk factors for hot flashes in midlife women J Womens Health (Larchmt ) 2003 12 459 472 12869293 10.1089/154099903766651586
Whiteman MK Staropoli CA Langenberg PW McCarter RJ Kjerulff KH Flaws JA Smoking, body mass, and hot flashes in midlife women Obstet Gynecol 2003 101 264 272 12576249 10.1016/S0029-7844(02)02593-0
Staropoli CA Flaws JA Bush TL Moulton AW Predictors of menopausal hot flashes J Womens Health 1998 7 1149 1155 9861592
Nachtigall LE Nachtigall MJ Menopausal changes, quality of life, and hormone therapy Clin Obstet Gynecol 2004 47 485 488 15166873 10.1097/00003081-200406000-00023
Treatment of menopause-associated vasomotor symptoms: position statement of The North American Menopause Society Menopause 2004 11 11 33 14716179 10.1097/01.GME.0000108177.85442.71
Feldman BM Voda A Gronseth E The prevalence of hot flash and associated variables among perimenopausal women Res Nurs Health 1985 8 261 268 3852361
Deecher DC Physiology of thermoregulatory dysfunction and current approaches to the treatment of vasomotor symptoms Expert Opin Investig Drugs 2005 14 434 448 10.1517/13543784.14.4.435
Kronenberg F Hot flashes: epidemiology and physiology Ann N Y Acad Sci 1990 592 52 86 2197954
Freedman RR Pathophysiology and treatment of menopausal hot flashes Semin Reprod Med 2005 23 117 125 15852197 10.1055/s-2005-869479
Utian WH Janata JW Kingsberg SA Patrick LD Aso T and Yanaihara T Determinants and quantification of quality of life after the menopause: the Utian Menopause Quality of Life score The Menopause at the Millenium 2000 Taylor & Francis 141 144
Vliet EL Menopause and perimenopause: the role of ovarian hormones in common neuroendocrine syndromes in primary care Prim Care 2002 29 43 67, vi 11856658
McVeigh C Perimenopause: more than hot flushes and night sweats for some Australian women J Obstet Gynecol Neonatal Nurs 2005 34 21 27 15673642 10.1177/0884217504272801
Hays J Ockene JK Brunner RL Kotchen JM Manson JE Patterson RE Aragaki AK Shumaker SA Brzyski RG LaCroix AZ Granek IA Valanis BG Effects of estrogen plus progestin on health-related quality of life N Engl J Med 2003 348 1839 1854 12642637 10.1056/NEJMoa030311
Utian WH Janata JW Kingsberg SA Schluchter M Hamilton JC The Utian Quality of Life (UQOL) Scale: development and validation of an instrument to quantify quality of life through and beyond menopause Menopause 2002 9 402 410 12439099 10.1097/00042192-200211000-00005
Reynolds F Exploring self-image during hot flushes using a semantic differential scale: associations between poor self-image, depression, flush frequency and flush distress Maturitas 2002 42 201 207 12161044 10.1016/S0378-5122(02)00082-8
Ledesert B Ringa V Breart G Menopause and perceived health status among the women of the French GAZEL cohort Maturitas 1994 20 113 120 7715462 10.1016/0378-5122(94)90006-X
Burton WN Pransky G Conti DJ Chen CY Edington DW The association of medical conditions and presenteeism J Occup Environ Med 2004 46 S38 S45 15194894
Bobula JD Vasomotor symptoms and quality of life (QoL) in postmenopausal women Value Health 2003 6 Number 6 707
Stewart WF Van Rooyan JB Cundiff GW Abrams P Herzog AR Corey R Hunt TL Wein AJ Prevalence and burden of overactive bladder in the United States World J Urol 2003 20 327 336 12811491
Polo-Kantola P Erkkola R Sleep and the menopause J Br Menopause Soc 2004 10 145 150 15667750 10.1258/1362180042721076
1998 Women and Sleep Poll 1998 National Sleep Foundation
Young T Finn L Austin D Peterson A Menopausal status and sleep-disordered breathing in the Wisconsin Sleep Cohort Study Am J Respir Crit Care Med 2003 167 1181 1185 12615621 10.1164/rccm.200209-1055OC
Moe KE Hot flashes and sleep in women Sleep Med Rev 2004 8 487 497 15556380 10.1016/j.smrv.2004.07.005
Amore M Di Donato P Papalini A Berti A Palareti A Ferrari G Chirico C De Aloysio D Psychological status at the menopausal transition: an Italian epidemiological study Maturitas 2004 48 115 124 15172085 10.1016/j.maturitas.2003.08.010
Collins A Landgren BM Reproductive health, use of estrogen and experience of symptoms in perimenopausal women: a population-based study Maturitas 1994 20 101 111 7715461 10.1016/0378-5122(94)90005-1
Soares CN Joffe H Steiner M Menopause and mood Clin Obstet Gynecol 2004 47 576 591 15326420
Bromberger JT Meyer PM Kravitz HM Sommer B Cordal A Powell L Ganz PA Sutton-Tyrrell K Psychologic distress and natural menopause: a multiethnic community study Am J Public Health 2001 91 1435 1442 11527777
Dennerstein L Guthrie JR Clark M Lehert P Henderson VW A population-based study of depressed mood in middle-aged, Australian-born women Menopause 2004 11 563 568 15356410 10.1097/01.GME.0000113844.74462.F6
Cohen LS Soares CN Poitras JR Prouty J Alexander AB Shifren JL Short-term use of estradiol for depression in perimenopausal and postmenopausal women: a preliminary report Am J Psychiatry 2003 160 1519 1522 12900318 10.1176/appi.ajp.160.8.1519
Shepherd JE Effects of estrogen on congnition mood, and degenerative brain diseases J Am Pharm Assoc (Wash ) 2001 41 221 228 11297335
Jacobs DM Tang MX Stern Y Sano M Marder K Bell KL Schofield P Dooneief G Gurland B Mayeux R Cognitive function in nondemented older women who took estrogen after menopause Neurology 1998 50 368 373 9484355
Tosteson AN Weinstein MC Cost-effectiveness of hormone replacement therapy after the menopause Baillieres Clin Obstet Gynaecol 1991 5 943 959 1822828
Tosteson ANA Weinstein MC Schiff I Lobo RA Cost-effectiveness analysis of hormone replacement therapy Treatment of the Postmenopausal Woman: Basic and Clinical Aspects 1999 51 Second Philadelphia, Pennsylvania, Lippincott Williams &Wilkins 557 565
Whittington R Faulds D Hormone replacement therapy: I. A pharmacoeconomic appraisal of its therapeutic use in menopausal symptoms and urogenital estrogen deficiency Pharmacoeconomics 1994 5 419 445 10147233
Hersh AL Stefanick ML Stafford RS National use of postmenopausal hormone therapy: annual trends and response to recent evidence JAMA 2004 291 47 53 14709575 10.1001/jama.291.1.47
Utian WH Pieter van Keep Memorial Lecture. Menopause--a modern perspective from a controversial history Maturitas 1997 26 73 82 9089556 10.1016/S0378-5122(96)01092-4
The 2002 Gallup study of consumers' and physicians' changing HRT concerns. Qualitative Phase I & II 2002 MS 22085 Princeton, New Jersey, Multi-Sponsor Surveys, Inc 1 90
Ohsfeldt RL Gavin NI Thorp JM Medical care costs associated with postmenopausal estrogen plus progestogen therapy Value Health 2004 7 544 553 15367250 10.1111/j.1524-4733.2004.75006.x
Weinstein MC Estrogen use in postmenopausal women--costs, risks, and benefits N Engl J Med 1980 303 308 316 6770270
Botteman MF Shah NP Lian J Pashos CL Simon JA A cost-effectiveness evaluation of two continuous-combined hormone therapies for the management of moderate-to-severe vasomotor symptoms Menopause 2004 11 343 355 15167315 10.1097/01.GME.0000097742.96468.68
Coyle D Cranney A Tugwell P Economic evaluation of norethisterone acetate/ethinylestradiol (FemHRT) for women with menopausal symptoms Pharmacoeconomics 2003 21 661 669 12807367
N Z F B B J J K A re-assessment of the cost-effectiveness of hormone replacement therapy in Sweden--results based on the Women's Health Initiative randomised controlled trial. SSE/EFI Working Paper Series in Economics and Finance 2004 571 Stockholm, Sweden, Stockholm School of Economics, Centre for Health Economics 1 38 15280008
Fugate SE Church CO Nonestrogen treatment modalities for vasomotor symptoms associated with menopause Ann Pharmacother 2004 38 1482 1499 15292498 10.1345/aph.1D610
Barrett B Byford S Knapp M Evidence of cost-effective treatments for depression: a systematic review J Affect Disord 2005 84 1 13 15620380 10.1016/j.jad.2004.10.003
Source Naturals -Hot Flash, 180 tablets 2005
Maca 750 for hot flashes, night sweats, fatigue and low libido 2005
Sleep & Slim. The healthy all natural way to lose weight 2005
Femforte - managing the symptoms of menopause 2005
Kronenberg F Fugh-Berman A Complementary and alternative medicine for menopausal symptoms: a review of randomized, controlled trials Ann Intern Med 2002 137 805 813 12435217
|
16083502
|
PMC1190205
|
CC BY
|
2021-01-04 16:38:14
|
no
|
Health Qual Life Outcomes. 2005 Aug 5; 3:47
|
utf-8
|
Health Qual Life Outcomes
| 2,005 |
10.1186/1477-7525-3-47
|
oa_comm
|
==== Front
Int J Health GeogrInternational Journal of Health Geographics1476-072XBioMed Central London 1476-072X-4-191607640210.1186/1476-072X-4-19MethodologyComparison of spatial scan statistic and spatial filtering in estimating low birth weight clusters Ozdenerol Esra [email protected] Bryan L [email protected] Su Young [email protected] Melina S [email protected] Department of Earth Sciences, 236 Johnson Hall, University of Memphis, Tennessee, 38152, USA2 Department of Preventive Medicine, University of Tennessee Health Science Center, Memphis, Tennessee, USA2005 2 8 2005 4 19 19 6 6 2005 2 8 2005 Copyright © 2005 Ozdenerol et al; licensee BioMed Central Ltd.2005Ozdenerol et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The purpose of this study is to examine the spatial and population (e.g., socio-economic) characteristics of low birthweight using two different cluster estimation techniques. We compared the results of Kulldorff's Spatial Scan Statistic with the results of Rushton's Spatial filtering technique across increasing sizes of spatial filters (circle). We were able to demonstrate that varying approaches exist to explore spatial variation in patterns of low birth weight.
Results
Spatial filtering results did not show any particular area that was not statistically significant based on SaTScan. The high rates, which remain as the filter size increases to 0.4, 0.5 to 0.6 miles, respectively, indicate that these differences are less likely due to chance. The maternal characteristics of births within clusters differed considerably between the two methods. Progressively larger Spatial filters removed local spatial variability, which eventually produced an approximate uniform pattern of low birth weight.
Conclusion
SaTScan and Spatial filtering cluster estimation methods produced noticeably different results from the same individual level birth data. SaTScan clusters are likely to differ from Spatial filtering clusters in terms of population characteristics and geographic area within clusters. Using the two methods in conjunction could provide more detail about the population and spatial features contained with each type of cluster.
==== Body
Background
Spatial analysis of birth related morbidity (e.g., low birthweight) provides a way to identify inadequate health care access as well as potential environmental and behavioral causal factors. Cluster analysis is frequently used to identify an unusually high occurrence of morbidity that is clustered in space and time [1]. The purpose of this study is to examine the spatial characteristics of low birthweight using two distinct cluster analysis techniques and to compare the resultant clusters in terms of their socio-economic characteristics [2,3]. Both Spatial Scan Statistic (SaTScan) and Spatial filtering techniques are designed to overcome limitations in presenting data in an aggregated form where spatial patterns of low birth weight may vary in relation to the level of geographic area used.
This study will assess the spatial characteristics of infant birth weight throughout Shelby County, Tennessee from 2000 to 2002 with the use of Kulldorff's Spatial Scan Statistics or SaTScan and Rushton's Spatial filtering methods. Comparing the results of the two methods using the same input gives us more insight in to the spatial distribution of birth morbidity data. This is particularly relevant since there is little or no evidence of comparison of these two methods in the literature. The research questions are as follows: (1) Are low birthweight births clustered significantly in relation to maternal residence from 2000–2002 in Shelby County? (2) To what extent will the total area within SaTScan clusters differ from the total area within Spatial filtering clusters? (3) To what extent will the maternal and familial characteristics of those births within SaTScan clusters differ from those within Spatial filtering clusters? To assess both methods, we used individual point data by address-matching the latitude-longitude coordinates of the maternal residences at the time of delivery and assigning these to specific grid locations.
This case study demonstrates how the two methods better reflect spatial variation when individual point data is used. A methodological comparison can provide insight about the limitations and benefits of varying approaches when mapping morbidity. Especially since methodological studies of these techniques are scarce. This study also demonstrates a strategy for dealing with the issue of geographical scale, which is central to this type of small-area ecological study [4]. Information obtained from this study provides a useful foundation for prospective environmental health tracking. By comparing the results across different spatial scales, we hope to derive more reliable information on an important health concern in this region.
Cluster analysis of low birth weight
Studies have examined the spatial characteristics of low birth weight and other health outcomes in relation to contaminant exposures using geographic information systems (GIS) [5-7]. Results can vary significantly depending upon the level of geographic scale that is analyzed [8-10]. For example, spatial patterns of birth morbidity vary in relation to area-based census tract, block group, and zip code level measures of socio-economic status [12]. Additionally, aggregation bias represents an inherent source of error for these analyses [13]. Aggregation bias results from the rather arbitrary means by which GIS aggregate individual cases at some geographic organizational unit such as a census tract. The coarser the spatial scale the higher the potential for aggregation bias. SaTScan and Spatial filtering techniques differ in how each aggregates individual level data. Consequently, aggregation bias may manifest itself distinctively in one method versus the other.
Using GIS to study the spatial characteristics of low birthweight is complicated by the lack of methodological guidance and the scarcity of accurate integrated spatial-morbidity databases [14-18]. A representative, but not exhaustive, set of methods exist for analyzing spatial clusters of birth morbidity with individual point data or aggregated data that still maintains the stability of the estimated rates by constructing a continuous smoothed map [19-21]. SaTScan and Spatial filtering techniques are the most commonly used spatial analysis methods in epidemiological research. The SaTScan method has been applied much more broadly and more frequently than the Spatial filtering approach.
Spatial filtering process
The Spatial filtering methodology employs non-parametric statistical techniques as a tool in exploratory spatial data analysis. This method has been used to study clusters of congenital anomalies, infant mortality, and other forms of birth morbidity [3,18,20]. It works well with both aggregated data and individual point data. The estimated rate at a particular grid point can be defined as the observed rate within a fixed distance from the grid point. Once estimated rates are assigned to each grid point, isarithmic maps can be constructed in GIS. We assume that the probability of a case resulting to a low birthweight birth is equal to the proportion of all births in the region that resulted in low birthweight [3]. A 'smoothed' probability map can be drawn where the significance levels of high rates of low birth weight by percentage for each individual circle is calculated and mapped in isarithmic form.
SaTScan process
The SaTScan methods has been used to study clusters of cancer morbidity and mortality, sudden infant death syndrome, congenital anomalies, and infectious diseases [2,4,5]. SaTScan estimates the probability that the frequency of events per trial at each vertex surpasses that expected by chance. It then creates an isoplethic map that shows the estimated probabilities. SaTScan uses circles and a non-parametric test statistic. It takes into account the observed number of low birth weight births inside and outside the circle when calculating the highest likelihood for each circle. SaTScan uses a circular window of different sizes that scans the study area until a certain percent (e.g., half) of the total population is included. This circle is the most probable cluster, and has a rate that is the least likely to happen by chance alone. SaTScan also accounts for multiple testing through the calculation of the highest likelihood of occurrence for all possible cluster locations and sizes [2,5,21]. Although a range of probabilities can be displayed using SaTScan, only the most highly significant estimates are displayed in this paper.
As mentioned earlier, SaTScan accounts for multiple testing and only conducts one test for the whole collection of window location and sizes [2,5]. It tests the null hypothesis against the alternative hypothesis that there is an elevated rate of low birth weight within the windows as compared to the outside. The method uses the likelihood ratio λ as the test statistic [22]. The significance of the test statistic λ is determined by a large number of replications of the data set generated under the null hypothesis in a Monte Carlo simulation. The likelihood ratio λ for each replica is computed, and the result is significant at the 0.05 level if the λ value of the real data set is among the top 5% of all the values, including the replicas. Secondary clusters with lower significance can also be identified. SaTScan generates an ASCII output file, which contains the log likelihood ratios and their significance levels for the census areas. In this study, the output file was imported into Arc GIS 9 to create cluster maps to visually examine and compare the clusters.
Methodology
We conducted a retrospective ecological study of birth weight in Shelby County, Tennessee that included births from the years 2000 to 2002. Birth data was obtained from fixed-width electronic birth certificate files from the Tennessee Department of Health. Demographic data was obtained from the 2000 U.S. Census, Summary Files 1 and 3 [23]. Using the Arc GIS 9 software, the mothers' addresses were automatically matched with a digital street file. The digital street network used in this study was the Environmental System Research Institute (ESRI) street map, which was derived from the 2000 Census Topologically Integrated Geographic Encoding and Referencing system (TIGER) files [23]. The location of each address is shown in this example only as a generalized location in order to preserve the confidentiality of the individual records. Personal information was never identified or displayed in this study.
Birth weights of less than 2,500 g were defined as a low birth weight. There were a total of 42,394 births, including 4,794 low birth weight infants – comprising 11.30 % of the total births in Shelby County during the years of 2000 till 2002. After completing the address matching, each birth record was characterized by unique latitude and longitude location coordinates. We first analyzed these birth records as point patterns in their own right by applying the Spatial filtering technique and compared the results to SaTScan clusters.
Spatial filtering method
By address matching birth records to a digital road map, we were able to compute low birth weight rates for each location on a grid, which covers the entire Shelby county area at approximately 0.4 mile intervals. In order to make any general conclusions about the results of Spatial filtering, we used multiple filter sizes such as 0.4 mile, 0.5 and 0.6 miles. We conducted a sensitivity analysis using differing Spatial filter sizes and the 0.4 size appeared to provide the optimal distance for this study. Progressively larger Spatial filtering of data removes local spatial variability, which eventually produces an approximate uniform pattern of low birth weight. We did not aggregate cases to the geometric or geographic centroid of the administrative units such as zip codes, and census tracts. Given the approximate 0.4 mile distance interval between grid intersections, there were 5,928 grid points in Shelby County. Meaningful low birth weight rates were estimated for 841 grid points that had at least 40 births within their 0.4 mile vicinity. The Spatial filter area surrounding each of these points is the area from which an estimate of the low birth weight rate is made. We counted the normal births and low birth weights within the area and assigned the observed rate to the location. When we repeated this for a grid of such estimates, we could interpolate the low birth weight rate as a continuous spatial distribution. Neighbouring grid points share circular patterns that overlap, thereby sharing observations. Isarithmic maps with a constant range of values were constructed in GIS after the estimated rates were assigned to grid points.
For each birth location, we generated a random number from a uniform distribution in the range of 1 through 1,000. For each of the 841 grid point locations, 1,000 Monte Carlo simulations were made and the 1,000 different low birth weight rates were rank-ordered. The percent of the simulated rates at each grid location that were less than the observed rate for the same grid location were computed and the levels of statistical significance were portrayed as isolines. Because testing the rates against 1,000 simulations is a form of exploratory spatial analysis, methods of representing the results are discretionary, and the investigator can adjust the results based on level of significance. For example, the isolines representing the significance levels of 80 %, 85 %, 90 % and 95 % could be color-coded. Additionally, the isolines are overlaid on the significant SaTScan clusters for the purpose of comparison.
These probabilities, portrayed as isarithmic maps, show areas that have significantly high rates of low birth weight. The isarithmic maps have many advantages in comparison with other conventional thematic maps that provide an indication of the level of a disease by area. They are not constrained by the borders of geographic units, and sudden transitions between levels of two neighbouring areas are avoided [24]. We used the inverse-distance weighing interpolation technique in constructing the isarithmic maps. Since inverse distance weighing represents the average of the values of the surrounding points, weighed by the inverse of the distance to those points, the process is based on the assumption of positive spatial autocorrelation depicting a continuous gradient exists between points in a linear way [24].
SaTScan method
In our second method, we applied the SaTScan method developed by Kulldorff to detect local clusters [22]. We compared the smoothed maps with the data derived through SaTScan in order to determine whether areas with statistically significant rates were retained or smoothed out. The spatial level of data and time period used is the same one used previously in the Spatial filtering method. Since both methods perform better on the point data, we did not include any analyses of aggregated data. We chose the Bernoulli model, which required information about the location of a set of cases and controls. We selected low birth weight geocoded points as cases and normal birth weight geocoded points as the controls. We employed non-overlapping grid buffers spaced at 0.4 mile intervals and maximum spatial cluster sizes of 0.4 mile, 0.5 and 0.6 mile respectively just like in the previous method.
Results
Significant clusters by method
In Figure 1, 2, and 3, we display the isolines of 95 % level Spatial filtering clusters using 0.4, 0.5 and 0.6 miles filter sizes for lump sum years of 2000–2002. These are overlaid on the most likely significant clusters of SaTScan for the purpose of comparison. We evaluated the effect of changes in filter size by creating maps with different filter sizes. The high rates of low birth weight remained as we increased the filter sizes to 0.4, 0.5 to 0.6 miles. This indicates that these differences are less likely due to chance. The 0.4 mile filter size or smaller filter sizes showed the local variability much better than the larger filter sizes of 0.5 and 0.6 miles (Figure 1). The 0.4 mile filter size resulted to five clustered areas. On the other hand, clusters increased in size and additional clusters emerged towards the northeastern part of the county when we used the 0.5 mile filter size (Figure 2). As we increased the filter size to 0.6 mile, the localized clusters unioned to a larger uniform pattern covering the western portion of the county (Figure 3). Table 1 shows the total area of the clusters for both methods with varying filter sizes. Once we applied a larger spatial filter size as 0.8 mile, the Spatial filtering technique lost the ability to detect elevated rates except for the most densely populated areas of Memphis.
Figure 1 Areas with statistically significant high rates of low birth weights, Shelby, TN, 2000–2002. The maps show SaTScan clusters with a maximum spatial cluster size of 0.4 miles. It also shows significant Spatial filter clusters with a maximum 0.4 mile filter size.
Figure 2 Areas with statistically significant high rates of low birth weights, Shelby, TN, 2000–2002. The maps show SaTScan clusters with a maximum spatial cluster size of 0.5 miles. It also shows significant Spatial filter clusters with a maximum 0.5 mile filter size.
Figure 3 Areas with statistically significant high rates of low birth weights, Shelby, TN, 2000–2002. The maps show SaTScan clusters with a maximum spatial cluster size of 0.6 miles. It also shows significant Spatial filter clusters with a maximum 0.6 mile filter size.
Table 1 Area of clusters by method
Methods Filtering Sizes or Maximum Spatial Cluster Sizes (miles) Significance Total Area of Shelby County (sq. meter) Clustered Area (sq. meter) % of Total Shelby County
Spatial filtering 0.40 ≥ 0.95 level 6,676,687 0.32%
0.50 ≥ 0.95 level 27,191,960 1.32%
0.60 ≥ 0.95 level 119,492,954 5.80%
SaTScan 0.40 P1 (most likely) = 0.002 935,139 0.05%
P2 (secondary) < 0.05 2,061,075,094 1,523,635 0.07%
0.50 P1 (most likely) = 0.001 1,500,352 0.07%
P2 (secondary) < 0.05 2,828,068 0.14%
0.60 P1 (most likely) = 0.001 2,344,277 0.11%
P2 (secondary) < 0.05 5,145,388 0.25%
On the other hand, SaTScan gave consistent results with larger filter sizes. The areas that have statistically significant rates adjusted after multiple testing also showed up as high rate areas on the Spatial filtering smoothed maps of low birth weight rates. SaTScan clusters were discrete compared to the continuous distribution of Spatial filtering clusters. The most likely SaTScan clusters and the continuous Spatial filtering clusters were concentrated in the western portion of Shelby County. Secondary clusters with less significance emerged as the cluster size increased. The resultant clusters were circular in shape within the predefined maximum spatial cluster size. The most likely clusters appeared in the same locations with larger radii with the exception of new secondary clusters in the vicinity.
Table 2 illustrates maternal and familial characteristics by cluster estimation method and type. The ethnic, economic, and educational characteristics of mothers whose births are in the cluster are quite similar irrespective of filter size. With both methods, the clusters become increasingly heterogeneous as the filter size increases. Although one would expect more heterogeneity as the sample size increases, it is possible that outer portions of the cluster start to encroach upon better educated and more affluent communities. The only major difference between the two methods was the total number of births within each respective type of cluster. Spatial filtering clusters include almost 3 times as many total births than does SaTScan at the 0.4 mile filter size. This may due in part to the fact that the total area covered by Spatial filtering clusters is almost 3 times greater than the total area covered by SaTScan clusters.
Table 2 Maternal and familial characteristics by cluster estimation method and type
Methods Filtering Sizes or Maximum Spatial Cluster Sizes (miles) Significance Total Births within Cluster Maternal Ethnicity within Cluster Average Percentage of Mothers having Some College Education within Cluster Average Percentage of Families Below Poverty level
Caucasian African American Others
Spatial filtering 0.40 ≥ 0.95% level 1,250 1.4% 98.3% 0.2% 13.9% 48.7%
0.50 ≥ 0.95% level 2,509 2.1% 97.6% 0.3% 16.2% 40.6%
0.60 ≥ 0.95% level 7,288 6.5% 92.4% 1.1% 19.6% 36.0%
SaTScan 0.40 P1 (most likely) = 0.002 141 0.0% 100.0% 0.0% 14.9% 49.7%
P2 (secondary) < 0.05 311 1.0% 98.4% 0.6% 9.6% 61.2%
0.50 P1 (most likely) = 0.001 184 0.0% 100.0% 0.0% 12.5% 31.9%
P2 (secondary) < 0.05 437 5.0% 93.8% 1.1% 16.9% 34.7%
0.60 P1 (most likely) = 0.001 268 0.0% 99.6% 0.4% 11.6% 58.1%
P2 (secondary) < 0.05 820 3.0% 96.3% 0.6% 13.2% 43.2%
Discussion
The purpose of this study is to examine the spatial and population (e.g., socio-economic) characteristics of low birthweight using two different cluster estimation techniques. We compared the results of Kulldorff's Spatial Scan Statistic with the results of Rushton's Spatial filtering technique across increasing sizes of spatial filters (circle). Spatial filtering results did not show any particular area that was not statistically significant based on SaTScan. The high rates, which remain as the filter size increases to 0.4, 0.5 to 0.6 miles, respectively, indicate that these differences are less likely due to chance. The maternal characteristics of births within clusters differed considerably between the two methods. Progressively larger Spatial filters removed local spatial variability, which eventually produced an approximate uniform pattern of low birth weight.
As shown here both methods can be used for aggregated as well as point level data [21]. Although the two methods produced similar clusters there were some differences between them. Spatial filtering provided estimated risks for low birth weight incidence for each location in the map while SaTScan provides the statistical significance of the likely clusters after adjusting for multiple testing. Spatial filtering calculated risk estimates while using predetermined circle sizes defined either by geographical or "constant or near constant population size rather than constant geographic size" (p.2400) [21]. SaTScan on the other hand, uses circles of different sizes when searching over a grid [5]. The size of clusters identified by Spatial filtering in this example depended on the identified filter size, such as the radius of the circles.
As evidenced by this study, different filter sizes should be used to construct spatial filter maps when evaluating the two methods in terms of cluster size. Based on our findings, the Spatial filtering technique may provide many advantages over SaTScan. The resultant SaTScan detected cluster is a non-continuous circular shape, which often conceals spatial pattern – even though the actual geographic coordinates of each case and control are used through the Bernoulli method. Spatial filtering, on the other hand, treats low birth weight rates as a continuous spatial distribution.
The results from both softwares can easily be incorporated into GIS. SaTScan results provide the radius, latitude and longitude coordinates, and the P-value for the most likely clusters in a database or ASCII format. On the other hand, the Distance Mapping and Analysis Program (DMAP) produces morbidity rates using Spatial filters and tests for significance using Monte Carlo simulations. Its results are isarithmic maps that exhibit a continuous spatial distribution.
Spatial filtering and DMAP
DMAP provides more consistent results if the total number of points is within the 20,000 range and when a finer grid size is used. It is preferable to break up the data to smaller units and time periods such as annual years to receive consistent results with Monte Carlo simulations. In our analysis, DMAP ran smoothly with a maximum data size of three years (42,394 births) and a maximum filter size of 0.6 mile. We did not receive consistent results when we used larger data sets and larger filter sizes. We believe this is related to the memory constraints of the current version of the DMAP software. When used with a smaller dataset, varying filter sizes of Spatial filtering showed cluster patterns very well and with consistent results.
SaTScan and DMAP
With the current version of SatScan, we did not run into the problem of running large data sets. Specifically with our dataset, we ran SaTScan and got consistent results with more than 0.6 mile maximum cluster size. SaTScan can examine temporal effects much better than Spatial filtering method due to its ability to handle large data sets. With SaTScan, the use of discrete circular shapes represent only approximate locations of concentrated data counts [25]. Consequently, this technique does not provide useful information about the absolute proximity of clusters to point sources of contamination. However, some techniques have been proposed to improve SaTScan limitations [18,21].
Conclusion
This study suggests that both estimation methods provide a useful way to characterize the spatial aspects of this birth outcome. The literature has strongly advocated the use of GIS in surveillance of the maternal environment and its impact on birth outcomes [12,25-33]. Yet, there are relatively few efforts to integrate and/or compare analytical techniques. SaTScan and Spatial filtering cluster estimation methods produced noticeably different results from the same individual level birth data. SaTScan clusters are likely to differ from Spatial filtering clusters in terms of population characteristics and geographic area within clusters. Using the two methods in conjunction could provide more detail about the population and spatial features contained with each type of cluster.
First, the two methods yielded many significant spatial clusters of low birthweight in Shelby County. We fully expected to find these clusters since the birth outcomes in this Midsouth area are historically among the nation's worst. The annual infant mortality rate in parts of Memphis frequently exceed 16 per 1,000 live births as opposed to about 10 per 1,000 live births in the state and about 7 per 1,000 live births nationally [34]. In 2002, 14% of children in Memphis were born prematurely as opposed to 12% nationally, while approximately 12% of children were born with low birth weight as opposed to 8% nationally [35]. Since 1996, about 40% of child deaths under the age of 18 resulted from prematurity [36].
Second, Spatial filtering clusters appear to cover much more geographic area han did SaTScan clusters. As we discussed earlier, this is likely due to the difference in basic assumptions between the two models. The two methods differ in their estimation of significance; SaTScan accounts for multiple testing of the highest likelihood of occurrence for all possible cluster locations and sizes while Spatial filtering does not. In addition, there may be somewhat of a "ceiling effect" with SaTScan. This maximum value ensures that the detected clusters, regardless of their location and size, are clusters detected without any pre-selection bias. The maximum allowed value of a spatial cluster does not mean that one has to pre-specify the size of a cluster before running an analysis. It simply means the largest allowed cluster would contain 50% of the at-risk population in the study area. This maximum value is reasonable because a cluster is expected to concentrate in certain areas of the study region. If a cluster covers most of the study area, then the location and size of the study area is no longer meaningful in that study area. Consequently, SaTScan clusters have an inherent but adjustable "cap" on cluster size whereas Spatial filtering is somewhat less limited. The tendency or capacity of Spatial filtering to yield clusters with considerably more geographic variability than SaTScan raises the issue of sample generalizability. Although not demonstrated in this study, the potential for greater variability of the characteristics of births within a Spatial filtering cluster may provide some analytic benefits.
Third, the maternal and familial characteristics of births contained within the two methods were remarkably alike. Additionally, changing the level of geographic scale resulted in very similar patterns between the two methods with respect to these characteristics. As the level of scale increased the sample became increasingly heterogeneous. We know that geographic scale is an important consideration in such investigations irrespective of the analytic method used [7]. This study found that clusters of low birthweight in Shelby County might extend into less impoverished, better educated, and more ethnically diverse communities.
We do not contend that either cluster estimation method is inherently superior. Instead this study underscores the need for an exploratory, integrative, and multi-scalar approach to assessing geographic patterns of disease, since different methods identify different patterns. First, both methods should be compared again using a Poisson approach. A temporal analysis of low birth weight could be conducted. Secondly, the two methods should be compared using different forms of chronic morbidity (i.e., congenital anomalies). Since other diseases differ in prevalence, population, and/or spatial characteristics, the results of the two methods might differ accordingly. Finally, the two methods should also be compared with respect to additional spatial characteristics. For example, the density or distribution of point sources of pollutants within either type of cluster could be examined on varying geographic scales.
List of abbreviations
DMAP Distance Mapping and Analysis Program
GIS Geographic information systems
SaTScan Spatial Scan Statistic
ESRI Environmental System Research Institute
TIGER Topologically Integrated Geographic Encoding and Referencing System
Authors' contributions
BW obtained data for this research. EO and BW conceptualized and conducted the analysis, and drafted the manuscript. EO directed development and interpretation of the spatial analysis. SK worked with EO on the methodology for imputation. MM worked with BW in the preparation and editing of the manuscript. All authors participated in the preparation and approved the final version of the manuscript.
Acknowledgements
We thank the Tennessee Department of Health for providing birth outcome data.
==== Refs
Alexander FE Cuzick J Elliott P, Cuzick J, English D, Stern R Methods for the assessment of disease clusters Geographical & Environmental Epidemiology: Methods for Small-Area Studies 1992 Oxford, UK , Oxford University Press 238 247
Kulldorff M Prospective time periodic geographical disease surveillance using a scan statistic J R Stat Soc [Ser A] 2001 164 61 72 10.1111/1467-985X.00186
Rushton G Krishnamurti D Krishnamurthy R Song H Spatial analysis of infant mortality rates in Des Moines, Iowa, 1989-1992: Washington D.C.. 1995
Gregorio DI Samociuk H Breast cancer surveillance using gridded population units, Connecticut, 1992-1995 Ann Epidemiol 2003 13 42 49 12547484 10.1016/S1047-2797(02)00258-2
Cromley EK McLafferty SL GIS and public health 2002 New York , The Guilford Press
English PB Kharrazi M Davies S Scalf R Waller L Neutra R Changes in the spatial pattern of low birth weight in a southern California county: the role of individual and neighborhood level factors Soc Sci Med 2003 56 2073 2088 12697198 10.1016/S0277-9536(02)00202-2
Krieger N Chen JT Waterman PD Soobader MJ Subramanian SV Carson R Choosing area based socioeconomic measures to monitor social inequalities in low birth weight and childhood lead poisoning: The Public Health Disparities Geocoding Project (US) J Epidemiol Community Health 2003 57 186 199 12594195 10.1136/jech.57.3.186
Choi KM Serre ML Christakos G Efficient mapping of California mortality fields at different spatial scales J Expo Anal Environ Epidemiol 2003 13 120 133 12679792 10.1038/sj.jea.7500263
Diem JE A critical examination of ozone mapping from a spatial-scale perspective Environ Pollut 2003 125 369 383 12826415 10.1016/S0269-7491(03)00110-6
Schneider D Greenberg MR Donaldson MH Choi D Cancer clusters: The importance of monitoring multiple geographic scales Soc Sci Med 1993 37 753 759 8211291 10.1016/0277-9536(93)90369-F
Krieger N Chen JT Waterman PD Rehkopf DH Subramanian SV Race/ethnicity, gender, and monitoring socioeconomic gradients in health: a comparison of area-based socioeconomic measures--the public health disparities geocoding project Am J Public Health 2003 93 1655 1671 14534218
Richards T Croner C Rushton G Brown C Fowler L Geographic information systems and public health: mapping the future Public Health Rep 1999 114 359 373 10501137 10.1093/phr/114.4.359
Bonner MR Han D Nie J Rogerson P Vena JE Freudenheim JL Positional accuracy of geocoded addresses in epidemiologic research Epidemiology 2003 14 408 412 12843763
Brody JG Vorhees DJ Melly SJ Swedis SR Drivas PJ Rudel RA Using GIS and historical records to reconstruct residential exposure to large-scale pesticide application J Expo Anal Environ Epidemiol 2002 12 64 80 11859434 10.1038/sj.jea.7500205
Maantay J Mapping environmental injustices: pitfalls and potential of geographic information systems in assessing environmental health and equity Environ Health Perspect 2002 110 Suppl 2 161 171 11929725
McElroy JA Remington PL Trentham-Dietz A Robert SA Newcomb PA Geocoding addresses from a large population-based study: lessons learned Epidemiology 2003 14 399 407 12843762
Rushton G Public health, GIS, and spatial analytic tools Annu Rev Public Health 2003 24 43 56 12471269 10.1146/annurev.publhealth.24.012902.140843
Kafadar K Smoothing geographical data, particularly rates of disease Stat Med 1996 15 2539 2560 8961462 10.1002/(SICI)1097-0258(19961215)15:23<2539::AID-SIM379>3.3.CO;2-2
Rushton G Lolonis P Exploratory spatial analysis of birth defect rates in an urban population Stat Med 1996 15 717 726 9132899 10.1002/(SICI)1097-0258(19960415)15:7/9<717::AID-SIM243>3.0.CO;2-0
Talbot TO Kulldorff M Forand SP Haley VB Evaluation of spatial filters to create smoothed maps of health data Stat Med 2000 19 Part 17/18 2399 2408 10960861 10.1002/1097-0258(20000915/30)19:17/18<2399::AID-SIM577>3.3.CO;2-I
Kulldorff M A spatial scan statistic Communications in Statistics: Theory and Methods 1997 26 1481 1496
Census US Census 2000 Gateway
Turnbull BW Iwano EJ Burnett WS Howe HL Clark LC Monitoring for clusters of disease: application to leukemia incidence in upstate New York Am J Epidemiol 1990 132 (Supplement) S136 S143 2356825
Burkom HS Biosurveillance applying scan statistics with multiple, disparate data sources J Urban Health 2003 1 Supplement 1 57 65
Ali M Emch M Donnay JP Yunus M Sack RB Identifying environmental risk factors for endemic cholera: a raster GIS approach Health Place 2002 8 201 210 12135643 10.1016/S1353-8292(01)00043-0
Cockings S Dunn CE Bhopal RS Walker DR Users' perspectives on epidemiological, GIS and point pattern approaches to analysing environment and health data Health Place 2004 10 169 182 15019911 10.1016/j.healthplace.2003.09.001
Cromley EK GIS and disease Annu Rev Public Health 2003 24 7 24 12668753 10.1146/annurev.publhealth.24.012902.141019
Croner CM Public health, GIS, and the internet Annu Rev Public Health 2003 24 57 82 12543872
Dangendorf F Herbst S Reintjes R Kistemann T Spatial patterns of diarrhoeal illnesses with regard to water supply structures--a GIS analysis Int J Hyg Environ Health 2002 205 183 191 12040916 10.1078/1438-4639-00151
Elgethun K Fenske RA Yost MG Palcisko GJ Time-location analysis for exposure assessment studies of children using a novel global positioning system instrument Environ Health Perspect 2003 111 115 122 12515689
Reif JS Keefe T Stallones L Reproductive, Neurobehavioral, and Other Disorders in Communities Surrounding the Rocky Mountain Arsenal Govt Reports Announcements & Index (GRA&I) 1996
Stallones L Nuckols J Berry J Surveillance around hazardous waste sites: geographic information systems and reproductive outcomes Environ Res 1992 59 81 92 1425521 10.1016/S0013-9351(05)80227-0
TDH Tennessee Vital Statistics Summary Resident Data 2002 2004 Nashville, TN 1 9
MOD Peristats
Williams BL Descriptive Characteristics of Childhood Mortality in Shelby County Tennessee from 1995-2003 2004 Child Death Review Team - Memphis and Shelby County Health Department 15609827
|
16076402
|
PMC1190206
|
CC BY
|
2021-01-04 16:39:06
|
no
|
Int J Health Geogr. 2005 Aug 2; 4:19
|
utf-8
|
Int J Health Geogr
| 2,005 |
10.1186/1476-072X-4-19
|
oa_comm
|
==== Front
J NeuroinflammationJournal of Neuroinflammation1742-2094BioMed Central London 1742-2094-2-151596703510.1186/1742-2094-2-15ResearchInterleukin 1 receptor antagonist knockout mice show enhanced microglial activation and neuronal damage induced by intracerebroventricular infusion of human β-amyloid Craft Jeffrey M [email protected] D Martin [email protected] Emmet [email protected] Eldik Linda J [email protected] Center for Drug Discovery and Chemical Biology, Northwestern University, Chicago, IL, USA2 Cell and Molecular Biology, Northwestern University Feinberg School of Medicine, Chicago, IL, USA3 Molecular Pharmacology and Biological Chemistry, Northwestern University Feinberg School of Medicine, Chicago, IL, USA4 Obstetrics and Gynecology, Northwestern University Feinberg School of Medicine, Chicago, IL, USA5 Department of Obstetrics and Gynecology, Evanston Northwestern Healthcare, Evanston, IL, USA2005 20 6 2005 2 15 15 24 5 2005 20 6 2005 Copyright © 2005 Craft et al; licensee BioMed Central Ltd.2005Craft et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Interleukin 1 (IL-1) is a key mediator of immune responses in health and disease. Although classically the function of IL-1 has been studied in the systemic immune system, research in the past decade has revealed analogous roles in the CNS where the cytokine can contribute to the neuroinflammation and neuropathology seen in a number of neurodegenerative diseases. In Alzheimer's disease (AD), for example, pre-clinical and clinical studies have implicated IL-1 in the progression of a pathologic, glia-mediated pro-inflammatory state in the CNS. The glia-driven neuroinflammation can lead to neuronal damage, which, in turn, stimulates further glia activation, potentially propagating a detrimental cycle that contributes to progression of pathology. A prediction of this neuroinflammation hypothesis is that increased IL-1 signaling in vivo would correlate with increased severity of AD-relevant neuroinflammation and neuronal damage.
Methods
To test the hypothesis that increased IL-1 signaling predisposes animals to beta-amyloid (Aβ)-induced damage, we used IL-1 receptor antagonist Knock-Out (IL1raKO) and wild-type (WT) littermate mice in a model that involves intracerebroventricular infusion of human oligomeric Aβ1–42. This model mimics many features of AD, including robust neuroinflammation, Aβ plaques, synaptic damage and neuronal loss in the hippocampus. IL1raKO and WT mice were infused with Aβ for 28 days, sacrificed at 42 days, and hippocampal endpoints analyzed.
Results
IL1raKO mice showed increased vulnerability to Aβ-induced neuropathology relative to their WT counterparts. Specifically, IL1raKO mice exhibited increased mortality, enhanced microglial activation and neuroinflammation, and more pronounced loss of synaptic markers. Interestingly, Aβ-induced astrocyte responses were not significantly different between WT and IL1raKO mice, suggesting that enhanced IL-1 signaling predominately affects microglia.
Conclusion
Our data are consistent with the neuroinflammation hypothesis whereby increased IL-1 signaling in AD enhances glia activation and leads to an augmented neuroinflammatory process that increases the severity of neuropathologic sequelae.
Alzheimer's diseaseamyloid betaanimal modelglial activationinterleukin-1microglia
==== Body
Background
There is increasing evidence that CNS inflammation (termed neuroinflammation) driven by abnormal or prolonged glia activation contributes to the pathogenesis and progression of both acute and chronic disorders [1,2]. Normally, glia respond to stresses by a transient activation that serves a homeostatic function. However, increased levels of inflammatory and oxidative stress molecules produced by chronically activated glia can lead to neuron damage or death, which can induce further glial activation, thus leading to a self-propagating, detrimental cycle of neuroinflammation and neurodegeneration [3]. A large body of evidence [4-8] suggests that targeting this glia-neuron cycle represents an attractive potential strategy for development of new therapeutic approaches to AD that would alter disease progression. To this end, a more detailed understanding of the proteins, pathways, and inflammatory responses involved in neuroinflammation relevant to AD progression is critical.
One of the biochemical responses of glia to both acute and chronic conditions of brain damage is increased production of the pro-inflammatory cytokine IL-1. An extensive body of research strongly suggests that IL-1 has an integral role in AD pathogenesis and progression. First, analysis of AD brain tissue demonstrates IL-1 overproduction, primarily in the activated microglia that surround β-amyloid (Aβ) plaques and neurons containing neurofibrillary tangles [9,10], the two neuropathological hallmarks of AD. This finding is complemented by the revelation that this overproduction of IL-1 closely corresponds to the level of neuropathology found in a given brain region [11]. Second, cell-based studies show that IL-1 can elicit the production of a number of detrimental molecules from microglia, astrocytes, and neurons. For example, IL-1 can stimulate the production of α 1 anti-chymotrypsin, IL-6, S100B, and inducible nitric oxide synthase [12-15], which are themselves increased in the AD brains [2]. These molecules, either by themselves or by stimulating the production of other molecules, contribute to a neuroinflammatory cascade that has been suggested to result in cell injury, dysfunction, and death in AD[16]. This hypothesis is supported by the neuroprotection observed following suppression of the neuroinflammatory cascade in AD animal models [4,5]. Finally, multiple studies examining IL-1 genetics have shown that polymorphisms in the IL-1gnd IL-1 receptor genes increase the risk of AD by as much as three times in a homozygous carrier [17,16].
All these studies to date are consistent with the hypothesis that increased brain IL-1 levels or activity would correlate with increased severity of AD-relevant neuroinflammation and neuronal damage. To test this hypothesis, we used interleukin-1 receptor antagonist knockout (IL1raKO) mice, which have enhanced IL-1 signaling because of the loss of the IL-1 receptor's physiological antagonist. We induced AD-relevant neuroinflammation and neuronal damage by intracerebroventricular (ICV) infusion of human Aβ1–42 in a mouse experimental model previously developed by us [4,5], and determined the degree of glia activation and neuroinflammation and synaptic degeneration in the hippocampus. We report here that IL1raKO mice are significantly more susceptible than WT mice to the neuroinflammatory and neurodegenerative sequelae of Aβ infusion, supporting the concept that elevated IL-1 signaling in the brain participates in AD pathogenesis.
Methods
Interleukin-1 receptor antagonist knockout mice (IL1raKO)
IL1raKO mice were derived as previously described [18] and the colony maintained by mating of heterozygous littermates. Homozygous IL1raKO mice and WT littermates were selected following genotyping, and were allowed to mature until 16 weeks of age before surgery. All mice were kept at the Center for Comparative Medicine (CCM) at Northwestern University Feinberg School of Medicine. All animal procedures were approved by the Animal Care and Use Committee at Northwestern University.
Aβ infusion
ICV Infusion of human oligomeric Aβ1–42 or vehicle into IL1raKO and WT littermates was performed essentially as described [4]. Briefly, four-month-old mice (n = 5–12 per group) were anesthetized with 2% vaporized isoflurane, and an Alzet micro-osmotic pump (Durect, Model #1002) was attached to a pre-cut 2.5 mm long cannula (Plastics One) stereotaxically implanted into the right lateral cerebral ventricle (at coordinates -1.0 mm mediolateral, -0.5 mm anterioposterior from Bregma; -2.0 mm dorsal-ventral from skull). Pumps contained either vehicle (4 mM Hepes + 250 μg/ml human high-density lipoprotein, HDL) or oligomeric Aβ1–42 (45 μg; American Peptide) [19] dissolved in vehicle. Since HDL normally carries Aβ in serum, it was used in the pump to reduce Aβ aggregation and facilitate better delivery to the neuropil [20,21]. Osmotic pumps were partially coated with paraffin to a point 5 mm above the distal end of the pump. This slows the osmotic passage of water into the pumps' gel casings and has been shown in ex vivo experiments to reduce the infusion rate to ~1.6 μg/3.5 μl per day for ~28 days (data not shown).
At 42 days after the start of Aβ infusion, mice were anesthetized with pentobarbital (50 mg/kg) and perfused with a Hepes buffer containing a protease inhibitor cocktail. The superior portion of the cranium was then incised, and brains were removed and longitudinally bisected. In order to exclude the potential that one side of the brain may possess more significant pathology following a unilateral infusion and, therefore, confound the results and/or conclusions, only the right half of the brain was fixed in a paraformaldehyde/ phosphate buffer solution and embedded in paraffin for histological examination, while the hippocampus was isolated from only the left hemisphere and snap frozen for biochemical evaluation. In addition, the Alzet pumps were examined to insure that the paraffin coating was intact and the reservoir solution was expelled.
Biochemical analysis of inflammatory and neural markers
Hippocampal soluble extracts were prepared by dounce and sonication in a Hepes buffer containing a protease inhibitor cocktail followed by centrifugation and collection of supernatant as described [4]. Levels of the pro-inflammatory cytokines IL-1β, tumor necrosis factor (TNF)α, S100B, and the presynaptic protein synaptophysin in hippocampal supernatants were determined by ELISA as previously described [4]. Western blots of hippocampal supernatants (10 μg supernatant protein loaded per lane) were done with an antibody to postsynaptic density protein-95 kDa (PSD-95) (1:100,000 dilution; Upstate Biotechnology) as described [4]. Antibodies against β-actin (1:500,000 dilution, Sigma) were used to confirm equal protein loading among the samples. Densitometry was done with ImageQuant software (Molecular Dynamics).
Histology
Immunohistochemical detection of activated astrocytes and microglia was performed on 10 μm sections with anti-GFAP (1:1500 dilution; Sigma) and anti-F4/80 (1:100 dilution; Serotek) antibodies, respectively, as previously described [4]. Cell counts were determined by two blinded observers and subsequently analyzed as follows. For microglia and astrocyte analysis, all diaminobenzidine (DAB)-stained cell bodies were manually counted in the hippocampus (excluding the fimbria) of three F4/80- and GFAP- labeled sections positioned at -1.8, -2.1, and -2.3 mm from Bregma. In all studies, concordance between observers was within 10% or the section was removed from analysis.
Statistical analyses
Experimental and control groups were compared as four independent groups using one-way ANOVA with SNK post-hoc analysis using a statistical software package (GraphPad Prism version 4.00, GraphPad Software, San Diego CA, ). GraphPad Prism was also used to construct Kaplan-Meier mortality curves and assess for significance. Statistical significance was assumed when p < 0.05.
Results
Increased mortality in Aβ-infused IL1raKO mice
In our previous studies utilizing the Aβ-infusion model [4,5], we found that intra-, peri-, and post-operative animal mortality was approximately 1–2%. Mortality in IL1raKO mice that received an Aβ infusion was much higher, reaching 50% by the time of sacrifice at 42 days (Fig 1). In sharp contrast, no animal mortality was experienced in the IL1raKO mice that received a vehicle infusion, or in WT littermates infused with either Aβ or vehicle.
Figure 1 Increased mortality in IL1raKO mice during Aβ infusion. Alzet pumps containing Aβ1–42 or vehicle were surgically implanted in IL1raKO and WT littermate mice (n = 10–12 mice per Aβ-infused group; n = 5 mice per vehicle-infused group), and post-operative survival was monitored for 42 days. Kaplan-Meier survival curves show that WT mice infused with vehicle or Aβ, and IL1raKO mice infused with vehicle experienced no mortality during the time period. In contrast, Aβ-infused IL1raKO mice experienced a 50% mortality rate (6 of the 12 animals died before 42 days). This mortality was significantly different from the other experimental and control groups (error bars = SEM; p < 0.05).
Enhanced microglial responses in Aβ-infused IL1raKO mice
Based on the high mortality seen in Aβ-infused IL1raKO mice, the infusion experiment was repeated with additional mice to allow survival of enough KO mice for subsequent analyses. At 42 days after the start of surgery, mice were sacrificed and hippocampal tissue analyzed. Measurement of microglia activation endpoints (Fig 2) revealed no significant differences in the basal levels of the pro-inflammatory cytokines IL-1β (Fig 2A) and TNFα (Fig 2B) in vehicle-infused IL1raKO and WT mice. There was a slight increase in the numbers of F4/80 positive microglia in vehicle-infused IL1raKO mice compared with the WT counterparts (Fig 2C). However, in IL1raKO mice infused with Aβ, the intensity of the microglial response, as measured by several biochemical and histological endpoints, was much greater than in Aβ-infused WT mice. For example, IL-1β levels were significantly greater in Aβ-infused IL1raKO compared to Aβ-infused WT mice (Fig 2A). Likewise, TNFα levels were significantly higher following Aβ infusion in IL1raKO mice versus WT littermates (Fig 2B). Finally, the number of activated microglia as measured by F4/80 immunostaining was greater in Aβ-infused IL1raKO mice versus their WT counterparts (Fig 2C). Representative photomicrographs from the hippocampus of WT and IL1raKO mice infused with Aβ (Fig 2D and 2E, respectively) demonstrate the extent of this microglial activation.
Figure 2 Microglia activation following Aβ infusion. WT and IL1raKO mice infused with Aβ or vehicle for 28 days were sacrificed on day 42 (n = 5–10 mice/group survived for analysis). Brains were bisected and the right side of the brain was processed for immunohistochemistry while the left hippocampus was dissected and used for biochemical analysis. A) Levels of the pro-inflammatory cytokine IL-1β were significantly higher in IL1raKO mice infused with Aβ compared to WT mice infused with Aβ. B) TNFα levels also showed a stronger upregulation in Aβ-infused IL1raKO mice compared to Aβ-infused WT mice. C) F4/80 immunostaining for activated microglia also revealed a significant increase in IL1raKO mice infused with Aβ versus WT mice infused with Aβ. Representative photomicrographs of F4/80-positive microglia in the hippocampus of a D) WT mouse infused with Aβ, and E) IL1raKO mouse infused with Aβ. Arrowheads point to microglia cell bodies. Bar in D-E = 50 μm (error bars = SEM; * Significantly different, p < 0.01; ***Significantly different, p < 0.001).
Astrocyte activation in Aβ-infused IL1raKO mice
Unlike the findings above with microglia endpoints, we observed no significant differences in astrocyte activation state between IL1raKO and WT mice following Aβ infusion. For example, levels of the astrocyte-derived cytokine S100B were significantly upregulated following Aβ infusion for both the WT and IL1raKO mice (Fig 3A); however, there was no significant difference in the magnitude of this increase between the Aβ-infused WT and IL1raKO mice. These results were also seen by glial fibrillary acidic protein (GFAP) immunohistochemistry. As shown in Fig 3B, there were significant increases in GFAP staining in the hippocampus of all mice following Aβ infusion, and the numbers of GFAP-positive astrocytes were similar in WT and IL1raKO mice.
Figure 3 Astrocyte activation following Aβ infusion. WT and IL1raKO mice were infused with Aβ or vehicle, and brains prepared as in Figure 2. A) Levels of the pro-inflammatory astrocyte-derived cytokine S100B showed a similar degree of upregulation in Aβ-infused IL1raKO and WT mice. B) Numbers of GFAP-positive astrocytes were increased to a similar degree in both WT and IL1raKO mice infused with Aβ. (error bars = SEM; p > 0.05 between Aβ-infused IL1raKO and WT mice).
Increased synaptic degeneration in Aβ-infused IL1raKO mice
Given the significant increase in microglial neuroinflammation following Aβ infusion in IL1raKO mice, it was important to investigate the effect of this enhanced neuroinflammation on neuronal responses. Therefore, two different biochemical markers of synaptic degradation were examined. Aβ infusion led to a reduction in levels of the postsynaptic protein PSD-95 in both the IL1raKO and WT mice compared to vehicle-infused mice; however, the reduction was significantly greater in IL1raKO mice compared to WT (Fig 4A). Similarly, levels of the presynaptic protein synaptophysin were reduced in Aβ-infused mice compared to vehicle-infused mice and, while not quite reaching statistical significance (p = 0.07), the reduction in synaptophysin levels was greater in Aβ-infused IL1raKO mice compared to Aβ-infused WT (Fig 4B).
Figure 4 Loss of synaptic markers following Aβ infusion. WT and IL1raKO mice were infused with Aβ or vehicle, and brains prepared as in Figure 2. A) Aβ-infused mice had reduced hippocampal PSD-95 levels compared to vehicle-infused mice, and there was a significantly larger decrease in IL1raKO mice infused with Aβ versus their WT counterparts (error bars = SEM; * p < 0.05). B) The presynaptic marker synaptophysin was reduced in Aβ-infused mice compared to the vehicle-infused mice. In addition, the reduction in synaptophysin in Aβ-infused mice was greater in IL1raKO mice compared to WT mice, although the difference did not quite reach statistical significance (error bars = SEM; p = 0.07).
Discussion
The principle finding of this study is that enhanced IL-1 signaling results in increased mortality, microglial neuroinflammation, and neuronal damage following a chronic infusion of human Aβ1–42 in a murine model. These data provide further support for the idea that IL-1 is an important component in the neuroinflammation cascade that drives AD progression.
An extensive body of evidence indicates the importance of IL-1 in regulating susceptibility to CNS injury. For example, IL-1β levels in cerebrospinal fluid (CSF) are substantially increased shortly after severe traumatic brain injury in humans, and the magnitude of this increase is directly proportional to intracranial pressure [22]. Animal studies have also demonstrated the importance of IL-1 in mediating damage following both an acute insult, such as neonatal hypoxia-ischemia [23], and the progressive neurodegeneration that follows mild acute insults in rodents [24]. This is not unexpected given the array of potentially detrimental molecules produced by the CNS in response to increased production of IL-1. For example, IL-1β and/or IL-1α have been implicated in the production of other pro-inflammatory cytokines such as S100B [14]. Furthermore, IL-1β can stimulate glial iNOS production [15], which in turn can greatly increase the oxidative stress experienced by the brain and potentially lead to neuronal damage through protein nitration pathways [25].
More relevant to the current study, there is increased IL-1 signaling in chronic neurodegenerative diseases. In addition to the IL-1 overexpression and disease-relevant distribution in AD [26,10], IL-1 is also increased in other chronic conditions that involve neurodegeneration. These include Down's syndrome [26], which possesses many of the neuropathological hallmarks of AD, Creutzfeldt-Jakob disease [27], and HIV dementia [28]. In particular, in vivo rodent models of AD have also revealed a correlation between the extent of neuropathology and the level of IL-1 production [4,5,21]. Most importantly, a number of different therapeutic interventions targeted towards decreasing neuroinflammation have been shown to both decrease IL-1 production and reduce the amount of synaptic degeneration and neuron death [8,4,6]. These observations support the utility of measuring IL-1β levels, in terms of demonstrating a linkage to disease progression and monitoring response to therapeutic interventions that result in attenuation of disease.
The results of the current study, in which a rodent model that has increased IL-1 signaling due to loss of the IL-1 receptor's physiologic antagonist shows enhanced Aβ-induced neuroinflammation and neuronal damage, are consistent with previous work in the field. The increases in TNFα levels and F4/80-positive cells document that enhanced IL-1 signaling stimulates a robust and generalized microglia response following Aβ infusion. These observations also illustrate the escalating, cyclical nature of the Aβ-induced neuroinflammatory response, since with enhanced IL-1 signaling there are also increased levels of IL-1β itself. This is similar to findings with the IL1raKO mouse in models of systemic inflammation [18]. In addition, the resultant increased neuroinflammation in the IL1raKO mice infused with Aβ was accompanied by an exacerbation in the loss of synaptic markers, especially PSD-95. This particular finding, in conjunction with our similar findings in Aβ-infused S100B overexpressing transgenic mice [29], strongly argues for the conclusion that animals predisposed to neuroinflammation suffer more severely from neurodegenerative sequelae following Aβ infusion. Evidence from the epidemiological assessment of AD risk factors also supports this conclusion. Previous head injury, for example, is a significant environmental risk factor for development of AD in which it is hypothesized that IL-1-mediated neuroinflammation plays a key role [30,31].
A somewhat surprising finding was that, unlike the enhanced microglia and neuronal responses in the Aβ-infused IL1raKO mice compared to WT mice, the astrocyte responses to Aβ infusion were very similar in the two mouse strains. Both IL1raKO and WT mice showed similar upregulation of S100B levels and GFAP immunoreactivity after Aβ infusion. A possible explanation is that, at the time point examined (42 days), astrocyte responses had not yet reached their maximum following Aβ infusion. This possibility indicates a need for future studies to examine the temporal development of microglia, astrocyte, and neuronal responses after start of Aβ infusion.
The IL1raKO mice infused with Aβ experienced extensive mortality during the course of the experiment, despite minimal mortality of other strains of mice in our previous studies [4,5,29]. At first inspection, this increased mortality could be explained by the pro-inflammatory status of IL1raKO mice, which may predispose them to systemic septic-like episodes at a higher frequency than their WT littermates, especially following a major surgical operation to place an indwelling pump and ICV catheter. However, the lack of mortality in the IL1raKO mice that received a vehicle infusion would argue against this conclusion. A more intriguing possibility is that these mice died either directly or indirectly from a more severe neuroinflammatory response to Aβ than the mice that survived. The robust and consistent neuroinflammation, which is one of the key hallmarks that characterizes the Aβ infusion model, supports this conclusion as a distinct possibility. While quite interesting, especially in light of a similar syndrome afflicting a subset of individuals enrolled in the now discontinued Aβ vaccine trials [32], elucidation of the mechanisms underlying the increased mortality will require additional research.
Conclusion
The major finding of this study is the demonstration that IL1raKO mice show selective up-regulation of microglial neuroinflammation and increased neuronal damage following Aβ infusion when compared to WT littermates. The susceptibility of the IL1raKO mice to increased Aβ-induced neuroinflammation was demonstrated by biochemical and histological measurements of microglial activation. This increase in microglial activation in the IL1raKO mice is also associated with an increase in the degree of synaptic degeneration observed following Aβ infusion, suggesting that enhanced IL1 signaling leads to deleterious neuroinflammation that either directly damages neurons and/or potentiates the neurotoxic effects of Aβ. These data provide further support for the hypothesis that increases in the level of IL1 signaling in the AD brain can be detrimental through the cytokine's role as a key component of the neuroinflammatory cascade that contributes to progression of neuropathology. It also suggests that manipulation of IL-1 signaling and other neuroinflammatory mediators and pathways could be utilized to develop clinically meaningful, disease-modifying AD therapies.
Competing interests
The authors declare that they have no competing interests in the outcome, results, or conclusions of these studies.
Authors' contributions
JMC helped conceive the study and conducted animal surgeries, care, and biochemical/ histological assays. DMW helped conceive the study, interpret the results, and assist in the preparation of the manuscript. EH developed and provided the IL1raKO mice and gave helpful advice for handling and care of the animals. LVE helped conceive the study, analyze data and assist in the preparation of the manuscript.
Acknowledgements
We thank Sara Medgysi for assistance with mouse colony maintenance and assays. These studies were supported in part by the Institute for the Study of Aging (DMW) and by NIH grants R37 AG13939 (LVE), R01 AG20243 (LVE), and P01 AG21184 (LVE, DMW). JMC is a predoctoral fellow of the Center for Drug Discovery and Chemical Biology training program, supported in part by NIH T32 AG00260, a predoctoral fellowship from the Pharmaceutical Research and Manufacturers of America Foundation (PhRMA), and a Ruth L. Kirschstein NRSA fellowship F30 NS46942.
==== Refs
Allan SM Rothwell NJ Inflammation in central nervous system injury Philos Trans R Soc Lond B Biol Sci 2003 358 1669 1677 14561325 10.1098/rstb.2003.1358
Akiyama H Barger S Barnum S Bradt B Bauer J Cole GM Cooper NR Eikelenboom P Emmerling M Fiebich BL Finch CE Frautschy S Griffin WS Hampel H Hull M Landreth G Lue L Mrak R Mackenzie IR McGeer PL O'Banion MK Pachter J Pasinetti G Plata-Salaman C Rogers J Rydel R Shen Y Streit W Strohmeyer R Tooyoma I Van Muiswinkel FL Veerhuis R Walker D Webster S Wegrzyniak B Wenk G Wyss-Coray T Inflammation and Alzheimer's disease Neurobiol Aging 2000 21 383 421 10858586 10.1016/S0197-4580(00)00124-X
Griffin WST Sheng JG Royston MC Gentleman SM McKenzie JE Graham DI Roberts GW Mrak RE Glial-Neuronal interactions in Alzheimer's disease: the potential role of a 'cytokine cycle' in disease progression Brain Pathol 1998 8 65 72 9458167
Craft JM Watterson DM Frautschy SA Van Eldik LJ Aminopyridazines inhibit beta-amyloid-induced glial activation and neuronal damage in vivo Neurobiol Aging 2004 25 1283 1292 15465624 10.1016/j.neurobiolaging.2004.01.006
Craft JM Van Eldik LJ Zasadzki M Hu W Watterson DM Aminopyridazines attenuate hippocampus-dependent behavioral deficits induced by human beta-amyloid in a murine model of neuroinflammation J Mol Neurosci 2004 24 115 122 15314259 10.1385/JMN:24:1:115
Hu W Ranaivo HR Craft JM Van Eldik LJ Watterson DM Validation of the neuroinflammation cycle as a drug discovery target using integrative chemical biology and lead compound development with an Alzheimer's disease-related mouse model Cur Alzheim Res 2005 2 197 205 10.2174/1567205053585828
Mirzoeva S Sawkar A Zasakzki M Guo L Velentza AV Dunlap V Bourguignon JJ Ramstrom H Haiech J Van Eldik LJ Watterson DM Discovery of a 3-amino-6-phenyl-pyridazine derivative as a new synthetic antineuroinflammatory compound J Med Chem 2002 45 563 566 11806708 10.1021/jm015573g
Lim GP Yang F Chu T Chen P Beech W Teter B Tran T Ubeda O Ashe KH Frautschy SA Cole GM Ibuprofen suppresses plaque pathology and inflammation in a mouse model for Alzheimer's disease J Neurosci 2000 20 5709 5714 10908610
Griffin WS Sheng JG Roberts GW Mrak RE Interleukin-1 expression in different plaque types in Alzheimer's disease: significance in plaque evolution J Neuropathol Exp Neurol 1995 54 276 281 7876895
Sheng JG Mrak RE Griffin WS Glial-neuronal interactions in Alzheimer disease: progressive association of IL-1alpha+ microglia and S100beta+ astrocytes with neurofibrillary tangle stages J Neuropathol Exp Neurol 1997 56 285 290 9056542
Sheng JG Mrak RE Griffin WS Microglial interleukin-1 alpha expression in brain regions in Alzheimer's disease: correlation with neuritic plaque distribution Neuropathol Appli Neurobiol 1995 21 290 301
Das S Potter H Expression of the Alzheimer amyloid-promoting factors α1-antichymotrypsin and apolipoprotein E is induced in astrocytes by IL-1 Neuron 1995 14 447 456 7857652 10.1016/0896-6273(95)90300-3
Woiciechowsky C Schoning B Stoltenburg-Didinger G Stockhammer F Volk HD Brain-IL-1 beta triggers astrogliosis through induction of IL-6: inhibition by propranolol and IL-10 Med Sci Monit 2004 10 BR325 330 15328477
Sheng JG Ito K Skinner RD Mrak RE Rovnaghi CR Van Eldik LJ Griffin WS In vivo and in vitro evidence supporting a role for the inflammatory cytokine interleukin-1 as a driving force in Alzheimer pathogenesis Neurobiol Aging 1996 17 761 766 8892349 10.1016/0197-4580(96)00104-2
Akama KT Van Eldik LJ Beta-amyloid stimulation of inducible nitric-oxide synthase in astrocytes is interleukin-1beta- and tumor necrosis factor-alpha (TNFalpha)-dependent, and involves a TNFalpha receptor-associated factor- and NFkappaB-inducing kinase-dependent signaling mechanism J Biol Chem 2000 275 7918 7924 10713108 10.1074/jbc.275.11.7918
Mrak RE Griffin WS Interleukin-1, neuroinflammation, and Alzheimer's disease Neurobiol Aging 2001 22 903 908 11754997 10.1016/S0197-4580(01)00287-1
Nicoll JA Mrak RE Graham DI Stewart J Wilcock G MacGowan S Esiri MM Murray LS Dewar D Love S Moss T Griffin WS Association of interleukin-1 gene polymorphisms with Alzheimer's disease Ann Neurol 2000 47 365 368 10716257 10.1002/1531-8249(200003)47:3<365::AID-ANA13>3.0.CO;2-G
Hirsch E Irikura VM Paul SM Hirsh D Functions of interleukin 1 receptor antagonist in gene knockout and overproducing mice Proc Natl Acad Sci U S A 1996 93 11008 11013 8855299 10.1073/pnas.93.20.11008
Dahlgren KN Manelli AM Stine WB JrBaker LK Krafft GA LaDu MJ Oligomeric and fibrillar species of amyloid-beta peptides differentially affect neuronal viability J Biol Chem 2002 277 32046 32053 12058030 10.1074/jbc.M201750200
Frautschy SA Yang F Calderon L Cole GM Rodent models of Alzheimer's disease: rat Aβ infusion approaches to amyloid deposits Neurobiol Aging 1996 17 311 321 8744413 10.1016/0197-4580(95)02073-X
Frautschy SA Hu W Kim P Miller SA Chu T Harris-White ME Cole GM Phenolic anti-inflammatory antioxidant reversal of Aβ-induced cognitive deficits and neuropathology Neurobiol Aging 2001 22 993 1005 11755008 10.1016/S0197-4580(01)00300-1
Hayakata T Shiozaki T Tasaki O Ikegawa H Inoue Y Toshiyuki F Hosotubo H Kieko F Yamashita T Tanaka H Shimazu T Sugimoto H Changes in CSF S100B and cytokine concentrations in early-phase severe traumatic brain injury Shock 2004 22 102 107 15257081 10.1097/01.shk.0000131193.80038.f1
Hu X Nesic-Taylor O Qiu J Rea HC Fabian R Rassin DK Perez-Polo JR Activation of nuclear factor-kappaB signaling pathway by interleukin-1 after hypoxia/ischemia in neonatal rat hippocampus and cortex J Neurochem 2005 93 26 37 15773902 10.1111/j.1471-4159.2004.02968.x
Basu A Lazovic J Krady JK Mauger DT Rothstein RP Smith MB Levison SW Interleukin-1 and the interleukin-1 type 1 receptor are essential for the progressive neurodegeneration that ensues subsequent to a mild hypoxic/ischemic injury J Cereb Blood Flow Metab 2005 25 17 29 15678109 10.1038/sj.jcbfm.9600002
Sarchielli P Galli F Floridi A Floridi A Gallai V Relevance of protein nitration in brain injury: a key pathophysiological mechanism in neurodegenerative, autoimmune, or inflammatory CNS diseases and stroke Amino Acids 2003 25 427 36 14661102 10.1007/s00726-003-0028-6
Griffin WS Stanley LC Ling C White L MacLeod V Perrot LJ White CL 3rdAraoz C Brain interleukin 1 and S-100 immunoreactivity are elevated in Down syndrome and Alzheimer disease Proc Natl Acad Sci U S A 1989 86 7611 7615 2529544
Van Everbroeck B Dewulf E Pals P Lubke U Martin JJ Cras P The role of cytokines, astrocytes, microglia and apoptosis in Creutzfeldt-Jakob disease Neurobiol Aging 2002 23 59 64 11755020 10.1016/S0197-4580(01)00236-6
Mrak RE Griffin WS The role of chronic self-propagating glial responses in neurodegeneration: implications for long-lived survivors of human immunodeficiency virus J Neurovirol 1997 3 241 246 9291232
Craft JM Watterson DM Marks A Van Eldik LJ Enhanced susceptibility of S-100B transgenic mice to neuroinflammation and neuronal dysfunction induced by intracerebroventricular infusion of human beta-amyloid Glia 2005 available online Apr 4
Griffin WS Sheng JG Gentleman SM Graham DI Mrak RE Roberts GW Microglial interleukin-1 alpha expression in human head injury: correlations with neuronal and neuritic beta-amyloid precursor protein expression Neurosci Lett 1994 176 133 136 7830934 10.1016/0304-3940(94)90066-3
Mortimer JA van Duijn CM Chandra V Fratiglioni L Graves AB Heyman A Jorm AF Kokmen E Kondo K Rocca WA EURODEM Risk Factors Research Group Head trauma as a risk factor for Alzheimer's disease: a collaborative re-analysis of case-control studies Int J Epidemiol 1991 20 S28 35 1833351
Orgogozo JM Gilman S Dartigues JF Laurent B Puel M Kirby LC Jouanny P Dubois B Eisner L Flitman S Michel BF Boada M Frank A Hock C Subacute meningoencephalitis in a subset of patients with AD after Abeta42 immunization Neurology 2003 61 46 54 12847155
|
15967035
|
PMC1190207
|
CC BY
|
2021-01-04 16:38:22
|
no
|
J Neuroinflammation. 2005 Jun 20; 2:15
|
utf-8
|
J Neuroinflammation
| 2,005 |
10.1186/1742-2094-2-15
|
oa_comm
|
==== Front
J Neuroengineering RehabilJournal of NeuroEngineering and Rehabilitation1743-0003BioMed Central London 1743-0003-2-261609553310.1186/1743-0003-2-26ResearchStride-to-stride variability while backward counting among healthy young adults Beauchet Olivier [email protected] Véronique [email protected] François R [email protected] Reto W [email protected] Laboratory of Physiology and Physiopathology of Exercise and Handicap, Faculty of Medicine, University of Saint-Etienne, France2 Department of Geriatrics, Saint-Etienne University Hospitals, Saint-Etienne, France3 Department of Rehabilitation and Geriatrics, Geneva University Hospitals, Geneva, Switzerland2005 11 8 2005 2 26 26 4 3 2005 11 8 2005 Copyright © 2005 Beauchet et al; licensee BioMed Central Ltd.2005Beauchet et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Little information exists about the involvement of attention in the control of gait rhythmicity. Variability of both stride time and stride length is closely related to the control of the rhythmic stepping mechanism. We sought 1) to determine whether backward counting while walking could provoke significant gait changes in mean values and coefficients of variation of stride velocity, stride time and stride length among healthy young adults; and 2) to establish whether change in stride-to-stride variability could be related to dual-task related stride velocity change, attention, or both.
Methods
Mean values and coefficients of variation of stride velocity, stride time and stride length were recorded using the Physilog®-system, at a self-selected walking speed in 49 healthy young adults (mean age 24.1 ± 2.8 years, women 49%) while walking alone and walking with simultaneous backward counting. Performance on backward counting was evaluated by recording the number of figures counted while sitting alone and while walking.
Results
Compared with walking alone, a significant dual-task-related decrease was found for the mean values of stride velocity (p < 0.001), along with a small but significant increase for the mean values and coefficients of variation of stride time (p < 0.001 and p = 0.015, respectively). Stride length parameters did not change significantly between both walking conditions. Dual-task-related increase of coefficient of variation of stride time was explained by changing stride velocity and variability between subjects but not by backward counting. The number of figures counted while walking decreased significantly compared to backward counting alone. Further, the dual-task related decrease of the number of enumerated figures was significantly higher than the dual-task related decrease of stride velocity (p = 0.013).
Conclusion
The observed performance-changes in gait and backward counting while dual tasking confirm that certain aspects of walking are attention-demanding in young adults. In the tested group of 49 young volunteers, dual tasking caused a small decrease in stride velocity and a slight increase in the stride-to-stride variability of stride time, while stride velocity variability was not affected by the attention-demanding task. The increase in stride time variability was apparently the result of a change in gait speed, but not a result of dual tasking. This suggests that young adults require minimal attention for the control of the rhythmic stepping mechanism while walking.
Dual-taskStride-to-stride variabilityAttentionGait controlHealthy young adults
==== Body
Background
Dual-task related gait changes are usually interpreted as interference caused by competing demands for limited attentional resources [1], highlighting the idea that walking is not only an automatic process but also an attention-demanding task. For example, it has been shown that healthy young adults devote attention to the control of balance during single-limb support in an anxiety provoking condition [2]. The involvement of attention in the control of the walking-related rhythmic stepping mechanism remains less clear, with only a few and contradictory published results in the literature [3-7].
Stride time and stride length variability are both parameters that are related to the control of the rhythmic stepping mechanism [8]. In motor control in general, high variability is related to major attention involvement [9], whereas low variability reflects automatic processes that require minimal attention [9,10]. Performing a motor task while walking, such as carrying a cup [7] has been related to an increased variability of stride time, but not stride length. Verbal fluency task is a frequently used attention-demanding task in dual-task paradigm [1]. In contrast to attention-demanding motor tasks and to their older counterparts, young adults showed no significant change in stride-to-stride variability while performing a verbal fluency task [3,5,6]. Recently, Beauchet et al. [11] reported that, compared to a verbal fluency task, backward counting out loud from 50 significantly increased the coefficient of variation (CV) of stride time in a group of older adults aged 75 years and older who had a broad range of cognitive function abilities (e.g., some had mild dementia). The authors suggested that these findings could be explained by a possible age-related difficulty in the ability to appropriately allocate attention between both tasks due to a major competitive interaction with executive function while dual tasking. Little information is available about the impact of backward counting on stride-to-stride variability in healthy young adults. The only published study using this mental arithmetic task in a small group of healthy young adults showed that the CV of stride length while backward counting did not change compared with walking alone [12]. No data are available about the impact of backward counting on stride time variability.
Previous studies have shown that stride time variability increases when stride velocity decreases [13-16]. Because stride velocity often decreases under dual-task condition [1,12], dual-task related increase in CV of stride time could be provoked either by stride velocity decrease, the attention-demanding task, or both. The understanding of the role of stride velocity, as a potential confounder in the relationship between stride time variability and the involvement of attention in gait control is important. In contrast to stride time variability, variability of stride length in young adults remained low across different gait speeds while walking alone [17]. Furthermore, no significant stride length changes appeared under dual-task condition [3,5,6]. Such results suggest a constant stereotype pattern for stride length regulation, independent of gait speed.
We hypothesized that backward counting could provoke significant changes in stride time variability but not in stride length variability related to different attention involvement, independently of dual-task related changes in stride velocity among healthy young adults. The aim of this study was 1) to determine whether backward counting while walking could provoke significant gait changes regarding mean values and coefficients of variation of stride velocity, stride time and stride length among healthy young adults; and 2) to establish whether possible significant changes in stride-to-stride variability could be related to dual-task related stride velocity changes, backward counting, or both.
Methods
Participants
Forty-nine healthy young adults (25 men and 24 women, mean age 24.1 ± 2.8 years, range: 20–30 years) were recruited from the campus of Saint-Etienne University after having given their written informed consent. The young adults reported no physical and mental disorders. They took no medication. The study was approved by the local ethics committee and conducted in accordance with the ethical standards set forth in the declaration of Helsinki (1983).
Tasks
The participants were asked to perform, in randomized order, the following tasks to the best of their capacity: counting backward aloud starting from 50 while sitting on a chair and while walking. For the dual-task condition, subjects were not specifically instructed to prioritize either one of both tasks, but were asked to perform the combined task at their best and at normal self-selected walking speed. Before testing, a trained evaluator gave standardized verbal instructions regarding the test procedure with visual demonstration of the walking test. To familiarize participants to the Physilog®-system [18,19], subjects completed 2 walking trials before recording. Participants' subjectively perceived gait safety while walking was measured with a visual analogue scale (score from 0 = safe to 10 = very unsafe) after each walking trial. All subjects reported zeroes under both conditions. Each subject completed one trial for each recorded walking condition. The subjects walked on a 20-meter walkway in a well-lit environment, at self-selected speed, and wearing their own footwear.
Apparatus
Stride parameters were obtained using Physilog® [18,19]. Physilog® is a validated ambulatory gait analysis system based on miniature kinematic sensors (i.e. gyroscopes) attached on body segments and connected to a portable data logger worn at the waist. In this study, lower limb movement during walking was measured using 4 miniatures gyroscopes (Murata, ENC-03J) attached with a rubber band, respectively, to each shank and each thigh. After each walking trial, data were transferred from the data logger to a personal portable computer via an interface cable for analysis and storage. The temporal and spatial gait parameters were estimated from the angular velocity of the lower limbs. Gait phases were determined from the precise moments of heel-strike and toe-off. These events gave rise to distinctive features of the shank angular velocity signals in the form of rather negative peaks. An algorithm based on wavelet transformation was used to enhance the estimated times and, thus, to determine mean gait cycle duration (i.e., stride time). Mean stride length was calculated on the basis of double-segment gait model involving both shank and thigh. Mean stride velocity was defined as the average of all strides' instantaneous walking speeds, calculated from mean stride length and mean stride time.
Study variables and outcomes
Stride velocity, stride time and stride length were measured during walking on a 20-meter walkway with Physilog® [18,19]. To assure that gait parameters were collected while steady state walking, the first and last 2.5 meters corresponding to the acceleration and deceleration phase of each trial were excluded from analysis. The enumerated figures (i.e., subtractions of one) and errors of subtractions were recorded with a tape recorder. We defined the number of enumerated figures while walking as the number achieved during the time interval needed to walk over the 15 meters distance. The corresponding number at rest was defined as the number of figures that participants enumerated during the same time interval while sitting on a chair.
The following outcomes were used: 1) mean and standard deviation of mean values of stride velocity, stride time, stride length and number of enumerated figures under single and dual-task condition; 2) mean and standard deviation of CV (CV = ([standard deviation/mean] × 100) of stride velocity, stride time and stride length; and 3) normalized dual-task-related variation of gait speed and counting performance expressed as mean and standard deviation of dual-task-related mean value changes in stride velocity and number of enumerated figures under dual-task condition, calculated with following formula:
Statistical analysis
Main outcome measures such as stride velocity, stride time and stride length were summarized using means and standard deviations. The normality of the parameters' distribution was verified with a skewness and kurtosis tests before and after applying usual transformations to normalize non-Gaussian variables by taking the logarithmic transformation. First, all comparisons of the main outcome measures were performed with paired samples t-test. Second, two balanced analysis of covariance (ANCOVA) with a repeated measures design was performed, once for mean stride time and a second time for CV of stride time, to estimate the effects of counting backward, stride velocity and subjects (corresponding to the variability between subjects) without interaction terms, while adjusting for walking speed. For computing the error term, subjects were nested within walking conditions. P < 0.05 was considered statistically significant. All statistics were performed using the Stata Statistical Software 2003.
Results
As shown in Table 1, dual-task related decrease in mean value of stride velocity was significant compared to walking alone (p < 0.001), whereas the CV of stride velocity did not change significantly (p = 0.097). Both mean value and CV of stride time were significantly higher while backward counting compared to walking alone (respectively, p < 0.001 for mean value and p = 0.015 for CV). No significant dual-task related changes in mean value and CV of stride length were found compared to walking alone (p = 0.414 and p = 0.275). Furthermore, significantly fewer figures were enumerated under dual-task than under single-task condition (p < 0.001). All subjects performed the mental arithmetic task without errors of subtractions. The ANCOVA models (Tables 2 and 3) revealed that both stride time parameters were significantly associated with walking speed and subject's effect (p < 0.010) but not with the simultaneous task of backward counting (p = 0.227 for mean value and p = 0.330 for CV). Moreover, R-squared values showed that the variance explained by the ANCOVA models was high for the mean value and CV of stride time (respectively 0.98 and 0.83). As the interaction term between task and velocity was neither significant for means of stride time nor for CV of stride time, we only reported the models without interaction. As depicted in Figure 1, the number of enumerated figures showed a higher decrease from walking alone to walking with backward counting than the dual-task related decrease in stride velocity (p = 0.013).
Table 1 Mean values and standard deviations of gait and backward counting parameters under single and dual-task condition among healthy young adults (n = 49)
Single task Dual-task P-value*
Stride velocity
Mean value (cm/sec) 129.7 ± 13.5 122.9 ± 16.0 <0.001
CV (%) 4.3 ± 2.0 4.7 ± 1.1 0.097
Stride time
Mean value (ms) 1066.9 ± 81.7 1129.5 ± 138.1 <0.001
CV (%) 1.8 ± 0.8 2.1 ± 1.1 0.015
Stride length
Mean value (cm) 137.4 ± 11.6 136.9 ± 11.7 0.414
CV (%) 3.9 ± 1.0 4.1 ± 1.0 0.275
Number of enumerated figures 18.9 ± 5.1 16.1 ± 3.8 <0.001
±: Standard deviation
CV: Coefficient of Variation = ([standard deviation/mean] × 100)
*: Compared to single task and based on paired samples t-test and use of normalized value by taking the logarithmic transformation
Sec: Second
Table 2 F test and P-value of ANCOVA with a repeated measures (n = 98) design comparing mean value of stride time while walking at self-selected speed with and without backward counting, adjusted for walking speed (covariate) and subject effect (n = 49).
Source of variation Sum of square df|| Mean square F P-value
Backward counting* 0.001 1 0.0008 1.50 0.227 †
Subjects ‡ 0.431 48 0.0090 17.71 0.000
Log (Stride velocity) § 0.124 1 0.1242 244.80 0.000
Residual 0.024 47 0.0005
Total 1.008 97 0.0104
*: Backward counting coded as a binary variable (0 = walking alone, 1 = walking with backward counting), †: Box conservative estimate, ‡: Variability between subjects §: Normalized by taking the logarithmic transformation, ||: Degree of freedom
Table 3 F test and P-value of ANCOVA with a repeated measures (n = 98) design comparing coefficient of variation of stride time while walking at self-selected speed with and without backward counting, adjusted for walking speed (covariate) and subject effect (n = 49).
Source of variation Sum of square df|| Mean square F P-value
Backward counting* 0.060 1 0.0601 0.97 0.330 †
Subjects‡ 12.751 48 0.2656 4.28 0.000
Log (Stride velocity) § 0.464 1 0.4636 7.47 0.009
Residual 2.915 47 0.0620
Total 17.201 97 0.1773
*: Backward counting coded as a binary variable (0 = walking alone, 1 = walking with backward counting), †: Box conservative estimate, ‡: Variability between subjects §: Normalized by taking the logarithmic transformation, ||: Degree of freedom
Figure 1 Change* in mean value of stride velocity and enumerated figures from single to dual-task condition among healthy young adults (n = 49). Error bars reflect the standard deviation. *: Calculated from the normalized difference between walking alone and walking with counting backward, i.e.
Discussion
Our results show that, among a sample of healthy young University students, backward counting while walking provoked significant changes in gait and counting performance, with a greater dual-task effect on backward counting than on gait. The decrease in mean value of stride velocity under dual-task was solely related to the increase of mean value of stride time. Mean value and CV of stride length did not change during walking with simultaneous backward counting. Increased stride time variability in dual-task condition was explained by slower stride velocity and subjects' effect, but was not directly attributable to dual tasking. Further, the number of enumerated figures while walking decreased significantly.
Changes in gait patterns due to the simultaneous performance of a walking-associated task have been reported previously among healthy young adults and interpreted as interference related to competing demands for attention resources involved in both tasks [1]. Both dual-task-related performance changes in gait and backward counting found in our study support this statement. However, unlike previous results obtained in older adults that showed major dual-task related gait changes [1,11], only minor changes in gait parameters were found in our sample of healthy young adults. Several interpretations of these results are possible.
The explanation of dual-task interference is usually based on the assumption that attention resources are limited [20]. According to this theoretical approach, dual-task interference will only occur if the available central resource capacity is exceeded, provoking a performance decrease in one or both tasks. Therefore, interference suggests an overload of the central resources associated with an inability to appropriately adapt allocation of attention between two simultaneously performed tasks. The manner in which attention is divided between two tasks in dual-task paradigm mainly depends on both the priority given (or not) to one task and the attentional load of each task [1,20-22]. In our study, subjects were asked to combine both walking and backward counting without prioritizing either one of the tasks, creating a condition in which attention is divided. Both tasks used in our dual-task paradigm are relatively easy and do not require major attention. Backward counting out loud from 50 is a simple mental arithmetic task requiring low attention involvement in healthy young University students. Therefore, the total attentional load mobilized to simultaneously perform both tasks could not overload the available central resources, and thus only provoked little interference with minor gait changes. All significant dual-task related gait parameter changes in our study were relatively small. Gait speed decreased from 130 cm·s-1 to 123 cm·s-1 and the CV of stride time increased from 1.8 to 2.1%. In addition, although decrease of stride velocity while backward counting was related to increase of stride time, change in stride-to-stride variability for stride time was not associated with the attentional component of backward counting.
Previous studies have shown that dual-task related gait changes also depend on the type of measured stride parameters [1,7,8]. A change in single support time while performing a walking-associated attention-demanding task has been shown in healthy young adults [2], suggesting that young adults devote attentional resources to balance control during single-limb support. Few studies have explored the effect of a walking-associated task on the rhythmic stepping mechanism in young adults [4,6,7]. Our findings showed no significant effects of backward counting on means and CV of stride length. Furthermore, dual-task related changes in stride time could be explained by a decrease in stride velocity and variability between subjects, but apparently not on attentional components related to backward counting. Such results suggest that, in contrast to stride velocity, the control of the rhythmic stepping mechanism requires only minimal attention. Only two studies using a motor task as attention-demanding task while walking have shown significant modifications in stride time variability of young adults. Grabiner et al [7] found an increase in stride time variability while simultaneously carrying an 8-ounce cup placed in a saucer while walking. Ebersbach et al. [4] reported a significant decrease in stride time when walking with a rhythmic finger tapping task, interpreted as a magnet effect, a term used to describe the tendency of biological oscillators to attract each other. However, both studies did not examine the role of walking speed as a potential confounder in the relationship between stride time variability and the involvement of attention in gait rhythmicity control.
Most studies exploring dual-task related gait changes have focused on mean values of stride parameters [1], whereas stride-to-stride variability is considered as a sensitive marker for gait control [9,23,24]. Among the temporal gait parameters, stride time reflects the walking rhythm, and is therefore taken as an index of the rhythmic stepping mechanism control [9]. In older people, there is increasing evidence that stride time variability may be related to executive function. Recently, Hausdorff et al. [25] showed an association between high CV of stride time and a relative decline in executive function among healthy older adults, and Sheridan et al. [26] reported a similar relationship between high CV of stride time and impaired executive function in demented older adults. Furthermore, Beauchet et al. [11] recently reported a specific increase of CV of stride time in a group of older adults with a range of cognitive function abilities while backward counting, but not with a verbal fluency task. Whereas verbal fluency mainly relies on semantic memory [27], counting backward essentially depends on the working memory [28] and is therefore more directly related to executive functions. Thus, the dual-task-related increase in CV of stride time while counting backward could be related to competitive interaction with executive function.
The findings of the present study demonstrate that the dual-task related increase in mean value and CV of stride time was apparently related to stride velocity and subjects' effect, but not independently to attentional interference. Although the effect of stride velocity on variability is complex [29], similar positive correlations between increase in stride time variability and decrease of stride velocity have been reported previously [13-16]. Thus, it seems that in young adults the control of gait rhythmicity, stride velocity variability, stride length variability and likely stride time variability, is an automated process that demands little or no attention.
Interestingly, the decrease in the stride velocity during dual tasking was related to an increase in stride time but not to changes in stride length. This result confirms previous findings, which suggested that stride length is not affected by dual tasking, despite changes in gait speed and the performance of attention-demanding tasks [3,5,6,17]. Our subjects decreased stride velocity only by increasing their stride time, without modifying their stride length. This increase in stride time has been related to an increase in the double-support phase [1,2], which may serve to reduce attentional demands during the swing phase and lower the risk of a loss of balance under dual-task. Therefore, the change in the gait pattern during dual task might represent a strategy aimed at maintaining an optimal index of movement consistency in term of energy costs, attentional demand, and efficiency of gait control. The isolated increase in stride time under dual tasking may be explained by two interpretations. First, stride length and stride time could depend on different cerebral control areas. Second, stride time could be more sensitive to interference than stride length.
In our sample of young University students dual tasking had a greater effect on the performance of backward counting than it did on gait velocity. This result could be interpreted as an implicit strategy of the participants, in this specific dual-task situation, to rather give priority to gait safety than to arithmetic task performance. A similar strategy has been showed in older adults [30].
A possible methodological limitation of the present study might be related to the number of strides required to obtain a representative and suitable measure of stride-to-stride variability. Analyzing steady-state walking over 15 meters, the number of steps collected in our study was around 20, whereas Owings et al. claimed that accurate estimation of step kinematics variability required at least 400 steps [31]. Another question that calls for future study is how other, more difficult "dual-tasks" might affect the variability of gait in healthy young adults.
In conclusion, performance changes in gait and backward counting when both tasks are performed simultaneously confirm that walking is an attention-demanding task in young adults. Backward counting caused a small, but significant decrease in stride velocity. However, this dual-task did not affect stride length variability and the small change in stride time variability was apparently related to the change in mean stride velocity Apparently, young adults do not allocate much attention to the control of the rhythmic stepping mechanism of walking.
Conflict of interest statement
The author(s) declare that they have no competing interests.
Contributors
O Beauchet was the main investigator of the study, designed the study, participated in data analysis, and wrote the manuscript. V Dubost was responsible for data collection and participated in preparation and analyses of data, and writing of the manuscript. FR. Herrmann participated in the development of statistical analysis, analysis, and writing of the manuscript. RW Kressig participated in the development of statistical analysis, data analysis, and writing of the manuscript.
Acknowledgements
We are grateful to the participants for their cooperation. We also thank the Saint-Etienne University Hospitals for financial support.
==== Refs
Woollacott M Shumway-Cook A Attention and the control of posture and gait: a review of an emerging area of research Gait Posture 2002 16 1 14 12127181 10.1016/S0966-6362(01)00156-4
Gage WH Sleik RJ Polych MA McKenzie NC Brown LA The allocation of attention during locomotion is altered by anxiety Exp Brain Res 2003 150 385 394 12707746
Lajoie Y Teasdale N Bard C Fleury M Attentional demands for static and dynamic equilibrium Exp Brain Res 1993 97 139 144 8131825 10.1007/BF00228824
Ebersbach G Dimitrijevic MR Poewe W Influence of concurrent tasks on gait: a dual-task approach Percept Mot Skills 1995 81 107 113 8532444
Lajoie Y Barbeau H Hamelin M Attentional requirements of walking in spinal cord injured patients compared to normal subjects Spinal Cord 1999 37 245 250 10338343 10.1038/sj.sc.3100810
Haggard P Cockburn J Cock J Fordham C Wade D Interference between gait and cognitive tasks in a rehabilitating neurological population J Neurol Neurosurg Psychiatry 2000 69 479 486 10990508 10.1136/jnnp.69.4.479
Grabiner PC Biswas ST Grabiner MD Age-related changes in spatial and temporal gait variables Arch Phys Med Rehabil 2001 82 31 35 11239283 10.1053/apmr.2001.18219
Gabell A Nayak US The effect of age on variability in gait J Gerontol 1986 39 662 666 6491179
Newell KM Corcos DM Newell KM, Corcos DM Variability and motor control Issues in variability and motor control 1993 Champaign: Human Kinetics 1 12
Nutt JG Marsden CD Thompson PD Human walking and higher-level gait disorders, particularly in the elderly Neurology 1993 43 268 279 8437689
Beauchet O Dubost V Aminian K Gonthier R Kressig RW Dual-task Related Gait Changes in the elderly: Does The Type Of Cognitive Task matter? J Mot Behav 2005 37 259 264 15967751
Beauchet O Najafi B Dubost V Aminian K Mourey F Kressig RW Age-related decline of gait control under a dual-task condition J Am Geriatr Soc 2003 51 1187 1188 12890096 10.1046/j.1532-5415.2003.51385.x
Brach JS Berthold R Craik R Van Swearingen JM Newman AB Gait variability in community-dwelling older adults J Am Geriatr Soc 2001 49 1646 1650 11843998
Danion F Varraine E Bonnard M Pailhous J Stride variability in human gait: the effect of stride frequency and stride length Gait Posture 2003 18 69 77 12855302 10.1016/S0966-6362(03)00030-4
Heiderscheit BC Movement variability as a clinical measure for locomotion J Applied Biomech 2000 16 419 427
Van Emmerik RE Wagenar RC Winogrodzka A Wolters EC Identification of axial rigidity during locomotion in Parkinson disease Arch Phys Med Rehabil 1999 80 186 191 10025495 10.1016/S0003-9993(99)90119-3
Sekiya N Nagasaki H Ito H Furuna T Optimal walking in terms of variability in step length J Orthop Sports Phys Ther 1997 26 266 272 9353690
Aminian K Najafi B Bula C Leyvraz PF Robert P Spatio-temporal parameters of gait measured by an ambulatory system using miniature gyroscopes J Biomech 2002 35 689 699 11955509 10.1016/S0021-9290(02)00008-8
Aminian K Robert P Buchser EE Rutschmann B Hayoz D Depairon M Physical activity monitoring based on accelerometry: validation and comparison with video observation Med Biol Eng Comput 1999 37 304 308 10505379
Abernethy B Dual-task methodology and motor skills research: some applications and methodological constraints J Hum Mov Study 1988 14 101 132
Bloem BR Steijns JAG Smits-Engelsman BC An update on falls Curr Opin Neurol 2003 16 15 26 12544853 10.1097/00019052-200302000-00003
Bloem BR Valkenburg VV Slabbekoorn M Willemsen MD The Multiple Tasks Test: development and normal strategies Gait Posture 2001 14 191 202 11600322 10.1016/S0966-6362(01)00141-2
Hausdorff JM Cudkowicz ME Firtion R Wei JY Goldberger AL Gait variability and basal ganglia disorders: stride-to-stride variations of gait cycle timing in Parkinson's disease and Huntington's disease Mov Disord 1999 13 428 437 9613733 10.1002/mds.870130310
Hausdorff JM Zemany L Peng C Goldberger AL Maturation of gait dynamics: stride-to-stride variability and its temporal organization in children J Appl Physiol 1999 86 1040 1047 10066721
Hausdorff JM Yogev G Springer S Simon ES Giladi N Walking is more like catching than tapping: gait in the elderly as a complex cognitive task Exp Brain Res 2005
Sheridan PL Solomont J Kowall N Hausdorff JM Influence of executive function on locomotor function: divided attention increases gait variability in Alzheimer's disease J Am Geriatr Soc 2003 51 1633 1637 14687395 10.1046/j.1532-5415.2003.51516.x
Hausdorff JM Stride variability beyond length and frequency Gait Posture 2004 20 304 15531178 10.1016/j.gaitpost.2003.08.002
Hittmair-Delazer M Semenza C Denses G Concepts and facts in calculation Brain 1994 117 715 728 7922459
Smith EE Geva Jonides J Miller A Reuter-Lorenz P Koeppe RA The neural basis of task-switching in working memory: effects of performance and aging Proc Natl Acad Sci 2001 98 2095 2100 11172081 10.1073/pnas.98.4.2095
Li KZ Lindenberger U Freund AM Baltes PB Walking while memorizing: age-related differences in compensatory behavior Psychol Sci 2001 12 230 237 11437306 10.1111/1467-9280.00341
Owings TM Grabiner MD Measuring step kinematic variability on an instrumented treadmill: how many steps are enough? J Biomech 2003 36 1215 1218 12831749 10.1016/S0021-9290(03)00108-8
|
16095533
|
PMC1190208
|
CC BY
|
2021-01-04 16:37:40
|
no
|
J Neuroengineering Rehabil. 2005 Aug 11; 2:26
|
utf-8
|
J Neuroeng Rehabil
| 2,005 |
10.1186/1743-0003-2-26
|
oa_comm
|
==== Front
Malar JMalaria Journal1475-2875BioMed Central London 1475-2875-4-271597514610.1186/1475-2875-4-27ResearchClustered local transmission and asymptomatic Plasmodium falciparum and Plasmodium vivax malaria infections in a recently emerged, hypoendemic Peruvian Amazon community Branch OraLee [email protected] W Martin [email protected] Dionicia V [email protected] Jean N [email protected] Freddy F [email protected] Norma [email protected] Eugenia [email protected] Enrique J [email protected] Eduardo [email protected] Department of Medicine, Geographic Medicine, University of Alabama at Birmingham, Bevill Research Building BBRB-556, Birmingham, Alabama, 35294-2170, USA2 Direccion de Salud-Loreto, Ministerio de Salud (MINSA), Iquitos, Peru3 Instituto de Medicina Tropical "Alexander Von Humboldt", Universidad Peruana Cayetano Heredia, A.P. 4314 Lima 100, Lima, Peru2005 23 6 2005 4 27 27 13 3 2005 23 6 2005 Copyright © 2005 Branch et al; licensee BioMed Central Ltd.2005Branch et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
There is a low incidence of malaria in Iquitos, Peru, suburbs detected by passive case-detection. This low incidence might be attributable to infections clustered in some households/regions and/or undetected asymptomatic infections.
Methods
Passive case-detection (PCD) during the malaria season (February-July) and an active case-detection (ACD) community-wide survey (March) surveyed 1,907 persons. Each month, April-July, 100-metre at-risk zones were defined by location of Plasmodium falciparum infections in the previous month. Longitudinal ACD and PCD (ACP+PCD) occurred within at-risk zones, where 137 houses (573 persons) were randomly selected as sentinels, each with one month of weekly active sampling. Entomological captures were conducted in the sentinel houses.
Results
The PCD incidence was 0.03 P. falciparum and 0.22 Plasmodium vivax infections/person/malaria-season. However, the ACD+PCD prevalence was 0.13 and 0.39, respectively. One explanation for this 4.33 and 1.77-fold increase, respectively, was infection clustering within at-risk zones and contiguous households. Clustering makes PCD, generalized to the entire population, artificially low. Another attributable-factor was that only 41% and 24% of the P. falciparum and P. vivax infections were associated with fever and 80% of the asymptomatic infections had low-density or absent parasitaemias the following week. After accounting for asymptomatic infections, a 2.6-fold increase in ACD+PCD versus PCD was attributable to clustered transmission in at-risk zones.
Conclusion
Even in low transmission, there are frequent highly-clustered asymptomatic infections, making PCD an inadequate measure of incidence. These findings support a strategy of concentrating ACD and insecticide campaigns in houses adjacent to houses were malaria was detected one month prior.
==== Body
Introduction
In Peru, the history and prevalence of the human-malaria causing Plasmodium species parasites are different from what is found in the widely studied African, Asian or Pacific countries. The incidence of malaria in Peru before 1940 is not well-known, but between 1940 and 1960 there were reports of Plasmodium vivax and Plasmodium malariae infections in northern and central Peru. In the 1960's malaria transmission was very low due to an impressive eradication attempt. However, in the 1980's the eradication campaign was abandoned and house spraying programs were increasingly neglected. Plasmodium falciparum was first reported in Peru's Department of Loreto in 1988 [1]. Then, between 1991 and 1995, the annual infection prevalence dramatically increased in the coastal regions near Piura [2] (Figure 1A). In the region surrounding the capital city of Iquitos, there was a P. vivax and P. falciparum epidemic between 1995 and 1998 (Figure 1B). This is attributable to the abandonment of DDT campaigns leading to the increased geographic range or increased local abundance of the mosquito vector Anopheles darlingi [2,3].
Figure 1 A-B: History of malaria in Peru. Before 1960, there was P. vivax, P. malariae and a very limited number of P. falciparum cases. After DDT campaigns stopped in the 1960's, malaria went from a low annual parasitaemia incidence (API = 1000*slide positive/population), limited to the northern costal regions, to a high level in 1995 (A: directly from Roberts et al[1], with permission). In 1991 P. vivax reemerged and in 1994 P. falciparum emerged in Loreto. An epidemic ensued, focused near the capital city of Loreto, Iquitos, and a hypoendemic continues (B).
Since 1998, in Loreto, malaria has continued to be detected in low incidence, with the P. falciparum and P. vivax annual infection incidence varying between 5 and 50 cases/1000 persons (Figure 1B). Malaria infections are recorded by the Peruvian Ministry of Health, Ministerio de Salud Dirección de Salud Loreto (MINSA). The health system is highly centralized, with little access to private physicians even within the city of Iquitos [4]. The primary approach to malaria control is passive case detection in local MINSA health posts or centers/hospitals and administration of effective anti-malaria drug treatment at no cost to the patient [4]. This control method is in line with Pan American Health Organization (PAHO) program guidelines. Based upon this detection and treatment method, the malaria incidence is exceedingly low, but has been sustained since the emergence of P. falciparum and P. vivax 10 years ago. However, passive case detection might not accurately reflect the malaria transmission rate or be an effective malaria intervention strategy if there are local transmission clusters and asymptomatic infections are going undetected.
One potential explanation for the sustained low incidence of malaria is a clustering of infections into a small subgroup of the population. If this occurred, then one would expect that the actual infection frequency in some persons would be higher than that estimated by summarizing over the entire population.
Of particular interest is whether the malaria transmission is occurring within the suburban villages just outside of Iquitos, where MINSA regularly applies pyrethoid insecticides to houses [4], or whether infections are acquired through travel to more remote regions. Because passive case detection is summarized at the community level, by number of infections detected per number of individuals living within the community, the more local malaria transmission dynamics needed to target insecticide campaigns remain unresolved. Insecticide spraying must be better targeted, as the insecticides are expensive, short-lived and are rapidly washed away from the resin-rich houses during the frequent rains in this region [4]. If the transmission was occurring in remote regions, or if transmission was clustered within certain locations within the suburban villages, then insecticide campaigns could be targeted to these infection-prone regions.
Also, the occurrence of local transmission impacts the expectation of having asymptomatic infections. If the majority of infections occur within a sub-population who frequently travel or live in higher transmission regions, then these highly exposed individuals might receive the quantity of infections conventionally considered necessary develop protective immunity against symptoms during infection [5-17]. In this case, these individuals would likely have infections undetected by passive case detection. Conversely, if infections are acquired locally, then there would be low transmission in the whole population and one would expect most of the infections to be symptomatic. This expectation is based upon literature regarding the development of naturally acquired immunity to P. falciparum malaria requiring frequent infections over a period of 5 years and sustained frequent infections in order to maintain an ability to resist symptoms during malaria infection [5-17].
Having to conduct active case detection in order to detect asymptomatic infections would not be expected in a region with recent and low transmission. Data suggests that, early in the epidemic, symptoms were likely, although deaths were rare. From the MINSA records, Ambarru et al. (1998) calculated the malaria-attributable death rate in 1997 to be only 0.1% [1]. This low mortality during the epidemic of 1995–1998 might be partly attributable to under-reporting of mortality. Another possible explanation of low morbidity and mortality with P. falciparum infections is good health care. It is possible that people experience symptoms quickly and go to the health centers immediately. If this is the case, then passive case detection with administration of effective malaria treatment drugs at no cost to the patients is an effective intervention strategy which can explain both the low mortality and the failure of malaria transmission to increase in this region.
The frequency of asymptomatic malaria infection is not known in this or most low transmission regions. There has been a longitudinal study in Brazil which demonstrated asymptomatic P. falciparum and P. vivax malaria infections [18]. In Peru, studies have been limited by cross-sectional designs. At the end of the epidemic (1998) in one community just north of Iquitos, there was a comprehensive cross-sectional active and passive case detection study [19]. Roper et. al (2000) found that 90% of both P. falciparum infections and P. vivax infections detected were associated with symptoms and the parasite densities were similar to that expected in non-immune individuals [19]. However, in 1999, Roshanravan et al conducted a cross-sectional survey of 998 individuals living in communities south of Iquitos [20]. They found 13 individuals with P. falciparum and 30 with P. vivax by microscopy. Of these, only 8 and 19, respectively, reported a fever. The conclusion was that there were many low density malaria infections that were asymptomatic. However, the question remains whether these were early stage infections that would have subsequently produced symptoms and then have been detected at the health post.
Understanding the continued low malaria transmission and the potential for increased transmission in the Amazon Jungle region of Peru is contingent upon determining the prevalence of a capable malaria vector, clustering of infections within certain households and the presence of asymptomatic malaria. This present study, in a suburban community near Iquitos, determined the level of local malaria transmission, the frequency of individuals presenting with symptoms and the dynamics of P. falciparum and P. vivax infections during weekly active surveillance. The results suggest that, even in low transmission, active case detection provides critical insight into P. falciparum and P. vivax transmission and infection and that active case detection can be spatially and temporally targeted for a feasible malaria intervention strategy.
Methods
Study area
South of Iquitos, the San Juan District, is a focus of P. falciparum and P. vivax malaria transmission. The rural communities in San Juan have P. falciparum and P. vivax infections, mostly detected in the rainy season, which lasts from January to July. Typically, MINSA personnel detects, treats and documents malaria cases by passive-case detection. All persons presenting with a fever at the community-based MINSA health posts or a larger health clinic located in the urban region of San Juan are tested for malaria by microscopy. Individuals go to the MINSA health centers as there is no access to private doctors and anti-malaria treatment is given by MINSA at no cost to the patient. The anti-malaria drug supply is tightly controlled and given as observed therapy (see treatment schedule, below) each day to those infected.
The community in which this study was conducted, Zungarococha, was selected based upon the prevalence of P. falciparum malaria infections, the acceptance by the community and the fact that the community is composed of four villages each separated by approximately two kilometres and yet serviced by the same MINSA health post. Zungarococha (population N = 1907) is composed of 4 villages: Zungarococha town (ZG), Puerto Almendra (PA), Ninarumi (NR) and Llanchama (LL) (Figure 2). The road to LL was not consistently passable by a four-wheel drive truck, so it was not included in the active house-by-house follow-up. The environment and income level in these communities is similar. The village "ZG" is more developed and is the location of the MINSA health post. There is a bus that offers daily transportation between villages. All of the villages are primarily sustained by agriculture and fishing. Generally, women work in or near their homes. Men often work in agriculture or are construction/maintenance workers at the agricultural university or on commercial farms within the community. Fishing is also conducted on the rivers which border the villages. In each village, the houses are typically made of wood with resin-rich thatch roofs, although approximately 10% of the homes are cement-block. There are no screens on houses, and mosquitoes have easy entry. There is electricity in some homes and streets of ZG and PA. The homes are on average 8 × 12 metres in size and are often side-by-side. In most cases the houses are close together in lines that follow dirt streets. More than 70% of the houses are less than 6 metres apart. The homes are often located near the marshy areas of rivers or ponds. Generally, the houses face dirt streets while the back of the houses are open to more foliage-dense areas within 100 metres of rivers or ponds. Much of the cooking and many other activities occur in these "back-yard" riparian habitats. The exception to this village design is NR. In NR, half of the village is traditional with many homes close together, while the other half is more dispersed.
Figure 2 Schema of community in the passive and active surveillance study. Zungarococha is a suburban community approximately 5 kilometres (km) from Iquitos, Peru. Zungarococha is composed of four villages: Zungarococha town (ZG, N = 1293), Puerto Almendra (PA, N = 207), Ninarumi (NR, N = 472), and Llanchama (LL, N = 142). Passive case detection occurred in the MINSA health post. Active case detection first included a community survey in March, 2003. Then, from the P. falciparum infections detected in passive case detection or the community survey, at-risk regions were defined for each month April-July, 2003, based upon the location of the month prior's P. falciparum infections. The location of these P. falciparum infections were used to determine a 100-metre radius at-risk zone for the following month. In each month of April-July 2003, 27–39 sentinel houses (approximately 150 participants/month) were selected within each at-risk zone for one month of weekly prospective visits.
Entomology
Members of the community have been informed that An. darlingi mosquito is highly prevalent in their community and causes malaria transmission. Public health education is conducted both by local MINSA health promoters and radio announcements. Additionally, throughout the daily interactions with the community and village authorities, the MIGIA investigator team discussed how to decrease the possibility of An. darlingi biting. All of the households included in the survey answered a questionnaire which included questions on bed net and insecticide usage. More than 95% of the households reported using bed nets, although none of the households reported using insecticide treated bed nets or repellants in their homes. Approximately twice per year, the MINSA entomologists conduct fumigation campaigns and apply pyrethoid insecticides to all houses [4].
Entomology capture studies were conducted to determine the transmission potential within the Zungarococha villages. Mosquitoes were captured by mechanical aspiration from 7–12 pm, the known time of peak biting for the anthropophilic An. darlingi mosquito malaria vector in the Peruvian Amazon [4]. An. darlingi lives near and bites within human households (endophagic behavior). An. darlingi also bites near its breeding sites, which are typically shady edges of quiet rivers (riparian habitats) [21]. Each month, April-July, 2003, mosquitoes were collected on five porches each week for one month one the porches of sentinel houses (different houses each month). These houses were sentinel houses selected within the defined at-risk zones for a given month (see below), with five houses per at-risk zone enrolled each month for weekly entomological captures. Each entomology team (composed of two members) placed the captured mosquitoes in containers labeled by location and hour of collection. The mosquitoes were identified to the species level by expert MINSA entomologists by morphology while using a dissecting microscope. An. darlingi have an easily distinguishable one white front leg.
Passive case detection
This study had a full-time clinical team stationed at the Zungarococha Health Post for passive case detection throughout the year. The MINSA records of malaria detection within the Zungarococha Health Post were reviewed and complied in order to calculate the P. vivax and P. falciparum incidence. The number of malaria cases per given time were divided by the total population (N = 1,907) to determine the incidence by passive case detection. The records were matched by name to the individuals observed in active case detection (see below).
Active case detection
Active case detection was conducted during the months of March and July, 2003 (the malaria transmission season).
The active case detection began with a cross-sectional community-wide survey, wherein all family members were invited for malaria parasite testing and a health survey. Between March 31 and April 5, 2003, there were village-wide surveys of Zungarococha. For simplicity within this manuscript, the community survey is stated to have occurred in March, 2003.
From May until July, 2003, longitudinal active case detection included a monthly selection of houses (selected at random, or as close to random as possible for field conditions, see below) located within 100 metres of an "index" house. Houses were called "index" houses if they had at least one person detected with a P. falciparum infection in the previous month, detected through the community survey results, the health post detection, or the prior month's active case detection. An approximate 100 metre radius circle around the index houses (or a median point between index houses when there were two index houses spaced by less than 50 metres) was defined as an at-risk zone for P. falciparum transmission. If at-risk zones overlapped in a given month, houses were assigned to one of the zones. Each month, all houses within each 100 metre at-risk zone were identified by assigning house numbers (addresses) on a map of each village and entered as list of possible sentinel houses. Of all possible sentinel houses (generally, 20–50 potential houses existed within a given 100 metre at-risk zone), 10 potential sentinel houses from each at-risk zone were randomly selected. In June and July, this selection was done by a random number generator (conducted using Statistical Analysis Software Version 8, Cary, NC). The enrolment team for this study would be given this list and then would go to these potential sentinel houses one at a time to recruit all household members for one month of weekly active sampling until at least 30 individuals (living in 5–9 houses) were enrolled within each at-risk zone. The index house were included in this longitudinal active case detection, however, the results from active case detection are not included in this analysis due to these houses not being selected at random.
There were 5–6 at-risk zones during each month of active surveillance, with approximately 150 sentinel individuals enrolled each month, April-July, 2003. Each household member was visited 4 times (once per week for one month), unless there was a positive blood slide that resulted in extra visits. The recruitment and retention was high: 82% of the time all permanent members of each sentinel house consented and enrolled and 86% of consenting individuals completed participation in all 4 weekly visits.
For each month of active weekly surveillance, due to the relatively small size of each village (less than 600 square-metres) and the location of index cases, approximately the same zones were defined in successive months of active surveillance. Throughout the study, 70–90% of the houses within PA and NR, and 50–70% of the houses within ZG were included within the zones of our active surveillance.
Epidemiological questionnaire, physical and blood samples
At each visit, whether in the health post, the community visits or the weekly home-visits, a detailed epidemiological questionnaire was administered and a physical examination by a physician was conducted. Axillary temperature was measured by a digital thermometre. 0.25–3 ml of blood, by fingerprick or venipuncture, was collected. Individuals diagnosed with malaria had 3–6 ml of blood collected by venipuncture. Thin and thick blood smears and capillary haematocrit tubes were prepared. The remaining blood was stored in tubes with EDTA, separated by centrifugation, and the sera and packed blood cells were frozen within 18 hours. Haematocrit was measured as packed cell volume (PCV) after centrifugation.
Microscopy
There were two microscopists in this study, each having more than 15 years experience. Their ability to for accurate reading of blood slides is well known, and they are the designated quality control monitors for all MINSA malaria microscopy in the Department of Loreto. Blood smears were stained with Giemsa using standard procedure. Using 100× magnification to read the thick smear, all malaria species' trophozoites and gametocytes were counted separately. Microscopy fields were read to count at least 500 white blood cells (WBCs) before diagnosing an individual as negative by microscopy. The parasite density for each species (parasites/μl blood) was determined by number of parasite species (trophozoites and gametocytes counted separately) multiplied by 6,000 and then divided by the total number of WBCs counted. Using 6,000 RBCs per one WBC to determine parasite density was established by the MINSA microscopists after years of counting RBCs and WBCs in malaria infected patients from this region and their continued intermittent checks verify that this remains an appropriate average conversion factor.
Treatment
In passive case detection, where individuals presented at the Zungarococha Health Post, treatment was administered immediately upon diagnosis within the Health Post. In active case detection, household visits occurring in the afternoon, the blood slides were transported to the microscopists' laboratory for reading either on the same day or the following morning. Blood slides from individuals reporting febrile illness within the previous 2 days, having a detected fever ≥ 38.3°C, or having a haematocrit PCV <30% were flagged for immediate reading by microscopists. Blood slides from individuals not reporting the symptoms listed above were read within 6 days after the actual visit and collection of the blood slide. All individuals were revisited on "day i+7." If individuals found positive from the earlier blood slide were now symptomatic at this "day i+7" visit, they were given antimalaria drug treatment. If they remained asymptomatic, the blood smear was flagged for immediate reading. These participants were then re-visited in one day ("day i+8") for a confirmation slide and treatment of all individuals with a positive blood slide on "day i" or "day i+7". Even without symptoms, all cases were treated within 7 days of confirmation of malaria parasites. If a positive slide was detected at the last of the 4 weekly visits, then an extra visit, including an extra blood sample, was made. The physician-nurse team was in the vicinity throughout the study to take a blood smear from anyone presenting with malaria symptoms, even if they were not in the regularly scheduled weekly visits. Again, symptomatic individuals had their blood slides read within one day.
Whether in active or passive case detection, all treatments were given through the MINSA authorities, following the MINSA National Drug Policy Guidelines. P. vivax treatment is chloroquine (10 mg/kg for 3 days) with primaquine (0.5 mg/kg for 7 days). P. falciparum treatment is mefloquine (12.5 mg/kg daily for 2 days) with artesunate (4 mg/kg daily for 3 days) in non-pregnant patients older than 1 year of age. In pregnant women and infants, P. vivax is treated with chloroquine (10 mg/kg daily for 2 days and 5 mg/kg on third day) and P. falciparum is treated with clindomyacin (10 mg/kg 2 times daily for 5 days) and quinine (10 mg/kg 3 times daily for 7 days). The treatments are specific to Plasmodium species, as P. falciparum parasites from this region are highly resistant to chloroquine[22]. The treatments are effective at eliminating all stages of the Plasmodium species parasites to which they are directed. In all cases the treatments are observed, and a fingerprick blood sample is collected at 7 days and then again at 14 days later to confirm treatment success.
Data analysis
Data was entered using a programme in ACCESS developed specifically for this cohort project. The principle investigator developed a program that displays potential matching records by similarity (keyed to name, address, and age). The principle investigator and data management specialist determined when records from the health post matched individuals in our active follow-up. After the matching, all names and unnecessary information was removed from the analysis data set. Relational database management and statistical analysis was conducted using the Statistical Analysis Software Version 8 (SAS, Cary, NC). The Fisher's Exact Test (Fisher's exact) was used to compare the expected versus observed frequency of detecting infections in houses and to compare the frequency of febrile illness associated with P. falciparum versus P. vivax infections. Parasite densities were logarithmically transformed (base 10) in order to normalize the densities and calculate the geometric mean and 95% confidence interval (95%CI). A general linear model logistic regression (GLM logistic) was used to test the correlation between probability of febrile illness and parasite density, while controlling for age.
Human subjects ethical approval
All protocols were reviewed and approved by the University of Alabama at Birmingham, Universidad Peruana Cayetano Heredia and the Peruvian Ministerio de Salud (MINSA) Institutional Review Boards for the study of human subjects. Written, informed consent was obtained from all participants. In the case of minors less than seven years old, the parents or guardians gave consent. In the case of minors between seven and eighteen, both assent from the minor and consent from the parents or guardians was obtained prior to enrollment.
Results
Entomology
Anopheles darlingi is an anthropophilic mosquito that bites in human houses as well as in the riparian habitats[21] between the backs of houses and the quiet rivers which are within each of the Zungarococha villages. Between April, 2003 and July, 2003, 12,035 mosquitoes were captured, of which 87% were An. darlingi (Figure 3). These human-bait captures (by mechanical aspiration) were on the porches of houses within our at-risk zones for Plasmodium falciparum transmission, so defined by having a P. falciparum infection centrally located within the 100-metre radius circle one month prior. In total, there were 80 houses, each sampled 4 times (weekly), giving 160 entomological captures between the months of April and July, 2003. In April, the month with the highest An. darlingi biting rate, the mean household biting rate was 10–24 An. darlingi bites/person/hour in each village. The biting rate rapidly decreased in the months of May-July, 2003, leading to the dry months of August-December (data not shown). This data supports the previous MINSA reports that there is seasonal P. falciparum and Plasmodium vivax transmission by An. darlingi[4] and demonstrates high local transmission potential.
Figure 3 Mean An. darlingi captured, prepared to bite/person/hour and percent of mosquitoes that were An. darlingi. Captures were conducted by mechanical aspiration from 7–11 pm (the known peak biting time), on the porch of households4. An. darlingi are known to bite in and near human households[4]. There were 10 MINSA entomologists, working in pairs, to give a total of 5 entomologic captures/night. The households were chosen from the sentinel houses in at-risk zones selected for human infection surveillance. In at-risk zones where there were more than 5 houses, 5 of the sentinel houses were randomly selected for entomologic captures. Each month, 5 at-risk zones were included, with 5 houses each, for 4 weekly captures in each house.
Incidence of infection by passive case detection
In the health post, there were 60 P. falciparum and 539 P. vivax infections detected by passive case detection between January, 2003 and March, 2004 (one was a mixed-species infection). Only 11 P. falciparum and 121 P. vivax infections were detected during the low transmission months of January, 2003 and August-December, 2003. Considering infections detected February-July, 2003 and estimating by dividing the total infections detected by the total population size (1907), the passive-case detection incidence was 0.03 P. falciparum and 0.22 P. vivax infections/person/malaria season (Table 1). This is significantly lower than prevalence of infection detected in individuals living in at-risk zones for P. falciparum transmission.
Table 1 Comparison of P. falciparum and P. vivax passive and active case detection. The study includes individuals followed throughout the year-2003 malaria transmission season (February-July, 2003). Passive case detection was conducted at the Health Post, where individuals presenting with febrile illness are tested for malaria parasites by microscopy. Active case detection was performed in two ways. In March, a community survey was conducted where 957 of the population (N = 1907) were sampled. Then, April-July, sentinel individuals living in houses defined as at-risk zones for P. falciparum transmission were randomly selected. Individuals selected as sentinels were followed prospectively for one month with weekly visits. The passive and "passive and active" (active+passive) prevalence only considers individuals who were enrolled both in the community survey and enrolled as sentinels in at-risk zones for at least one month during these months.
P. falciparum P. vivax
Total malaria cases detected by passive case detection February - July, 2003 / total population 49/1907 = 0.03 418/1907 = 0.22
No. monthly active case intervals with weekly sampling in at-risk zones (No. of individuals) 592 (573) 592 (573)
Malaria infections detected by active+passive surveillance in at-risk zone sentinels Total: 74/573 = 0.13 Total: 224/573 = 0.39
Passive case detection, February – July 15a (20.8%) 48 (21.4%)
Community survey, March 12b (16.7%) 36b (16.1%)
Monthly active surveillance in at-risk zones, April – July 46 (62.5%) 140 (62.5%)
Difference in passive versus active+passive prevalence Attributable fraction potentially explained by sampling in at-risk zones and/or asymptomatic infections 0.13/0.03 = 4.33 0.39/0.22 = 1.77
a Of the remaining 34 P. falciparum infections detected in passive case detection, but not included in this count, at least 12 occurred in individuals living within at-zones. This is because these individuals were not selected as sentinels for the weekly active surveillance.
b Five additional P. falciparum and 36 additional P. vivax infections were detected in the community survey. However, they occurred in individuals not selected as sentinels for the weekly active surveillance.
Prevalence of infection by active case detection
There were 957 participants in the cross-sectional survey in March. Then, based upon the detection of P. falciparum infections in the community survey and Health Post, 100-metre radius circles within each village were defined, focused from a house where there was a P. falciparum infection detected the month prior. Each month, approximately 150 individuals living in 5–6 at-risk zones were selected as sentinels. There were a total of 593 instances of weekly active surveillance for one month between April and July, 2003. Only 19 individuals were enrolled for more than one month of active sampling. In total, 74 P. falciparum and 224 P. vivax infections were detected by microscopy (3 were mixed P. falciparum and P. vivax infections). Of these infections, 45 P. falciparum and 140 P. vivax infections were detected during the one month of weekly active case surveillance of randomly selected sentinel individuals living in at-risk zones.
The age and sex distribution of all 573 individuals enrolled as sentinels in the at-risk zones was investigated (Table 2). Gender was not associated with prevalence of infection. Children (0–12 years-of-age) had a lower prevalence of infection than older individuals, which was statistically significant (p < 0.0001 for both P. falciparum and P. vivax infection) (Table 2). From the background knowledge of the community, this increase in prevalence is attributable to young children more frequently being in beds, under bednets, during the peak An. darlingi biting time of 7–12 pm.
Table 2 Age and sex distribution of individuals enrolled for active surveillance in at-risk zones. Age groups were classified so as to group children and adolescents separately from adults, while maintaining at least 15 individuals in each age group for each infection status. The prevalence of infection was significantly different by age group, where the children had less P. falciparum and P. vivax infections (Fisher's Exact Test, p < 0.0001). There was no significant difference in the frequency of males in the total population versus the P. falciparum or P. vivax infected individuals as a whole or when stratifying by age group (Fisher's Exact Test, p > 0.2).
age group (years) Enrolled in at-risk zone for weekly sampling P. falciparum P. vivax active+passive prevalence
n = % Male N = % Male N = % Male P. falciparum P. vivax
0–11 214 49% 19 53% 48 52% 0.09 0.22
12–20 119 41% 19 47% 62 51% 0.16 0.52
21–104 240 46% 36 39% 114 48% 0.15 0.48
Total 573 46% 74 45% 224 50% 0.13 0.39
a Of the remaining 34 P. falciparum infections detected in passive case detection, but not included in this count, at least 12 occurred in individuals living within at-zones. This is because these individuals were not selected as sentinels for the weekly active surveillance. b Five additional P. falciparum and 36 additional P. vivax infections were detected in the community survey. However, they occurred in individuals not selected as sentinels for the weekly active surveillance.
Comparing passive and active case detection
The passive and active (active+passive) case detection prevalence was calculated from all individuals who had passive case detection in the health post, the community survey and active-weekly detection for at least one month. The active+passive prevalence was 0.13 P. falciparum and 0.39 P. vivax infections/person/malaria season (Table 1).
To compare the active+passive to the passive case detection a ratio was computed (Table 1). Because individuals in active case detection were randomly selected within each at-risk zone and had the same follow-up throughout the study, there was no obvious bias to the prevalence calculated within at-risk zones. Likewise, because there was a community survey, community-wide advisories to go to the Health Post upon experiencing malaria symptoms, and accessibility to free malaria treatment at the Health Post, there was no obvious bias to the incidence calculated from passive case detection in the whole community. The P. falciparum and P. vivax active+passive case detection was 4.33-times and 1.77-times higher, respectively, than the prevalence predicted by passive case detection in the health post (Table 1).
The greater increase in active+passive versus passive case detection prevalence for P. falciparum in comparison to P. vivax was expected. This is because the active case detection was focused in regions defined as at-risk for P. falciparum based upon the previous months' P. falciparum infections. Additionally, to explain the 4.33-times higher prevalence when including active-case detection in at-risk zones, the symptoms and dynamics of the infections was investigated. However, first, the analysis regarding the further clustering of infections within at-risk zones is presented.
Spatial and temporal clustering of infections within at-risk zones
There were P. falciparum and P. vivax infections in each village, with P. falciparum being more prevalent in NR and PA and P. vivax being more prevalent in ZG (Figure 4). The plateau of P. falciparum prevalence suggests that local transmission was decreasing after May, 2003. This is consistent with the entomologic results showing significantly low An. darlingi biting May-July, 2003.
Figure 4 Spatial and temporal clustering of P. falciparum (Pfalc) infections within at-risk zones. Pfalc detection prompted our selection of sentinel houses within at-risk zones in the following month. There was clustering of infections (outlined in black) in the same and adjacent houses. The houses followed a general linear arrangement, with consecutive numbers being contiguous houses. Generally, the house numbers wrap around in opposite order for opposing houses located across a street. Human subjects concern prohibits display of more precise house locations. All persons with active case detection are shown (except for ZG, where, for simplicity and space constraints, the many households that were negative for Pfalc are not listed). Asterisks are passive case detections in the Health Post occurring within the sentinel houses. The index houses defining at-risk zones for the following month are shown in green font. Active sampling events are shown as a ratio. The numbers of Pfalc infected individuals detected in the household are shown as the numerator. The numbers of persons in the house participating as sentinels are shown as the denominator. There were four instances of persons living in two houses: where members were sampled within one of the two houses and reported frequently eating diner and sleeping in both houses (these are shown as dashed-house codes).
A short-term clustering of infection within homes is expected if there is local entomologic transmission. There were many instances of detecting more than one P. falciparum or P. vivax infection within a given household (Figure 4). Considering the number of individuals living in each house, there was a significant clustering of infections within adjacent houses. The expected frequency of detecting a given number infections is the number of individuals living in households with the given infection history multiplied by the probability of the occurrence. For P. falciparum and P. vivax, separately, the expected frequency of having one, two, three or more infections was calculated and compared to the observed frequency. In each case, there was a significant difference between the expected and observed frequencies of having more than one infection in a given household (Fisher's Exact Test, p < 0.0001 for each species). Even more striking was the clustering when considering adjacent houses (Figure 4).
Of those surveyed over time, 3 individuals had more than one P. falciparum infection and 39 individuals had more than one P. vivax infection, separated by at least one month after malaria drug treatment, during the February-July, 2003 active+passive surveillance. Due to the malaria treatment protocol being effective at clearing all stages of the malaria parasites and the negative blood smears observed after the treatment, these following infections were termed "discrete." Detecting more than one discrete P. vivax infection within a given individual was more common. In fact, 4 persons had 3 discrete P. vivax infections. The 74 P. falciparum infections were detected in 69 individuals (4% of the individuals had more than one P. falciparum infection). The 212 P. vivax infections were detected in 173 individuals (23% of the individuals had more than one P. vivax infection). Detecting an individual with more than one infection for a given species was not different from the frequency expected based upon the active+passive prevalence (probability of having more than one infection is 0.132 for P. falciparum and 0.392 for P. vivax) (Fisher's Exact, P. falciparum p = 0.6197, P. vivax p = 0.4354).
The results show low-level local transmission even within at-risk zones, with the same and adjacent houses being at high risk in the following month (Figure 4). Even if there is long-term clustering of infections in the same households over successive seasons, the expected infection rate would still be less than 0.5 P. falciparum and less than 1.5 P. vivax infections/person/year. For example, of the 573 individuals in our at-risk zone active surveillance, there were 110 who lived in houses where more than one person within the house had a P. falciparum infection during the 2003 malaria season. The prevalence in this subpopulation within the at-risk zones is only 0.26 P. falciparum infections/person/malaria season. Therefore, even in the population at high P. falciparum infection risk, the transmission would be defined as hypoendemic based upon infection prevalence by standards based upon African transmission[5].
Travel away from homes
Participants were asked if they traveled away from their village for their occupations or other reasons during the hours of 6 pm-12 pm. Twenty-nine participants reported travel to a rural location or travel to any location by river within one month of active case detection. Of these 29 travelers, there were 7 (24%) P. vivax and 4 (14%) P. falciparum infections detected within one month of the time of travel. Therefore, this travel to rural regions with assumed higher malaria transmission (MINSA reports) did have a higher frequency of malaria infection. However, the other 58 P. falciparum and 205 P. vivax infections detected in active case detection were in 544 individuals not reporting travel to more rural regions. Therefore, the majority of infections detected in Zungarococha (94% for P. falciparum and 97% for P. vivax) could not be explained by travel away from their villages.
Description of infections detected in active surveillance
In the active case detection (community survey and weekly surveillance), 62 P. falciparum and 212 P. vivax infections were detected. The geometric mean parasite density, considering the first positive blood smear obtained in active surveillance, was 1282 P. falciparum parasites/μl blood (95%CI: 1201–1369) and 321 P. vivax parasites/μl blood (95%CI: 312–331). Of particular interest regarding the transmission potential, 53% of the P. falciparum and 22% of the P. vivax infections had gametocytes detected. The geometric mean gametocyte density was 151 P. falciparum gametocytes/μl blood (95%CI: 134–171) and 120 P. vivax gametocytes/μl blood. The higher gametocytemia frequency of P. falciparum relative to P. vivax is attributable to the difficulty in distinguishing P. vivax gametocytes from trophozoites by microscopy.
A classic indicator of clinical malaria is the presence of a fever at the time parasites are detected. The frequency of a measured fever in individuals with parasites in blood smears was 21% for P. falciparum cases and 14% for P. vivax cases, respectively (Fisher's exact: p = 0.2329). The symptoms reported in individuals with P. falciparum or P. vivax infections at any time within one month were considered. Of the 62 individuals with P. falciparum infections detected in active case detection, 41%, 3%, 6%, 6%, 5% and 6% reported fever, chills, headache, diarrhea, nausea/vomiting or body aches, respectively. Of the 212 individuals with discrete P. vivax infections detected in active case detection, 24%, 1%, 4%, 9%, 5% and 10% reported fever, chills, headache, diarrhea, nausea/vomiting or body aches, respectively. The proportion of individuals with a detected or reported fever (febrile illness) was significantly higher in P. falciparum versus P. vivax infections (Fisher's exact: p = 0.0096). There were no signs of severe malaria infections (hyperparasitaemia >100,000/μl, neurological symptoms, severe anemia, or respiratory distress) during active surveillance. One P. vivax infection was associated with jaundice.
The geometric mean parasite density of P. falciparum infections in the group with febrile illness was 4011/μl (95%CI: 811–9746). In individuals with no reported or detected fever at anytime within one month of the detected P. falciparum parasitaemia, the geometric mean parasite density was 973/μl (95%CI: 142–2078). The parasite density of P. vivax infections in the group with febrile illness was 636/μl (95%CI: 81–2477) versus 391/μl (95%CI: 173–1211) in the group without febrile illness. For P. falciparum, there was a positive correlation between probability of febrile illness and P. falciparum parasite density and (GLM logistic: p = 0.0203), independent of the negative correlation between probability of febrile illness and age (GLM logistic: p = 0.0471). P. vivax parasite density was not significantly correlated with probability of febrile illness and (GLM logistic: p = 0.0833).
For asymptomatic individuals, enrolled in weekly surveillance, blood smear microscopy results were not available and reported to the participant until the next scheduled visit, seven days later. This protocol was necessary to prioritize microscopy reading on symptomatic participants and to maintain a logical schedule of participant follow-up visits. Of the 45 individuals in the weekly follow-up visits who had P. falciparum infections, 30 individuals were asymptomatic with no indication of parasites at the time of sampling. However, upon reading their blood smears by microscopy, parasites were detected. The initial day of parasitaemia detection was termed "day i." The follow-up visit of these "day i" positive but asymptomatic individuals was termed "day i+7." At this regularly scheduled visit of these 30 individuals, the regular blood smear was collected and read by microscopy on the same day. However, sixteen individuals did not have parasitaemia upon the "day i+7" visit and remained asymptomatic. Six of the sixteen individuals were treated, even though they were negative by microscopy at the time of treatment, "day i+7". Ten of the sixteen individuals with earlier transient parasitaemias were not given treatment, opting for continued surveillance on the following day ("day i+8"). All ten of these asymptomatic individuals with transient parasitaemias on "day i", but negative on "day i+7" remained asymptomatic and were again negative "day i+8", and despite continued surveillance throughout the study remained without parasitaemia.
In the 14 cases where asymptomatic parasitaemias were not self-limited to being undetectable by microscopy, the parasite densities were often decreased in the "day i+7" blood sample taken just before treatment. From the 30 asymptomatic individuals, the geometric mean parasite density in the "day i" visits was 1047 parasites/μl blood (95%CI: 951–1153). Considering the 14 individuals with parasitaemia on "day i+7" the geometric mean parasite density was 905 parasites/μl blood (95%CI: 758–1082). The parasite density was higher in the "day i+7" versus the "day i" blood smear in only 9 of the 30 (20%) of the asymptomatic individuals. Therefore, it appeared that the majority of asymptomatic individuals were self-limiting their infections and would not have been detected with malaria if it were not for active case detection.
Discussion
Plasmodium falciparum and Plasmodium vivax malaria have been continuing in low transmission, as detected by passive case detection, since the 1995–1998 epidemic. This study determined the presence of a competent malaria vector, the local spatial and temporal clustering of infections and the prevalence of both symptomatic and asymptomatic parasitaemia in a suburban community.
The potential for local transmission by Anopheles darlingi was observed in each village of Zungarococha. Local transmission was demonstrated by finding that over 90% of the infections were in individuals not reporting travel outside of the village. The An. darlingi biting rate, especially in April, 2003, was high (10–24 An. darlingi bites/person/hour, with an estimated 5 hours of An. darlingi biting per night). Anopheles darlingi is the most competent malaria vector in South America. However, An. darlingi appears to be a less competent vector for malaria transmission than the Anopheles species prevalent in Africa[23]. Malaria studies in Africa, where the P. falciparum prevalence is higher, generally report that 1–20% of the important vector species are infected with P. falciparum sporozoites[24]. The high An. darlingi biting rate and the high proportion of human infections with gametocytes (53% of the P. falciparum infections) suggests ample potential for malaria transmission within each village.
The infection incidence by passive case detection in the Zungarococha Health Post was compared to the infection prevalence by active and passive case detection (active+passive) in 100-metre zones defined as at-risk to local P. falciparum transmission, based upon the detection of a P. falciparum infection within the preceding 30 days. The Health Post passive detection showed a population incidence of 0.03 P. falciparum and 0.22 P. vivax infections/person/malaria season. In individuals living in at-risk zones, participating in the cross-sectional survey and participating in weekly visits for one month as randomly selected as sentinels, the active+passive case detection prevalence was 0.13 P. falciparum and 0.39 P. vivax infections/person/malaria season. Therefore, the P. falciparum prevalence detected by active+passive case detection in at-risk zones was 4.77-fold higher than the passive case detection.
One attributable-factor for the 4.77-fold increase in detecting P. falciparum infections within at-risk zones versus the overall population prevalence is the clustering of infections within at-risk zones in Zungrococha. Passive case detection summarizes the infection incidence by considering the number of infections per the population size of the community. However, because infections were clustered in certain regions of the community, using the whole population as the denominator underestimated the prevalence. A household clustering of malaria infections was detected in a study by Brooker et al[25]. Brooker et al determined that households where children were low weight, households in lower altitudes and household where there were not drugs kept in the house were more likely to have P. falciparum infections[25]. In our cohort, the socioeconomic status, environment, house construction, and access to drugs (highly regulated by MINSA) are all very homogenous. Given the homogeneous environment, the same-house and adjacent-house clustering is most likely attributable to mosquito behavior: that is, to highly localized mosquito biting. The results are consistent with a concept that although An. darlingi has the potential to fly one kilometre or more to find a blood meal, An. darlingi and other Anopheles species frequently remain in highly restricted localities where hosts are readily available[4,21,26]. The clustering of infections suggests that An. darlingi return for multiple feedings in the same or adjacent houses. Future genetic typing of the malaria parasites both in the human host and mosquito vector will compare the genetic relatedness of the spatially (adjacent houses) and temporally (within one month) clustered infections versus the infections detected in other regions within the villages.
Another attributable-factor for the increased rate of detecting malaria infections within at-risk zones, as compared to the detection in the overall population, is the detection of asymptomatic infections. This study used a broad definition of symptoms: including any fever (reported or observed) or anemia (haematocrit <20% PCV) during the entire month of active follow-up. The symptoms reported from individuals detected with P. falciparum or P. vivax infections were relatively mild. There was a correlation between febrile illness and P. falciparum parasite density, which was independent of the correlation between febrile illness and age, as has been detected in other studies[27]. However, even with a wide range of parasite densities observed in this study, only 41% and 24% had a febrile illness within one month of the detected P. falciparum or P. vivax infection, respectively. This frequency of febrile illness is lower than that reported by a cross-sectional study in this region conducted by Roshanravan et al, in 1999, only one year after the epidemic's onset[20]. In the new study, presented here, the longitudinal dynamics of parasitaemia in asymptomatic individuals were investigated. In 80% of the asymptomatic cases, the parasitaemia decreased by the end of the one-week sampling interval (the one-week follow-up visit). Of the asymptomatic P. falciparum infections, 53% (16/30) had no detectable parasitaemia when they were re-visited or on successive checks for parasitaemia during the following 7 days. Moreover, 10 of the 16 individuals had continued surveillance by at least one additional blood sample 14 days later and continued surveillance by passive case detection throughout the malaria season. Thus, all available evidence suggested that these infections were self-limited.
Having infection follow-up more frequently than 3 times in one week (as in this study) has occurred in a few studies of individuals who had many prior malaria infections. Daily sampling for over two weeks on semi-immune individuals with asymptomatic P. falciparum infections in Papua New Guinea, Senegal and Tanzania showed that P. falciparum parasites remained detectable by microscopy in almost all blood samples [12-15]. As in the studies of semi-immune individuals described above [12-15], the parasite densities of asymptomatic individuals in this study were, by definition, ≥5,000 parasite/μl. Despite the similar parasite densities under consideration, in this low transmission Peruvian community, 16 of the 30 asymptomatic individuals (53%) had no detectable parasitemia in the subsequent follow-up visits. A precise understanding of self-limited infections will require further study of more individuals.
Considering the frequency of asymptomatic infections, with symptoms observed for one month, the results of this study suggest that approximately 40% of infections detected by active case detection would not have been detected by passive case detection in the Health Post. Asymptomatic malaria infections have also been detected in a longitudinal study in Brazil, where transmission was higher than this Peruvian cohort but still considered low transmission [18]. Of particular importance in this Peruvian population study, the prevalence of the sexual stage gametocytes was considered. There were asymptomatic cases with circulating gametocytes, providing the potential to infect mosquitoes and cause additional human cases. Therefore, although against the conventional thought that P. falciparum protective immunity requires years of frequent P. falciparum infections to develop [5-17], the need for active case detection in low transmission regions must be considered.
This study compared the malaria incidence by passive case detection for the total community with the prevalence rates in the 100-metre at risk zones for malaria. The limitation of not having an a priori-defined monthly control group in non-at-risk zones is that the incidence of asymptomatic malaria infections within these non-at-risk zones could not be calculated. Now that the possibility of asymptomatic infections in this community is known, a future study will include such control groups for conducting cost-benefit analysis.
The results of this present study, however, are not significantly limited by lacking precise knowledge of asymptomatic infections in non-at-risk zones. First, the attributable-fraction of the increased prevalence in at-risk zones can be estimated by assuming that the frequency of asymptomatic infections in at-risk zones was similar to that in non-at-risk zones. Supposing that, without active case detection, 40% of the P. falciparum infections would have gone undetected in the health post, then there is an approximately a 2.6-fold increase (4.33*0.60 = 2.6) in the P. falciparum active+passive prevalence versus the passive population incidence attributable to clustering of infections within at-risk zones. Second, the spatial and temporal clustering of infections was apparent within at-risk zones, with adjacent houses being at highest risk within at-risk zones. By focusing resources for active case detection in the at-risk zones, more than three times the malaria infections were detected and treated versus passive case detection. Future studies will determine the range of focused active case detection and fumigation campaigns to obtain the most cost-effective method for immediate public health and long-term decreases in malaria transmission.
Conclusion
The occurrence of clustered infections in at-risk zones and asymptomatic infections contributed to the higher prevalence by active case detection within at-risk zones versus passive case detection generalized to the community population. Currently, passive case detection and treatment is the primary approach to malaria control in Peru. This control method is consistent with PAHO's program guidelines, which do not advocate active case detection. However, according to this longitudinal study in a low transmission community, this control strategy only targets approximately one-third of the malaria infections. Therefore, in regions where active case detection cannot be sustained throughout a community, feasible alternatives to detect and limit malaria transmission are greatly needed.
Carter et al. discussed the importance of concentrating often limited and expensive resources and manpower where they would be most effective [26]. Studying spatial dynamics of malaria infection in low transmission regions might be a particularly effective model system for determining how to target malaria control efforts [26]. Continued investigation in this community will determine if there is a long-term clustering of infections in houses across malaria transmission seasons. At that point, a retrospective definition of malaria risk at the household level, using mapping aided by Global Positioning Satellite technology, can be used to better target control measures. Although more investigation and cost-benefit analysis is needed, the study presented here suggests one more immediate and readily applicable intervention strategy.
It is evident that passive case detection, even in low transmission, does not identify the regions most at risk for infection and has the potential of leaving many asymptomatic malaria cases undetected. However, community-wide active case detection in regions of low transmission is highly cost prohibitive for local governments. Additionally, by focusing in regions more at risk to transmission, community participation in active case detection is more easily sustained. The results of the present study suggest a specific public health strategy where resources for active case detection are limited: after detecting a malaria infection in a house in a given month by passive case detection, in the following month actively survey for malaria infections in this and adjacent houses and focus insecticide spraying in these at-risk houses.
Authors' contributions
OHB designed the project, supervised the field team and directed the research team. OHB performed the data management, analysis and writing of the manuscript. WMC coordinated collaboration with MINSA and provided critical background statistics on malaria in Iquitos, such as Figure 1B. DG assisted in developing the consent forms and questionnaires and supervised the storage of samples and integration with the Peruvian University, UPCH-IMTavH. JNH, physician and coordinator, conducted physical examinations, asked epidemiologic and clinical questionnaires as well as coordinated efforts in the field. FFA was the lead field entomologist as well as the participant recruitment specialist responsible for selecting the sentinel houses each month for active case detection and obtaining and continually verifying participants' informed consent. NR organized the field team, kept daily follow-up records and maintained logistics between the clinic sampling and laboratory team. EA conducted daily field visits, acquiring the microscopy results, recording data and determining haematocrit packed cell volume from each visit. EP supervised the morphologic characterization of the mosquitoes into species and conducted the entomology analysis. EG coordinated the collaboration between UAB, the Peruvian University, UPCH-IMTavH, and MINSA. All authors have read and approved of the content of this manuscript.
Acknowledgements
We would like to thank the Zungarococha community members and authorities for their continued consent and commitment to the Malaria Immunology and Genetics in the Amazon (MIGIA) Project. This work was made possible by pilot funds awarded to OLB as a new UAB faculty member, the Gorgas Memorial Institute and the Sparkman Center for International Health. We thank Dr. Carlos Vidal, the Director of MINSA DISA-Loreto, who organized the many members of MINSA DISA collaborating in the MIGIA project. We thank the Zungarococha Health Post physician, Dr. Willy Alava, for his expertise and dedication in providing health care to the community. We thank the microscopists, Mr. Ever Alvarez and Mr. Anibal Sanchez, for their untiring dedication and exceptional skill in reading the blood smears. We thank the health technicians, Ms. Zoila Reategui and Ms. Elva Sanchez, for their warm care of the community and efficient zeal in their work, preparing blood slides and assisting the doctors each day. We also thank Prof. Humberto Guerra, who, in addition to translation assistance, contributes to our keeping the highest standards of human subject research. We thank Prof. Donald Roberts for his insight and guidance in the biology and behavior of An. darlingi and his permission for us to use his previously published figure[1] in Figure 1A.
==== Refs
Aramburu Guarda J Ramal Asayag C Witzig R Malaria reemergence in the Peruvian Amazon region Emerg Infect Dis 1999 5 209 215 10221872
Roberts DR Laughlin LL Hsheih P Legters LJ DDT, global strategies, and a malaria control crisis in South America Emerg Infect Dis 1997 3 295 302 9284373
Manguin S Roberts DR Peyton EL Fernandez-Salas I Barreto M Fernandez Loayza R Elgueta Spinola R Martinez Granaou R Rodriguez MH Biochemical systematics and population genetic structure of Anopheles pseudopunctipennis, vector of malaria in Central and South America Am J Trop Med Hyg 1995 53 362 377 7485688
Arata AA Ruebush TK Porter CH Stein JD Lounibos LP Fernandez DM 60 ARN Malaria in the Peruvian Amazon: a review of the epidemiology, entomology and insecticide resistance of vectors Environmental Health Project 1999 Washington, DC: USAID
Snow RW Omumbo JA Lowe B Molyneux CS Obiero JO Palmer A Weber MW Pinder M Nahlen B Obonyo C Newbold C Gupta S Marsh K Relation between severe malaria morbidity in children and level of Plasmodium falciparum transmission in Africa [see comments] Lancet 1997 349 1650 1654 9186382 10.1016/S0140-6736(97)02038-2
Baird JK Krisin Barcus MJ Elyazar IR Bangs MJ Maguire JD Fryauff DJ Richie TL Sekartuti Kalalo W Onset of clinical immunity to Plasmodium falciparum among Javanese migrants to Indonesian Papua Ann Trop Med Parasitol 2003 97 557 564 14511553 10.1179/000349803225001472
Richard-Lenoble D Kombila M Guillemot O Guillemot F Monjour L Gendrel D [Anti-measles vaccination and malaria] Arch Fr Pediatr 1987 44 817 819 3328577
Colbourne MJ Malaria in gold Coast students on their return from the United Kingdom Trans R Soc Trop Med Hyg 1955 49 483 487 13267916 10.1016/0035-9203(55)90017-1
Greenwood B The impact of malaria chemoprophylaxis on the immune status of Africans Bull World Health Organ 1984 62 69 75 6397278
van Dijk M Mass chemoprophylaxis with chloroquine additional to DDT indoor spraying: report on a pilot project in the Demta region, Netherlands new Guinea Trop Geogr Med 1958 10 379 384 13635858
Metselaar D Relative increase in the prevalence of Plasmodium falciparum some years after the beginning of a house spraying campaign in Netherlands New Guinea Trans R Soc Trop Med Hyg 1960 54 523 528 13769878 10.1016/0035-9203(60)90026-2
Farnert A Snounou G Rooth I Bjorkman A Daily dynamics of Plasmodium falciparum subpopulations in asymptomatic children in a holoendemic area Am J Trop Med Hyg 1997 56 538 547 9180605
Smith T Beck HP Kitua A Mwankusye S Felger I Fraser-Hurt N Irion A Alonso P Teuscher T Tanner M Age dependence of the multiplicity of Plasmodium falciparum infections and of other malariological indices in an area of high endemicity Trans R Soc Trop Med Hyg 1999 93 15 20 10450421 10.1016/S0035-9203(99)90322-X
Zwetyenga J Rogier C Spiegel A Fontenille D Trape JF Mercereau-Puijalon O A cohort study of Plasmodium falciparum diversity during the dry season in Ndiop, a Senegalese village with seasonal, mesoendemic malaria Trans R Soc Trop Med Hyg 1999 93 375 380 10674080 10.1016/S0035-9203(99)90122-0
Bruce MC Donnelly CA Packer M Lagog M Gibson N Narara A Walliker D Alpers MP Day KP Age-and species-specific duration of infection in asymptomatic malaria infections in Papua New Guinea Parasitology 2000 121 247 256 11085245 10.1017/S0031182099006344
Branch OH Takala S Kariuki S Nahlen BL Kolczak M Hawley W Lal AA Plasmodium falciparum genotypes, low complexity of infection, and resistance to subsequent malaria in participants in the Asembo Bay Cohort Project Infect Immun 2001 69 7783 7792 11705960 10.1128/IAI.69.12.7783-7792.2001
Singer LM Mirel LB ter Kuile FO Branch OH Vulule JM Kolczak MS Hawley WA Kariuki SK Kaslow DC Lanar DE Lal AA The effects of varying exposure to malaria transmission on development of antimalarial antibody responses in preschool children. XVI. Asembo Bay Cohort Project J Infect Dis 2003 187 1756 1764 12751033 10.1086/375241
Alves FP Durlacher RR Menezes MJ Krieger H Silva LH Camargo EP High prevalence of asymptomatic Plasmodium vivax and Plasmodium falciparum infections in native Amazonian populations Am J Trop Med Hyg 2002 66 641 648 12224567
Roper MH Torres RS Goicochea CG Andersen EM Guarda JS Calampa C Hightower AW Magill AJ The epidemiology of malaria in an epidemic area of the Peruvian Amazon Am J Trop Med Hyg 2000 62 247 256 10813480
Roshanravan B Kari E Gilman RH Cabrera L Lee E Metcalfe J Calderon M Lescano AG Montenegro SH Calampa C Vinetz JM Endemic malaria in the Peruvian Amazon region of Iquitos Am J Trop Med Hyg 2003 69 45 52 12932096
Roberts DR Manguin S Rejmankova E Andre R Harbach RE Vanzie E Hakre S Polanco J Spatial distribution of adult Anopheles darlingi and Anopheles albimanus in relation to riparian habitats in Belize, Central America J Vector Ecol 2002 27 21 30 12125869
Huaman MC Roncal N Nakazawa S Long TT Gerena L Garcia C Solari L Magill AJ Kanbara H Polymorphism of the Plasmodium falciparum multidrug resistance and chloroquine resistance transporter genes and in vitro susceptibility to aminoquinolines in isolates from the Peruvian Amazon Am J Trop Med Hyg 2004 70 461 466 15155976
da Silva-Vasconcelos A Kato MY Mourao EN de Souza RT Lacerda RN Sibajev A Tsouris P Povoa MM Momen H Rosa-Freitas MG Biting indices, host-seeking activity and natural infection rates of anopheline species in Boa Vista, Roraima, Brazil from 1996 to 1998 Mem Inst Oswaldo Cruz 2002 97 151 161 12016435
Beier JC Killeen GF Githure JI Short report: entomologic inoculation rates and Plasmodium falciparum malaria prevalence in Africa Am J Trop Med Hyg 1999 61 109 113 10432066
Brooker S Clarke S Njagi JK Polack S Mugo B Estambale B Muchiri E Magnussen P Cox J Spatial clustering of malaria and associated risk factors during an epidemic in a highland area of western Kenya Trop Med Int Health 2004 9 757 766 15228485 10.1111/j.1365-3156.2004.01272.x
Carter R Mendis KN Roberts D Spatial targeting of interventions against malaria Bull World Health Organ 2000 78 1401 1411 11196487
Naik RS Branch OH Woods AS Vijaykumar M Perkins DJ Nahlen BL Lal AA Cotter RJ Costello CE Ockenhouse CF Davidson EA Gowda DC Glycosylphosphatidylinositol anchors of Plasmodium falciparum : molecular characterization and naturally elicited antibody response that may provide immunity to malaria pathogenesis J Exp Med 2000 192 1563 1576 11104799 10.1084/jem.192.11.1563
|
15975146
|
PMC1190209
|
CC BY
|
2021-01-04 16:24:13
|
no
|
Malar J. 2005 Jun 23; 4:27
|
utf-8
|
Malar J
| 2,005 |
10.1186/1475-2875-4-27
|
oa_comm
|
==== Front
Malar JMalaria Journal1475-2875BioMed Central London 1475-2875-4-341604278010.1186/1475-2875-4-34Case StudyCreating an "enabling environment" for taking insecticide treated nets to national scale: the Tanzanian experience Magesa Stephen M [email protected] Christian [email protected] Don [email protected] Jane E [email protected] Ritha JA [email protected] Karen [email protected] Andrew [email protected] Alex [email protected] National Institute for Medical Research, P.O. Box 9653, Dar es Salaam, Tanzania2 Swiss Tropical Institute, P.O. Box, 4002 Basel, Switzerland3 Population Services International, P.O. Box 33500, Dar es Salaam, Tanzania4 WHO Country Office, P.O. Box 9292, Dar es Salaam, Tanzania5 National Malaria Control Programme, Ministry of Health, P.O. Box 9083, Dar es Salaam, Tanzania2005 22 7 2005 4 34 34 15 3 2005 22 7 2005 Copyright © 2005 Magesa et al; licensee BioMed Central Ltd.2005Magesa et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Introduction
Malaria is the largest cause of health services attendance, hospital admissions and child deaths in Tanzania. At the Abuja Summit in April 2000 Tanzania committed itself to protect 60% of its population at high risk of malaria by 2005. The country is, therefore, determined to ensure that sustainable malaria control using insecticide-treated nets is carried out on a national scale.
Case description
Tanzania has been involved for two decades in the research process for developing insecticide-treated nets as a malaria control tool, from testing insecticides and net types, to assessing their efficacy and effectiveness, and exploring new ways of distribution. Since 2000, the emphasis has changed from a project approach to that of a concerted multi-stakeholder action for taking insecticide-treated nets to national scale (NATNETS). This means creating conditions that make insecticide-treated nets accessible and affordable to all those at risk of malaria in the country. This paper describes Tanzania's experience in (1) creating an enabling environment for insecticide-treated nets scale-up, (2) promoting the development of a commercial sector for insecticide-treated nets, and (3) targeting pregnant women with highly subsidized insecticide-treated nets through a national voucher scheme. As a result, nearly 2 million insecticide-treated nets and 2.2 million re-treatment kits were distributed in 2004.
Conclusion
National upscaling of insecticide-treated nets is possible when the programme is well designed, coordinated and supported by committed stakeholders; the Abuja target of protecting 60% of those at high risk is feasible, even for large endemic countries.
==== Body
Background
Like most countries in sub-Saharan Africa, Tanzania carries a heavy malaria disease burden. It is estimated that 28 million citizens are exposed to the risk of stable malaria, resulting in 16 million clinical episodes per year and 100,000 child deaths – over 25% of total deaths [1]. In addition, malaria represents the leading cause of outpatient attendance for children and adults (38% and 32%, respectively) and the cost to health services is, therefore, considerable. The negative impact on the socio-economic development of the country is undoubtedly large. In order to address this enormous burden, Tanzania recently developed a national malaria medium-term strategic plan [1]. With regard to primary prevention, the emphasis has been put on the widespread use of insecticide-treated nets (ITNs). Globally, ITNs have been shown to reduce morbidity and mortality from malaria substantially in over 80 settings [2].
A target of 60% coverage of ITNs by 2005 in the groups bearing the highest malaria risk (young children and pregnant women) was set at the meeting of African Heads of State on malaria in Abuja, Nigeria, in April 2000. This has provided a clear goal for the way forward and stressed the need to rapidly expand access to, and use of, ITNs. Here, a summary of the extensive Tanzanian experience with ITNs, as well and the long and complex process leading to a national scale-up of ITNs, are presented. This paper also attempts to review some of the issues that are crucial for this process, with the hope that this experience might prove useful to other countries.
Case Description
ITN developments in Tanzania, 1983 to 2005
Over the last twenty years, much work has taken place on ITNs in Tanzania (Table 1). Various research organisations, donor agencies, non-governmental organizations (NGOs), the private sector and government agencies have been involved in improving this tool and preparing large-scale expansion.
Table 1 The critical path of insecticide-treated nets (ITN) research and implementation in Tanzania, 1983 to 2004.
Critical path of insecticide-treated nets research and Implementation in Tanzania
Efficacy studies Effectiveness studies Policy developments Going to scale
Reducing malaria vector exposure (including net and insecticide developments) Reducing malaria morbidity and mortality Impact (morbidity and mortality) and cost assessment in pilot programmes National strategies and partnerships for an enabling environment National ITN strategy and policy NATNETS
1983–1995 1985–1995 1992–2000 1997–2000 >2000
A number of studies have focused on the entomological action of ITNs [3-6]. Following this initial work a number of small-scale trials demonstrated epidemiological impact under controlled conditions – also called efficacy [7-11].
Subsequently, the emphasis was put on larger trials and exploring community-wide benefits of ITNs on both morbidity and mortality. Impact under programme conditions (effectiveness) was demonstrated in 1997–2000 by the Kilombero Net Project (KINET) with a reduction of 27% in the risk of dying associated with ITN use [12]. Even larger beneficial effects were seen for anaemia and parasitaemia in children [13] and in pregnant women [14]. This programme also demonstrated the cost-effectiveness of ITNs under programme conditions [15], with a cost per death averted of USD 1018, making ITNs comparable to childhood immunization.
While studying impact much was also learned about different approaches to ITNs and insecticide distribution, either based on community mechanisms [16-18] or on social marketing [19,20]. Tanzania was also the place for an international ITN meeting in 1994 attempting to bring together all the early experience of ITN implementation, and resulting in the publication of the first book dedicated to this topic [21].
An important development in Tanzania was the design and testing of insecticide home treatment kits [22,23]. Further studies on the issue of net re-treatment highlighted the difficulties associated with sustaining this behaviour [24]. Recently, work was also done on the state of existing polyester nets [25], as well as the remarkable state of Olyset™ long-lasting nets 8 years after their introduction [26].
This large body of work demonstrated that ITNs were highly efficacious, as well as effective, that they represented a cost-effective intervention, and that feasible approaches existed for large-scale expansion. By 2000, two social marketing programmes were operating. Firstly, the Kilombero Net Project (KINET) managed by the Ifakara Health Research and Development Center (IHRDC) and the Swiss Tropical Institute (STI), which has been operating since 1997 in two southern districts [19]. Secondly, the Social Marketing for ITNs project (SMITN) implemented by Population Services International (PSI) which started in 9 districts in 1998 and which went national in mid-2000 (SMITN Phase 2).
From the population census in 2002, the Tanzania population has been estimated at 34,569,232 (mainland 33,584,607) living in 6,996,036 (mainland 6,811,087) households [27]. With a total population at risk of 28 million [28]. and two users on average per net, a minimum standing crop of 14 million treated nets are required to protect the whole population, while 8.4 million nets are required for an overall coverage of 60%. In addition, the same number of treatment kits are required in order to re-treat every net once per year until the introduction of long-lasting insecticidal nets (LLIN). While the official target of NATNETS is to protect 60% of those at high risk of dying from malaria (pregnant women and children under five years), a national programme must obviously also take into account the fact that the rest of the population wants to be protected. Overall ITN distribution targets are hence calculated for the whole population (bearing in mind that only high-risk groups will benefit from a subsidy).
At the 1999 International Conference on ITNs in Dar es Salaam, Tanzania, a call was made to go beyond projects and programmes towards changing norms on a national scale [29]. This ultimately requires working towards a future in which all those in need of protection in malarious areas enjoy easy access to ITNs and use them consistently.
A concerted national action for ITNs in Tanzania
During the International ITN Conference in Dar es Salaam in 1999, an opportunity was seized to hold the first ITN stakeholder meeting in the country with a focus on national scaling up. The meeting attracted representatives from the Tanzanian private sector (mosquito nets and insecticide manufacturers, marketing agencies), the public sector (Ministry of Health – MoH), the research and academic communities, NGO's, bilateral donors (UK, Netherlands, Switzerland) and multilateral agencies (UNICEF, WHO). A consensus was reached on the need for a forum including all stakeholders to plan an overall strategy of how Tanzania should proceed. A planning meeting was held in early November 1999 and at this meeting PSI was asked by the MoH to take the lead in organising a large meeting that took place in March 2000. Over 40 stakeholders, representing all major constituencies, attended the workshop. The key outcome was the establishment of a Task Force representing all the stakeholder groups and the framing of roles for all stakeholders (Figure 1). As a first step, this Task Force, under the chairmanship of the MoH, commissioned and facilitated the drafting of a "National Strategic Plan for Insecticide-Treated Nets in Tanzania", which was completed in September 2000, then refined and endorsed by all parties at a second workshop attended by over 50 stakeholders [30]. In a final step, MoH directors officially endorsed the strategic plan in November 2000. The three core concepts of the national ITN strategy (NATNETS) are:
Figure 1 Strategic framework for ITN Scaling up in Tanzania.
(1) Increased demand creation for ITNs.
(2) A national public-private partnership for developing a sustainable domestic commercial ITN market.
(3) Targeted subsidies aimed at high-risk groups.
A close collaboration among the public, private and NGO sectors was advocated around the issue of demand creation and increased supply and use of ITNs (Figure 1). The public sector role is focusing on consumer protection, policy and regulatory issues, as well as generic demand creation, in order to create an ITN-enabling environment. The NGO role focuses on more local, grass-root demand creation and support for specific niche supply. The commercial sector role focuses on supply and distribution, product development, and brand-specific demand creation. Underpinning these is the research community, assisting with product development, implementation research, market research, and monitoring and evaluation. Bilateral donors provide strategic funding support, as well as strategic thinking across sectors.
From 1999 to 2002, the MoH Task Force and its membership of stakeholders have facilitated and coordinated all ITN activities in the country. Key landmarks in this process were (1) a national ITN implementation plan and budget produced by the National Malaria Control Programme (NMCP) and (2) a successful application to the Global Fund to Fight AIDS, TB and Malaria in 2002 to support a targeted national ITN subsidy programme to increase coverage and equity (see below). So far, all stakeholders, including NGOs and multilateral partners have aimed towards coordinating all activities within NATNETS and no alternative programmes have been implemented in the country.
With a new ITN landscape emerging in the country, a Steering Committee, made up of Tanzanian Government departments, PSI, STI and key donors, was created in July 2002 as an overseeing body for NATNETS, under the chairmanship of the Chief Medical Officer (CMO). A second group with a wider membership was created to provide a more general stakeholder forum for all aspects related to the scaling-up process – the ITN consultative group. An ITN cell was also created within the NMCP in May 2003 to become the executive body coordinating and supporting the ITN process in the country (Figure 2). To this effect, a full time NATNETS Team Leader was recruited in May 2003, together with two professional officers. Support was provided by the MoH and the Swiss Agency for Development and Cooperation (SDC) through the Swiss Topical Institute.
Figure 2 The National Malaria Control Programme structure showing the position of the ITN cell and the main components of the national ITN strategy (NATNETS). CMO = Chief Medical Officer, DPS = Director of Preventive Services, NMCP = National Malaria Control Programme, IEC = Information, Education, Communication.
Practically, NATNETS is now based on three major operational components:
(1) The ITN cell – a coordination unit to create an enabling environment for taking ITNs to national scale;
(2) Strategic social marketing (SMARTNET) to massively upscale the supply of ITNs;
(3) The Tanzanian national ITN voucher scheme (TNVS) to selectively target pregnant women with highly subsidized nets.
The areas of responsibility of the key ITN stakeholders in the country are presented in Table 2, as an extension from Figure 1. Since these operational components are the basis of NATNET they will be reviewed in turn below.
Table 2 Shared roles within the national ITN initiative (NATNETS).
NATNETS National Malaria Control Programme (NMCP)
ITN cell within NMCP
• Overall coordination of ITN activities, including linkage with NGOs and commercial sector
• Advocacy (political, administrative)
• Reduction of taxes and tariffs
• Legislation on insecticides & consumer protection
• Quality control (nets and treatment)
• Technical support to districts
• Brokering agreements between public sector and commercial sector
• Resource mobilisation for other aspects of ITN Strategy
• Increasing role in national/local level demand creation
• Monitoring and evaluation
• Design & implementation of national voucher scheme Strategic social marketing (SMARTNET)
• Market research and creation of a national ITN logo
• Opening new markets in areas not yet reached by commercial markets and attracting commercial distribution into such areas
• Marketing support to national net manufacturers/distributors/retailers
• Distribution support to national net manufacturers/distributors/retailers
• National/local demand creation
• Net treatment kit promotion and distribution
• Close collaboration with ITN Cell
Additional key ITN stakeholders
Research community
• Monitoring and evaluation
• Implementation research
• Market research
• Product development Commercial sector
• Net production
• Net distribution
• Branded advertisement
• Product development
• Market research
Creating an enabling environment for ITNs
Neither the public sector nor the commercial sector alone can achieve the level of coverage required to meet and exceed the Abuja target. This is evidenced by the fact that the Government of Tanzania and its donor partners spend only USD 5.10 per capita per year on the entire health system [31]. Other partners including NGOs and the private sector spend an additional USD 0.80 per capita. Households, for their part, contribute the largest share of health expenditure, averaging USD 5.35 per capita. Within this limited budget, it would be impossible to allocate a substantial part of public spending to a single intervention for a single disease. Furthermore, the current health structures would be unduly stretched by the distribution of a large volume of valuable goods. There is ample evidence to suggest that nets are a highly attractive commodity once available within a reasonable distance, and at a low price [18,20,21].
As a result, there is a strong consensus both in Tanzania and internationally [32] that the best way forward is a well-coordinated partnership among all ITN stakeholders based on increased demand and supply, a vibrant commercial sector, and a targeted subsidy scheme for those most at risk. In order to create and sustain this partnership a shared vision is required, as well as a number of enabling factors. Some of the key enabling factors are: 1) removal of any form of taxation; 2) favourable insecticide regulatory conditions; 3) net quality control issues; 4) generic demand creation by the public sector; and 5) equity of access.
Taxes and Tariffs
Affordability of mosquito nets is the major single factor limiting wide and equitable coverage of mosquito nets, especially in rural settings. Removing all taxes and tariffs on nets, netting and insecticide is a relatively simple and very effective way to make ITNs more affordable. In 1995 imported nets and locally manufactured nets were subject to customs duty (tariff) and sales tax, (later value added tax – VAT). In addition, there were licensing requirements (including application and permit fees) and foreign exchange controls limiting imports.
National manufacturing companies experienced a difficult period in the sales of ready-made nets between 1980 and 1992 when the government levied a sales tax of 125%. At this point netting material was not taxed and the sole domestic manufacturer at that time (Sun Flag Ltd) found it more worthwhile to sell netting material to local tailors than to turn netting into nets. As a result, ready-made nets were luxury items for the majority of Tanzanians. In 1994, when ITNs were beginning to be recognized as important tools for malaria control, the sales tax was removed on ready-made nets. This resulted in a substantial price drop and the net manufacturing company began to experience a rise in sales. As a result, a second domestic textile manufacturer started production in 1998 (A-Z Textiles Ltd).
In May 1998, the MoH, with support from PSI and the Programme for Applied Technology (PATH), Canada, convened a meeting in Tanzania with the purpose of promoting the use of treated nets within the country with an emphasis on partnership between the private and public sectors. It brought together representatives of the MoH, Ministry of Finance, Ministry of Trade & Industries, textile manufacturers, insecticide manufacturers, NGO's, bilateral and multilateral organizations. The meeting unanimously agreed that VAT, which was to be introduced for the first time in Tanzania, was going to affect negatively the pricing of nets. It was decided that each stakeholder should petition against the proposed VAT on nets. The meeting document was circulated to the Permanent Secretaries of the Ministries of Health, Finance, Industries & Trade, to the World Bank, WHO and the media. The NMCP was urged to present the national ITN strategy to the MoH so that the Minister could present it at the next Parliamentary session.
Despite these efforts, VAT was introduced on nets and netting materials during the 1998/99 budget. But shortly after, thanks to the stakeholders' intensive lobbying, an amendment was issued by the Ministry of Finance waiving VAT on mosquito nets. Importers of ready-made nets had to pay only 5% import duty. Unfortunately, local producers still had to pay 10% import duty and 20% VAT on raw materials (polyester yarn, thread and packaging materials), utilities and machinery. This put them at a disadvantage compared to imported nets. Further lobbying was, therefore, undertaken. This finally resulted in the 1999/2000 Finance Bill that declared mosquito nets and insecticides "zero rated" items. The bill classified ready-made mosquito nets as essential drugs, like other pharmaceuticals, which attract only 5% import duty and are exempted from VAT. However, in revised legislation taking effect in July 2002, these provisions were again removed. In response to this, the national ITN Task Force had further consultations with government authorities including the Tanzania Revenue Authority. From the end of 2004, all netting items were again zero-rated for VAT and the VAT on imports of inputs (mainly the yarn) can be re-claimed against proof that the material was used for netting manufacture.
This experience shows clearly that success depends on continuously lobbying the finance and tax departments, arguing that the economic and health benefits of ITNs outweigh concerns about loss of government tax revenue. Tanzania's experience also illustrates the need for vigilance by the health authorities and ITN stakeholders since taxes and tariffs are constantly shifting. As a result of these efforts the local manufacturing of nets has been stimulated and the average retail price of Tanzanian nets has gone down from over USD 5 in 1995 to less than USD 3.5 in 2004. This has made ITNs substantially more affordable. With better market prospects, a third net manufacturer started selling nets in 1999 (TMTL Ltd), bringing the combined annual net production to over 5 million. Recently, a fourth manufacturer (Motex, Moshi Textile Mills, Ltd.) has started production. Taxes and tariffs on insecticide are also a potential issue, but so far all insecticide used for ITNs has been imported by tax-exempted projects and this issue has not arisen. It will, however, have to be dealt with in the future.
Regulatory issues concerning insecticides
Regulatory issues are important in a number of aspects including insecticide registration, the definition of authorized retail outlets, consumer rights and product quality. Efforts have been made to speed up the process of insecticide registration by proposing to fast-track those products already approved by the WHO Pesticides Evaluation Scheme (WHOPES).
Another issue was that previously only registered pharmacies were allowed to sell home insecticide re-treatment kits. After a number of meetings and a thorough review, the Tanzanian Pesticides Research Institute (TPRI) as well as the MoH agreed that such kits could be distributed more widely by the SMITN and KINET social marketing projects. The TPRI regulations had to be amended to this effect and thanks to prompt action by all concerned this was agreed in 2000. In a further step and following a private-public sector agreement in 2002, insecticide kits were bundled with all nets leaving Tanzanian factory doors for the domestic market. Hence, from the end of 2002 onwards, the vast majority of nets are sold with insecticide in the country (some unbundled, imported, poor quality Far-East Asian nets are sometimes found). SMARTNET research showed that 92% of all nets sold with an insecticide kit were treated.
Strategic social marketing
In 2002, the PSI ITN social marketing project (SMITN) took on a new role under the name "Strategic Social Marketing for Expanding the Commercial Market for ITNs in Tanzania" (SMARTNET). SMARTNET is a 5-year initiative managed by PSI-Tanzania and funded by the British and Netherlands development aid programmes. Under SMARTNET the role of social marketing evolved substantially compared to previous projects. SMARTNET's primary focus has moved away from promoting and distributing PSI over-branded ITNs, to assisting the Tanzanian net manufacturers to expand their wholesale and retail network in the country, especially in rural areas. It is expected that this will lead to a sustainable and vigorous net market by 2007. Support includes nation-wide multi- and single-brand advertisement of the net manufacturers' products, identification of primary agents/wholesalers/retailers, transport subsidies to remote locations and general support. All companies benefit basically from the same support, although in the end this support is obviously dependent on how active they become in developing the market. Multi-branded advertisement is achieved by including the logos of all manufacturers on the promotional materials developed by SMARTNET. An early success of SMARTNET has been to convince the Tanzanian net manufacturers to bundle all their nets destined for the Tanzanian market with insecticide treatment kits. This was made possible by supplying kits to them very cheaply (USD 0.15) and by making sure that all the manufacturers participated, hence avoiding unfair competition from untreated nets.
In 2004 the combined sales of Tanzanian manufacturers on the domestic market had reached nearly two million insecticide-bundled nets and the projections for 2005 are over 2.5 million nets as a result of the start of the Tanzanian National Voucher Scheme (TNVS). This illustrates clearly the growth potential of the market when assisted strategically by the public sector.
In addition, SMARTNET is continuing to develop a national distribution network for insecticide treatment kits (branded Ngao® and Ngao ya Maji®), since selling insecticide kits does not constitute an attractive long-term commercial market due to the development of long-lasting insecticidal nets (LLIN). The first of these, the Olyset® net developed by Sumitomo Corp. is now being manufactured by A-Z Textiles Ltd in Tanzania and will soon be available on the local market. SMARTNET intends to promote Olyset® nets as soon as they become available. In 2004, nearly 2.2 million additional insecticide treatment kits were sold.
By developing the supply chain, SMARTNET has been crucial to the success of the TNVS, which cannot work unless ITNs are available everywhere in the country. On the other hand, the TNVS helps develop and sustain the ITN market by generating a large demand for these products, especially among the poorer segments of the population (see below).
The Tanzanian National Voucher Scheme (TNVS)
The Tanzania National Voucher Scheme (TNVS) is a five-year scheme supported by the Global Fund to Fight AIDS, Tuberculosis and Malaria [33], which started operating in October 2004, after an 18 months preparation phase. The TNVS is modelled on a similar voucher scheme implemented by the KINET project from 1997 to 2001 [33]. It aims at giving every pregnant woman attending an antenatal clinic a printed voucher with a face value of TShs 2750 (USD 2.75 in 2004). This voucher can be used to purchase an ITN from any participating commercial retailer at a discounted price. The pregnant woman has to spend between USD 0.5 and USD 1.5 to get her ITN (current retail prices: USD 3 to 4, according to shape, colour and size). The retailers then redeem the voucher with their wholesaler and the wholesaler with regional teams working for the programme or with the net manufacturers. In addition, a free insecticide treatment kit is given to all women attending antenatal care, as well as to mothers attending vaccination clinics with their child at 3 and 9 months. Through this additional mechanism infants will be protected by a treated net from before birth until after the first year of life. The main aims of the TNVS is to provide a facilitated and equitable access to ITNs to those groups most at risk of the severe consequences of malaria, i.e. pregnant women (1.4 million every year) and their newborn children.
This approach was chosen instead of direct distribution of free nets for two main reasons: 1) avoiding the complex logistical problems of distributing ITNs through the already over-burdened public health system; 2) providing a subsidy mechanism that strengthens the development of a sustainable commercial distribution, especially in rural areas; 3) rapidly addressing equity concerns raised by a purely commercial market; 4) increasing the coverage of antenatal and MCH care.
By February 2005, preliminary results from the 7 initial regions where the TNVS was operating suggested:
• a major increase in the number of retail outlets selling ITNs: 700 retail outlets in 7 regions were involved in the TNVS, of which over 70% were new to the ITN business
• a major increase of sales figures at retail and wholesale level (50 – 80%)
• approximately 80% of the pregnant women in the TNVS areas using the voucher to buy an ITN
• attendance of ante-natal clinics by pregnant women earlier in their pregnancy. As a result of the TNVS clinics have indicated that they see many more women in their first trimester, making it possible to provide better overall health care
• active district involvement in ITN-upscaling activities as a result of intensive advocacy and negotiations with the different ministries involved
Hence, the TNVS is expected to support a rapid expansion of ITN use by both pregnant women and infants. Moreover, the TNVS is expected to support and encourage private sector involvement in the manufacture of ITNs and their delivery to poor rural communities, since there will be widespread and predictable demand for ITNs by pregnant women. The scheme is coordinated by the national ITN cell and implemented by tendered contractors. This was thought to be the most efficient and effective way of introducing a complex new public health service, and it could provide a model for future developments in the frame of health sector reforms. The ITN Steering Committee is overseeing the TNVS and there is a close coordination with the SMARTNET programme.
Costing NATNETS
Cost data for NATNETS are currently being collected systematically in the frame of routine evaluation and monitoring activities. Following cost figures are currently available.
The annual cost of the ITN cell amount to USD 364,000. This amount includes staff, management, running cost and technical support. No data are currently available on the cost of the SMARTNET component. Finally, the annual cost of the TNVS will amount to USD 5,385,000 once the programme operates at national level (1.2 million vouchers given out per year). Of this amount, USD 3,308,000 (61%) represent the value of the vouchers and the insecticide re-treatment kits, USD 1,384,000 (26%) will be the cost of logistics (distribution of vouchers and their redemption), and USD 693,000 (13%) will be used for training and promotion. Hence, the cost per net given out using a voucher amounts to USD 4.49, to which the recipient will have to add USD 0.5 to USD 1 to purchase an ITN.
Discussion and Lessons Learned
While this paper attempts to present the Tanzanian experience as objectively as possible, it is written by individuals who were all involved in these developments. In order to ensure an independent assessment of NATNETS, an evaluation and monitoring programme (including costing) is currently carried out by a joint team from the Ifakara Health Research and Development Center and the London School of Hygiene and Tropical Medicine. The evaluation does not entail measuring the impact of NATNETS on morbidity and mortality from malaria, since this has been demonstrated beyond doubt in many settings, including in Tanzania [2,12,13]. In any case, regular monitoring of child survival parameters will be available, either through existing demographic surveillance systems (of which there are currently four in the country) or through the regular Demographic and Health Surveys.
The strong development of the commercial sector for ITNs since 1995 is testimony to the success of NATNETS. So far, a number of key lessons have been learned from 1983 to 2004.
A coordinated Public-Private Alliance
Tanzania has been special in having not only a large number of ITN activities, but also a substantial group of motivated ITN stakeholders. From early on this group has acted in a concerted way through the national ITN Task Force and integrated all ITN activities in the country. As a result, many initiatives could be launched and coordinated to allow a true national scale-up to be considered from 2002 onwards. The role of the MoH has always been very supportive, especially in promoting an enabling environment. Politicians have been targeted regularly and have clearly supported this process. Finally, bilateral and multilateral donors have been generously supporting this process, providing a good model for the funding of large-scale malaria control initiatives. This public-private partnership could serve as a wider example in the health sector, nationally as well as internationally.
Coverage and equity
SMITN and KINET data suggest that affordability remains a significant obstacle to net use, especially for the poorest. Recent data confirm a socio-economically stratified gradient in treated and untreated net ownership and re-treatment rates [35,36], although this gap is narrowing over time. As household coverage rates increase overall, and with the beneficial effects of the TNVS, coverage should rapidly increase in the lower socio-economic quintiles and harder-to-reach rural areas.
The data in Table 3 show coverage figures from a household survey undertaken by NMCP in 14 districts in Tanzania in 2001 and 2003, prior to the TNVS (unpublished data). The rapid growth in coverage in children and pregnant women is apparent. It is certain that the 2004 national coverage figures will be substantially higher, given current ITN sales figures and the launch of the TNVS. A further NMCP household survey is planned for 2005.
Table 3 ITN coverage in 14 districts in Tanzania, 2001 and 2003. Source: National Malaria Control Programme. ITN = insecticide-treated net.
Target group Percentage sleeping under an ITN Percentage sleeping under an untreated net Total (ITN and untreated)
2001 2003 2001 2003 2001 2003
Children under 5 15 26 31 27 41 58
Pregnant women 8 21 28 21 29 49
Net manufacturers
Tanzania is in a unique situation in sub-Saharan Africa by having no less than four domestic mosquito net manufacturers with a total annual net production of over 5 million nets, of which nearly 2 million entered the domestic market in 2004. The Tanzanian manufacturers have made initial progress in expanding the distribution network for nets, but were unlikely to expand further to under-served rural areas on their own. This situation has called for specific strategies to support this development, leading to the launching of the SMARTNET and TNVS programmes.
Removing all taxes on nets and netting did lead to a substantial decrease in retail prices. As a result of lower prices, demand for nets increased dramatically and this allowed the manufacturers to develop the market. This lesson might be of use to other endemic countries which have not yet implemented ITN tax removal, one of the commitments from the Abuja Summit.
Treated versus untreated nets
The benefits of using a treated net compared to an untreated net are not universally understood and valued by consumers. Up to 2002, all ITN projects held jointly less than 30% of the market share, while the remaining 70% were untreated nets sold through commercial channels. Untreated nets have a clear competitive advantage because of the extreme price sensitivity of mosquito control products. Unfortunately, untreated nets provide less than half the public health impact of treated nets [2] and a significant effort is required to ensure that only treated nets are sold. An important step has been made in this direction with the 100% bundling policy agreed by all major net manufacturers in the country and the readiness to engage in the production of LLIN.
Main outstanding challenges
The most obvious current challenge is to make the TNVS operate smoothly at national scale, with operational stability and limited fraud.
Funding for the ITN cell is secured until 2008, that for SMARTNET until June 2007 and that for the TNVS until end 2007. Long-term financing needs to be secured for all components, both from external sources and from the existing MoH resources.
A second challenge will be to bring LLIN as quickly as possible onto the market. Currently, one brand of LLIN is available on the Tanzanian market but its production capacity is insufficient and its retail price is too high to make it widely available (over USD 7). Attention is currently devoted to secure access of the LLIN technology to the Tanzanian manufacturers of polyester nets.
Conclusion
Tanzania has been among the leading African countries in the development and promotion of ITNs. The Ministry of Health has realised and appreciated the role of ITNs in the control of malaria. The country enjoys a unique situation by having effectively stimulated a strong local net industry and distribution network. There is also a committed team of ITN stakeholders from the public and private sectors who have pioneered the scaling up process in a coordinated way, in the spirit of the Roll Back Malaria Partnership. As a result, different components have been implemented to complement each other. It is expected that the NATNETS initiative will enable Tanzania to become the first large African country to meet the Abuja target for ITN use, and hence reduce substantially and sustainably the morbidity and mortality due to malaria over the coming years.
List of Abbreviations
CMO Chief Medical Officer
DPS Director of Preventive Services
IEC Information, Education, Communication
IHRDC Ifakara Health Research and Development Center
ITNs Insecticide-Treated Nets
KINET Kilombero Net Project
LLIN Long-Lasting Insecticidal Net
MCH Mother and Child Health
MoH Ministry of Health
NATNETS National ITN Strategy
NGO Non-Governmental Organization
NMCP National Malaria Control Programme
PATH Programme for Applied Technology
PSI Population Services International
RBM Roll Back Malaria
SMARTNET Strategic Social Marketing for Expanding the Commercial Market
for ITNs in Tanzania
SMITN Social Marketing for ITNs Project
STI Swiss Tropical Institute
TNVS Tanzania National Voucher Scheme
TPRI Tanzania Pesticides Research Institute
USD United States Dollar
VAT Value-Added Tax
WHO World Health Organization
WHOPES World Health Organization Pesticides Evaluation Scheme
Authors' contributions
Stephen M. Magesa, conceived the idea of writing this publication, was involved in the design, providing intellectual content and the main writing of the manuscript. Christian Lengeler was involved in the design, providing intellectual content and substantial writing. Don deSavigny was involved in the design, providing intellectual content and writing. Jane E. Miller was involved in the design, providing intellectual and factual content and a critical review of the manuscript. Ritha J.A. Njau, was part of the team that conceived the idea of writing this publication, was involved in the design, providing intellectual content and a critical review of the manuscript. Karen Kramer provided some essential factual inputs and reviewed critically the manuscript. Andrew Y. Kitua, conceived the idea of writing this publication, was involved in the design, providing intellectual content and a critical review of the manuscript. Alex Mwita, provided intellectual and factual content and critically reviewed the manuscript
Acknowledgements
We thank all the ITN stakeholders in Tanzania for their concerted effort in pushing forward the scaling-up process, and especially Drs. G. Upunda, A. Mzige and A. Unwin. For all involved it has been a gratifying experience to work in such a well-coordinated and productive way towards a common goal. Special thanks are due to the national ITN Task Force members, the National Malaria Control Programme and the Ministry of Health, Tanzania. The governments of the United Kingdom, the Netherlands, Switzerland, Canada as well as WHO and UNICEF have contributed substantially to this process by supporting a number of important initiatives.
==== Refs
Ministry of Health National Malaria Medium-term Strategic Plan, July 2002–June 2006 2002 Dar es Salaam, Tanzania
Lengeler C Insecticide-treated bednets and curtains for preventing malaria 2004 Cochrane Database Systematic Reviews: Oxford, UK
Lines JD Myamba J Curtis CF Experimental hut trials of permethrin-impregnated mosquito nets and eave curtains against malaria vectors in Tanzania Med Vet Entomol 1987 1 37 51 2979519
Magesa SM Wilkes TJ Mnzava AE Njunwa KJ Myamba J Kivuyo MD Hill N Lines JD Curtis CF Trial of pyrethroid impregnated bednets in an area of Tanzania holoendemic for malaria. Part 2. Effects on the malaria vector population Acta Trop 1991 49 97 108 1680284 10.1016/0001-706X(91)90057-Q
Curtis CF Myamba J Wilkes TJ Comparison of different insecticides and fabrics for anti-mosquito bednets and curtains Med Vet Entomol 1996 10 1 11 8834736
Maxwell CA Myamba J Njunwa KJ Greenwood BM Curtis CF Comparison of bednets impregnated with different pyrethroids for their impact on mosquitoes and on re-infection with malaria after clearance of pre-existing infections with chlorproguanil-dapsone Trans R Soc Trop Med Hyg 1999 93 4 11 10492776 10.1016/S0035-9203(99)90158-X
Lyimo EO Msuya FH Rwegoshora RT Nicholson EA Mnzava AE Lines JD Curtis CF Trial of pyrethroid impregnated bednets in an area of Tanzania holoendemic for malaria. Part 3. Effects on the prevalence of malaria parasitaemia and fever Acta Trop 1991 49 157 163 1685296 10.1016/0001-706X(91)90034-H
Stich AH Maxwell CA Haji AA Haji DM Machano AY Mussa JK Mattaelli A Haji H Curtis CF Insecticide-impregnated bed nets reduce malaria transmission in rural Zanzibar Trans R Soc Trop Med Hyg 1994 88 150 154 8036655 10.1016/0035-9203(94)90271-2
Premji Z Lubega P Hamisi Y Mchopa E Minjas J Checkley W Shiff C Changes in malaria associated morbidity in children using insecticide treated mosquito nets in the Bagamoyo district of coastal Tanzania Trop Med Parasitol 1995 46 147 153 8533015
Fraser-Hurt N Felger I Edoh D Steiger S Mashaka M Masanja H Smith T Mbena F Beck HP Effect of insecticide-treated bed nets on haemoglobin values, prevalence and multiplicity of infection with Plasmodium falciparum in a randomized controlled trial in Tanzania Trans R Soc Trop Med Hyg 1999 93 47 51 10450426 10.1016/S0035-9203(99)90327-9
Maxwell CA Msuya E Sudi M Njunwa KJ Carneiro IA Curtis CF Effect of community-wide use of insecticide-treated nets for 3–4 years on malarial morbidity in Tanzania Trop Med Int Health 2002 7 1003 1008 12460390 10.1046/j.1365-3156.2002.00966.x
Schellenberg JR Abdulla S Nathan R Mukasa O Marchant TJ Kikumbih N Mushi AK Mponda H Minja H Mshinda H Tanner M Lengeler C Effect of large-scale social marketing of insecticide-treated nets on child survival in rural Tanzania Lancet 2001 357 1241 1247 11418148 10.1016/S0140-6736(00)04404-4
Abdulla S Schellenberg JA Nathan R Mukasa O Marchant T Smith T Tanner M Lengeler C Impact on malaria morbidity of a programme supplying insecticide treated nets in children aged under 2 years in Tanzania: community cross sectional study BMJ 2001 322 270 273 11157527 10.1136/bmj.322.7281.270
Marchant T Schellenberg JA Edgar T Nathan R Abdulla S Mukasa O Mponda H Lengeler C Socially marketed insecticide-treated nets improve malaria and anaemia in pregnancy in southern Tanzania Trop Med Int Health 2002 7 149 158 11841705 10.1046/j.1365-3156.2002.00840.x
Hanson K Kikumbih N Armstrong SJ Mponda H Nathan R Lake S Mills A Tanner M Lengeler C Cost-effectiveness of social marketing of insecticide-treated nets for malaria control in the United Republic of Tanzania Bull World Health Organ 2003 81 269 276 12764493
Njunwa KJ Lines JD Magesa SM Mnzava AE Wilkes TJ Alilio M Kivumbi K Curtis CF Trial of pyrethroid impregnated bednets in an area of Tanzania holoendemic for malaria. Part 1. Operational methods and acceptability Acta Trop 1991 49 87 96 1680283 10.1016/0001-706X(91)90056-P
Makemba AM Winch PJ Makame VM Mehl GL Premji Z Minjas JN Shiff CJ Treatment practices for degedege, a locally recognized febrile illness, and implications for strategies to decrease mortality from severe malaria in Bagamoyo District, Tanzania Trop Med Int Health 1996 1 305 313 8673832
Fraser-Hurt N Lyimo EO Insecticide-treated nets and treatment service: a trial using public and private sector channels in rural United Republic of Tanzania Bull World Health Organ 1998 76 607 615 10191557
Schellenberg JR Abdulla S Minja H Nathan R Mukasa O Marchant T Mponda H Kikumbih N Lyimo E Manchester T Tanner M Lengeler C KINET: a social marketing programme of treated nets and net treatment for malaria control in Tanzania, with evaluation of child health and long-term survival Trans R Soc Trop Med Hyg 1999 93 225 231 10492745 10.1016/S0035-9203(99)90001-9
Hanson K Jones C Social marketing of insecticide treated mosquito nets, Tanzania: End of phase 1 social and economic analysis London 2000
Lengeler C Cattani J deSavigny D (Editors) Net Gain, a new method for preventing malaria deaths 1996 Ottawa and Geneva: IDRC and WHO
Miller JE Jones CO Ndunguru S Curtis V Lines J A new strategy for treating nets. Part 2: users' perceptions of efficacy and washing practices and their implications for insecticide dosage Trop Med Int Health 1999 4 167 174 10223211 10.1046/j.1365-3156.1999.43378.x
Miller JE Buriyo A Karugila A Lines JD A new strategy for treating nets. Part 1: formulation and dosage Trop Med Int Health 1999 4 160 166 10223210 10.1046/j.1365-3156.1999.43377.x
Schellenberg JA Minja H Mponda H Kikumbih N Mushi A Nathan R Abdulla S Mukasa O Marchant TJ Tanner M Lengeler C Re-treatment of mosquito nets with insecticide Trans R Soc Trop Med Hyg 2002 96 368 369 12497970 10.1016/S0035-9203(02)90361-5
Erlanger TE Enayati AA Hemingway J Mshinda H Tami A Lengeler C Field issues related to effectiveness of insecticide-treated nets in Tanzania Med Vet Entomol 2004 18 153 160 15189240 10.1111/j.0269-283X.2004.00491.x
Tami A Mubyazi G Talbert A Mshinda H Duchon S Lengeler C Evaluation of Olyset insecticide-treated nets distributed seven years previously in Tanzania Malar J 2004 3 19 15225349 10.1186/1475-2875-3-19
National Bureau of Statistics accessed March 10th 2005
MARA (Mapping Malaria Risk in Africa/Atlas du Risque de la Malaria en Afrique) accessed March 10th 2005
deSavigny D Changing norms: Going beyond projects and programs. Introductory Keynote Address Proceedings of the 2nd International Conference on Insecticide Treated Nets: October 1999 1999 Dar es Salaam, Tanzania
PriceWaterhouseCoopers Lengeler C Mponda H National strategic plan for insecticide treated nets in Tanzania 2000 Dar es Salaam
Ministry of Health National Health Accounts, Draft Final Report 2001 Government of Tanzania. Dar es Salaam
RBM Scaling up insecticide-treated netting programmes in Africa; a strategic framework for coordinated national action 2002 Technical Support Network for Insecticide-Treated Materials. Geneva, Switzerland
National Malaria Control Programme Hati Punguzo – Tanzania national voucher scheme 2003 Dar es Salaam, Tanzania
Mushi AK Schellenberg JR Mponda H Lengeler C Targeted subsidy for malaria control with treated nets using a discount voucher system in Tanzania Health Policy Plan 2003 18 163 171 12740321 10.1093/heapol/czg021
Hanson K Worrall E SMITN 2, Tanzania: End-of-project household survey analysis 2002 London, Malaria Consortium
Nathan R Masanja H Mshinda H Schellenberg JA de Savigny D Lengeler C Tanner M Victora CG Armstrong Schellenberg JRM Mosquito nets and the poor: can social marketing redress inequities in access? Trop Med Int Health 2004 9 1121 1126 15482406 10.1111/j.1365-3156.2004.01309.x
|
16042780
|
PMC1190210
|
CC BY
|
2021-01-04 16:24:13
|
no
|
Malar J. 2005 Jul 22; 4:34
|
utf-8
|
Malar J
| 2,005 |
10.1186/1475-2875-4-34
|
oa_comm
|
==== Front
Microb Cell FactMicrobial Cell Factories1475-2859BioMed Central London 1475-2859-4-221604276910.1186/1475-2859-4-22ResearchAn encoded N-terminal extension results in low levels of heterologous protein production in Escherichia coli Orchard Samantha S [email protected] Heidi [email protected] Department of Bacteriology, University of Wisconsin, Madison, WI 53706, USA2 Department of Biology, San Diego State University, San Diego, CA 92182 USA2005 21 7 2005 4 22 22 14 6 2005 21 7 2005 Copyright © 2005 Orchard and Goodrich-Blair; licensee BioMed Central Ltd.2005Orchard and Goodrich-Blair; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The tdk gene (encoding deoxythymidine kinase) of the gamma-proteobacterium Xenorhabdus nematophila has two potential translation start sites. The promoter-distal start site was predicted to be functional based on amino acid sequence alignment with closely related Tdk proteins. However, to experimentally determine if either of the two possible start codons allows production of a functional Tdk, we expressed the "long-form" (using the promoter-proximal start codon) and "short-form" (using the promoter-distal start codon) X. nematophila tdk genes from the T7 promoter of the pET-28a(+) vector. We assessed Tdk production and activity using a functional assay in an Escherichia coli tdk mutant, which, since it lacks functional Tdk, is able to grow in 5-fluorodeoxyuridine (FUdR)-containing medium.
Results
Short-form Tdk complemented the E. coli tdk mutant strain, resulting in FUdR sensitivity of the strain. However, the E. coli tdk mutant expressing the long form of tdk remained FUdR resistant, indicating it did not have a functional deoxythymidine kinase enzyme. We report that long-form Tdk is at least 13-fold less abundant than short-form Tdk, the limited protein produced was as stable as short-form Tdk and the long-form transcript was 1.7-fold less abundant than short-form transcript. Additionally, we report that the long-form extension was sufficient to decrease heterologous production of a different X. nematophila protein, NilC.
Conclusion
We conclude that the difference in the FUdR growth phenotype between the E. coli tdk mutant carrying the long-or short-form X. nematophila tdk is due to a difference in Tdk levels. The lower long-form protein level does not result from protein instability, but instead from reduced transcript levels possibly combined with reduced translation efficiency. Because the observed effect of the encoded N-terminal extension is not specific to Tdk production and can be overcome with induction of gene expression, these results may have particular relevance to researchers attempting to limit production of toxic proteins under non-inducing conditions.
==== Body
Background
Proteins from one organism are often expressed in a different species for the purpose of protein purification or complementation studies. When such efforts fail due to non-production of the protein, the underlying cause of failure is often unclear [1]. Protein overproduction is known to induce a heat shock-like response, which results in increased proteolysis in the cell and therefore possible degradation of the desired protein [2]. Other factors such as degradation of the RNA transcript, efficiency of translation and toxic nature of the desired protein may also influence the level of protein production. Thus, common Escherichia coli strains used for protein overproduction include protease mutants (e.g. BL21; lon), RNase mutants (e.g. BL21 Star™; RNaseE mutant strain; Invitrogen, Carlsbad, CA) and those that provide tRNA synthetases corresponding to infrequently used codons (e.g. Rosetta™; Novagen, Madison, WI) to increase translational efficiency. To reduce production of potentially toxic proteins at inappropriate points in the growth of the host strain, heterologous genes are often fused to engineered promoters that limit gene expression under non-inducing conditions. However, all promoters have a degree of "leakiness" and allow some protein production even under non-inducing conditions.
The predicted Xenorhabdus nematophila Tdk protein is 70% identical to E. coli Tdk and has been shown to have deoxythymidine kinase activity [3], converting salvaged deoxythymidine to deoxythymidine monophosphate [4]. A translational start site was predicted for X. nematophila tdk based on alignment with tdk sequences from other organisms. However, X. nematophila tdk has an additional potential start codon 12 bp 5' from the predicted start site. As part of an effort to establish which start codon is used for native X. nematophila Tdk synthesis, the two forms were expressed from a heterologous promoter in E. coli. These studies revealed that, in contrast to short-form, long-form Tdk is not expressed. Furthermore, the additional four codons present in long-form Tdk are sufficient to decrease production of another unrelated protein, NilC.
Results and discussion
Long-form Tdk does not confer FUdR sensitivity to an E. coli tdk strain
To determine if both short-and long-form (with four extra amino acids) Tdk protein are functional, they were expressed in an E. coli tdk mutant strain (KY895) and their activity assessed in an FUdR growth assay. In an environment lacking in salvageable deoxythymidine and containing FUdR, cells with active Tdk are starved for thymidylate due to inhibition by Tdk-phosphorylated FUdR (FdUMP) of the endogenous synthesis pathway and thus cannot grow (Tdk+ = FUdRs). pETSTdk, the plasmid used to express short-form tdk, complemented the tdk defect in KY895, as demonstrated by the resulting FUdR sensitivity of the strain (Fig. 1; triangles). The FUdR sensitivity of the strain carrying plasmid pETSTdk was similar to that of a strain carrying wild-type E. coli tdk (data not shown). Thus, short-form Tdk appears to be expressed and active in this system. In contrast, KY895 carrying plasmid pETLTdk, designed to express long-form tdk, had an extended lag phase during growth in FUdR punctuated by a reproducible brief increase then decrease in optical density (Fig. 1; circles). This strain ultimately exhibited the same optical density (FUdR resistance) as the strain carrying the control vectors (Fig. 1; squares) and grew similarly to KY895 carrying plasmid pETSTdk or pET28a(+) in Luria-Bertani broth (LB [11]; data not shown). Two possible explanations for this lack of tdk complementation in FUdR-containing medium are that long-form Tdk was not adequately produced in this system or that the long-form protein was not active. Further experiments were aimed at distinguishing between these two possibilities.
Figure 1 Growth of E. coli KY895 carrying plasmids pTara and pET-28a(+) (squares), pETSTdk (triangles) or pETLTdk (circles) in FUdR-containing medium. Each data point represents the average optical density reading for six independent cultures for each strain. Error bars represent standard error of the mean.
Long-form Tdk is not detected by immunoblot analysis
Immunoblot analysis with polyclonal anti-Tdk serum of overnight cultures of E. coli carrying plasmids pTara (encoding arabinose-inducible T7 polymerase) and pETSTdk or pETLTdk failed to reveal long-form Tdk even though short-form Tdk was readily detected (Fig. 2A). These results indicate that the inability of pETLTdk to complement the tdk mutation in E. coli for growth in FUdR (above) was likely due to non-production of long-form Tdk and not necessarily lack of activity in long-form Tdk. Since the only difference between the long-and short-form proteins is the set of additional amino acid residues (MDGP) at the N-terminus, we concluded that these residues or the DNA or mRNA encoding them minimize tdk expression.
Figure 2 Immunoblot analysis of protein production using anti-Tdk serum (panels A and C-G) or anti-NilC serum (panel B). E. coli KY895 (panels A, B, F, G), BL21 Star™ (DE3) (panel C), KY895 clpP (panel D), or BL21 (DE3) (panel E) carrying plasmids pTara and pET-28a(+) (vector-only; lane 1), pETSTdk or pETSNilC (short-form; lane 2), pETLTdk or pETLNilC (long-form; lane 3), pETMAGPTdk or pETMAGPNilC (D2A change; lane 4), pETMDAPTdk or pETMDAPNilC (G3A change; lane 5) or pETMDGATdk or pETMDGANilC (P4A change; lane 6) was grown overnight in LB/kan/cam and separated cellular proteins were subjected to immunoblot analysis with the indicated antiserum. Arabinose (0.2%) was added to the culture medium for the strains in panels F and G, and IPTG (0.2 mM) was added to the culture medium for the strains in panel G. Similar volumes of cultures were used in each lane but cultures were not normalized for optical density or overall protein concentrations.
Long-form NilC-trunc is not detected by immunoblot analysis
To determine if the 5' extension of long-form tdk would reduce production of a different protein, the 12-nt extension was added to an X. nematophila gene, nilC, encoding an outer membrane-associated protein [5] and for which there were readily available antibodies. To create a cytoplasmic version of NilC, the DNA encoding the 21 amino acid long NilC N-terminal signal sequence was removed creating NilC-trunc (short form), to which the MDGP extension was added (long form). Long-form NilC-trunc was not detected by immunoblot analysis with polyclonal anti-NilC serum while short-form NilC-trunc was (Fig. 2B). Thus, the MDGP-encoding extension is sufficient to reduce production of at least two unrelated proteins.
Long-form tdk transcript is less abundant than short-form transcript but shares the same 5' end
Primer extension analysis was used to determine if the MDGP-encoding sequence affects the transcription (either level or start site) of tdk from pET-28a(+) or processing of the tdk transcript. Primer extension analysis indicated that long-form tdk transcript was made and at least some portion maintained its 5' end (Fig. 3). Relative transcript levels were quantified by real-time PCR with primers specific for a central portion of the tdk gene. For growth in LB broth to OD600~0.85, the long-form transcript was 1.7-fold less abundant that the short-form transcript. This lower level of long-form tdk transcript could be due to higher levels of long-form tdk mRNA degradation, reduced transcription of long-versus short-form tdk or to differences in plasmid copy number, possibly a result of the presence of long-form tdk or long-form Tdk production. To determine if this lower transcript abundance is due to increased susceptibility to degradation, protein production was monitored in an E. coli RNaseE mutant. RNaseE is one of the main E. coli mRNA-degrading enzymes [6] and we hypothesized that it differentially degrades long-versus short-form tdk transcript. In the RNaseE mutant, long-form Tdk was detected at low levels by immunoblot analysis (Fig. 2C). Thus in the absence of RNaseE, long-form tdk transcript was sufficiently produced and stable to act as a template for protein synthesis, indicating that RNaseE degrades long-form tdk transcript. However, even in the absence of RNaseE, long-form Tdk protein levels were ~4.7-fold lower than those of short-form Tdk produced under the same conditions. Therefore, RNaseE-mediated degradation of long-form tdk transcript is not the only factor contributing to the low levels of long-form Tdk protein production relative to short-form Tdk.
Figure 3 Primer extension analysis of short-and long-form X. nematophila tdk transcripts in E. coli. Total RNA was isolated from E. coli KY895 carrying plasmids pTara and pETSTdk or pETLTdk grown overnight to OD600 = 2.0. Primer tdkprext2 was end-labeled with 32P and used to produce a sequencing ladder from plasmids pETSTdk (left side; lanes C, T, A, G) and pETLTdk (right side; lanes C, T, A, G). A primer extension product was produced with end-labeled tdkprext2, total RNA template (left side: KY895 carrying pETSTdk; right side: KY895 carrying pETLTdk) and AMV-RT enzyme at 37°C (lanes "pe"). Arrows indicate the expected transcriptional start site from the T7 promoter of pET-28a(+).
Long-form Tdk protein is stable
The stabilities of short-and long-form Tdk proteins were measured to determine if differential susceptibility to post-translational proteolysis is the cause of the low protein levels of long-versus short-form Tdk. Measurement of long-form Tdk stability was made possible by inducing tdk expression by addition of arabinose to the culture medium (arabinose induces expression of pTara-encoded T7 RNA polymerase, which is responsible for transcription from the pET-28a(+) promoter controlling tdk expression). Pulse-chase labeling of cells and immunoprecipitation experiments revealed that long-form Tdk protein is present at a low level with such induction (Fig. 4), although this level was still 13-fold lower on average than that of short-form Tdk and was insufficient for detection by immunoblotting (Fig. 2F). Over the 5 min time course of the experiment, long-form Tdk was at least as stable as short-form Tdk (Fig. 4), indicating that the difference in protein levels between these two forms is not due to differential proteolysis. Further support for this hypothesis comes from the fact that long-form Tdk was not detected in E. coli KY895 clpP and E. coli BL21(DE3) (lon) protease mutant strains (Fig. 2D–E).
Figure 4 Quantification of Tdk protein levels and stability. Levels of protein were determined by 35S-methionine labeling of whole cells and subsequent immunoprecipitation with polyclonal anti-Tdk antiserum. E. coli KY895 carrying plasmids pTara and pETSTdk (Short-form Tdk) or pETLTdk (Long-form Tdk) was labeled with 35S-methionine for 3 min followed by a chase with excess unlabeled methionine. Samples were removed for processing at 20 s, 2 min and 5 min following the chase and proteins reactive with anti-Tdk serum were immunoprecipitated and separated as described in Methods. Following autoradiography, intensity of the bands corresponding to labeled Tdk were determined and background corrected against a pET-28a(+) vector-only control. The calculated band intensities are expressed in arbitrary units.
Long-form Tdk is expressed in the presence of IPTG and arabinose
Arabinose was used to induce T7 RNA polymerase for the 35S-methionine labeling study, resulting in detectable long-form Tdk protein levels by autoradiography (Fig. 4) but not immunoblot analysis (Fig. 2F). Neither arabinose nor IPTG (de-repression of the pET-28a(+) T7 promoter) were used for the other protein production studies reported herein. When both IPTG and arabinose were added to cultures of KY895 carrying pTara and pETLTdk, long-form Tdk was detected by immunoblot analysis at levels similar to short-form Tdk in comparably induced cultures (Fig. 2G). Thus, increased transcription of long-form tdk is sufficient to promote high long-form protein levels. These data support the hypothesis that the low level of protein expressed from genes with the MDGP-encoding extension results at least in part from a reduction in mRNA levels. These results further indicate that, in non-inducing conditions in cases when overproduction of a toxic protein is desired, the encoded N-terminal extension described here might be beneficial in reducing "leaky" protein production.
Changing the nucleotide and/or amino acid identity of the long-form extension results in variable protein production
To determine the relative importance of each codon in the extension region of long-form tdk and nilC-trunc to the resulting observed protein levels, we systematically changed the aspartic acid, glycine and proline codons to alanine codons and monitored the resulting effects on protein production. The levels of the altered long-form proteins were assessed by immunoblot analysis following normalization of samples by optical density (Table 2; see also Fig. 2A–B). The only change that caused an increase in the level of both long-form Tdk and NilC-trunc protein production was a glycine (GGG) to alanine (GCG) codon change (Fig. 2A and 2B). An aspartic acid (GAC) to alanine codon (GCC) change did not alter the level of production of long-form NilC-trunc but did allow detection of long-form Tdk (Fig. 2A). Additional changes were made in the long-form extension region to assess the importance of the position and order of the amino acid residues, particularly glycine (Table 2). All extensions that result in low levels of protein production (similar to that of MDGP) include a glycine codon in the third or fourth position, even when the initiating methionine residue is predicted to be cleaved from synthesized proteins (see Table 2 for a list of extensions where the N-terminal f-Met residue is expected to be cleaved by PepM following protein synthesis [7]). Thus, the position of the glycine residue or codon nucleotides appears to impact long-form protein production.
Table 2 X. nematophila tdk plasmid design and resulting Tdk protein levels
Oligonucleotide name Oligonucleotide sequence (5'->3') a Resulting plasmid name N-terminus of engineered proteinb Relative amount of protein expressed (% of short-form) c
NcoITdkshort CCATGGCTCAGCTTTATTTTTAT pETSTdk MAQLY... 100
NcoITdklong CCATGGACGGGCCAATGGCTCAG pETLTdk MDGPMAQLY... 0
NcoIMDGATdk CCATGGACGGGGCAATGGCTCAG pETMDGATdk MDGAMAQLY... 0
NcoIMDPGTdk CCATGGACCCAGGGATGGCTCAGCTTTATTTTTATTATTCTGC pETMDPGTdk MDPGMAQLY... 0
NcoIMDG(GGU)PTdk CCATGGACGGTCCAATGGCTCAG pETMDG(GGU)PTdk MDGPMAQLY... 0
NcoIMAAGTdk CCATGGCCGCCGGGATGGCTCAGCTTTATTTTTATTATTCTG pETMAAGTdk MAAGMAQLY... 0
NcoIMDAGTdk CCATGGACGCCGGGATGGCTCAGCTTTATTTTTATTATTCTG pETMDAGTdk MDAGMAQLY... 0.2
NcoIMAGPTdk CCATGGCCGGGCCAATGGCTCAG pETMAGPTdk MAGPMAQLY... 1
NcoIMGPDTdk CCATGGGGCCAGACATGGCTCAGCTTTATTTTTATTATTCTGC pETMGPDTdk MGPDMAQLY... 13
NcoIMAAAGTdk CCATGGCCGCCGCCGGGATGGCTCAGCTTTATTTTTATTATTCTG pETMAAAGTdk MAAAGMAQLY... 15
NcoIMAPGDTdk CCATGGCCCCAGGGGACATGGCTCAGCTTTATTTTTATTATTCTG pETMAPGDTdk MAPGDMAQLY... 47
NcoIMGDPTdk CCATGGGGGACCCAATGGCTCAGCTTTATTTTTATTATTCTGC pETMGDPTdk MGDPMAQLY... 77
NcoIMGAATdk CCATGGGGGCCGCCATGGCTCAGCTTTATTTTTATTATTCTG pETMGAATdk MGAAMAQLY... 80
NcoIMAGATdk CCATGGCCGGGGCCATGGCTCAGCTTTATTTTTATTATTCTG pETMAGATdk MAGAMAQLY... 88
NcoIMDAPTdk CCATGGACGCGCCAATGGCTCAG pETMDAPTdk MDAPMAQLY... 93
NcoIMAPDGTdk CCATGGCCCCAGACGGGATGGCTCAGCTTTATTTTTATTATTCTG pETMAPDGTdk MAPDGMAQLY... 103
aUnderlined nucleotides indicate engineered restriction sites used in cloning bAn italicized M indicates an N-terminal f-Met residue that, based on the identity of the second amino acid residue, is predicted to be posttranslationally removed by PepM cAmount of Tdk protein detected by immunoblot analysis of E. coli KY895 carrying pTara and the pET-28a(+)-derived plasmid listed and expressed relative to the level of detected short-form Tdk and background corrected against the pET-28a(+) vector-only level.
Conclusion
A version of X. nematophila Tdk protein with four extra amino acids (long-form Tdk) was not detected by immunoblot analysis of cells carrying plasmid pETLTdk even though the cells could express Tdk lacking these extra amino acids (short-form Tdk encoded on pETSTdk) under the same non-inducing growth conditions. Even with arabinose induction of transcription, long-form Tdk protein was present at a level ~13-fold lower than the short-form. The same extension reduced production of another protein, NilC, suggesting the effect of the extension is not specific to Tdk. Long-form Tdk protein was as stable as short-form Tdk protein, and its levels were not affected by the absence of Lon or ClpP proteases. Therefore the difference in long versus short-Tdk protein levels is not likely due to differential proteolysis of the two protein forms. The long-form tdk mRNA transcript was present at a 1.7-fold lower level than short-form tdk transcript and increased levels of transcript are sufficient to overcome low protein levels, supporting the hypothesis that the low level of long-form Tdk production is due in part to low long-form tdk transcript levels. The bacterial N-end rule [8] states that proteins containing N-terminal arginine, lysine, leucine, phenylalanine, tyrosine and tryptophan residues are particularly sensitive to Clp-dependent degradation, reducing their in vivo half-lives. The phenomenon reported herein is distinct from the N-end rule as none of the known destabilizing residues occur in the MDGP N-terminal extension, nor does the extension result in protein degradation by Clp (Fig. 2D). Since the same pET-28a(+)-based expression system was used for both Tdk and NilC-trunc production, the observed phenotype may result from a combination of the MDGP extension and the pET-28a(+) expression system. These results may be relevant to researchers attempting to limit production of potentially toxic proteins from the pET-28a(+) expression plasmid under non-inducing conditions in an E. coli host, while still allowing protein overproduction under inducing conditions.
Methods
Bacterial strains
Strain KY895 ([9]; λ- tdk-1 IN(rrnD-rrnE)1, ilv-276) was obtained from the E. coli Genetic Stock Center, strain BL21 Star™ (DE3) (F-ompT hsdSB (rB- mB-) gal dcm rne131 (DE3)) was obtained from Invitrogen (Carlsbad, CA) and strain BL21 (DE3) (F- ompT hsdSB (rB- mB-) gal dcm (DE3)) was obtained from Novagen (Madison, WI). To construct the E. coli KY895 clpP mutant strain (HGB871), primers clpPtetKWfor and clpPtetKWrev (Table 1) were used to amplify the tetracycline resistance gene from Tn10. The primers have 40 nt of clpP sequence at their 5' ends. The amplified fragment was electroporated into E. coli DY330 (W3110 ΔlacU169 gal490 λc1857 Δ(cro-bioA)) as described elsewhere [10]. Under the conditions used, this strain allowed homologous recombination of the linear PCR product into the DY330 chromosomal clpP locus, replacing the region beginning 2 bp 5' of the clpP start codon and ending 2 bp 3' of the clpP stop codon with a tetracycline resistance marker. P1 transduction was used to move the mutated locus into E. coli strain KY895 using standard techniques and selecting for desired transductants on tetracycline. PCR analysis was used to confirm the insertion.
Table 1 Oligonucleotide primers used in this studya
Oligonucleotide name Sequence (5'->3')b Target c
BamHInilC GGATCCGACGATGCCTTAATGCGACAG nilC
BamHItdk GGATCCCTTAGTGATTTTATACG 3' of tdk
clpPtetKWford ATCGGTACAGCAGGTTTTTTCAATTTTATCCAGGAGACGG
CAAGAGGGTCATTATATTTCG 5' of E. coli clpP/Tn10 (tetr)
clpPtetKWrevd GCCGCCCTGGATAAGTATAGCGGCACAGTTGCGCCTCTGG
GACTCGACATCTTGGTTACCG 3' of E. coli clpP/ Tn10 (tetr)
Ecrecaminfor GAAAGCGGAAATCGAAGGCG E. coli recA
Ecrecaminrev CATCACACCAATTTTCATACGG E. coli recA
EctdkQPfor TTTGGTGCCGGGAAAGTC E. coli tdk
EctdkQPrev CTTGTTGTCTGGTTAAAAACTGG E. coli tdk
NcoIMAGPNilC CCATGGCCGGGCCAGCTAGAGGAGGGGGTTCTCACC nilC
NcoIMDAPNilC CCATGGACGCGCCAGCTAGAGGAGGGGGTTCTCACC nilC
NcoIMDGANilC CCATGGACGGGGCAGCTAGAGGAGGGGGTTCTCACC nilC
NcoIMDGPNilC CCATGGACGGGCCAGCTAGAGGAGGGGGTTCTCACC nilC
NcoINilC CCATGGCTAGAGGAGGGGGTTCTCACC nilC
tdkprext2 AATAAAAATAAAGCTGAGCC tdk
XntdkQPfor2 CTATCCGCTGATGCTTTGTTG tdk
XntdkQPrev2 TACAATTTCACAAAGCTGCTC tdk
aAdditional oligonucleotides used in this study are listed in Table 2
bUnderlined nucleotides indicate engineered restriction sites used in cloning
cOligonucleotide primers are specific for X. nematophila unless noted otherwise; Integrated DNA Technologies (Coralville, IA) synthesized the oligonucleotides used in this study.
dItalicized nucleotides indicate clpP-region sequence
Growth conditions
Cultures were grown in a tube roller at 30°C in Luria-Bertani (LB) broth [11] except for the FUdR growth assays, which were performed in semi-defined medium containing 5-fluorodeoxyuridine ([12]; FUdR obtained from Fisher Scientific, Pittsburgh, PA) at 37°C with shaking in a microplate reader (Molecular Devices, Sunnyvale, CA) as described previously [3]. LB agar (20 g l-1) plates and all liquid media were supplemented when appropriate with kanamycin (kan; 20 μg ml-1), chloramphenicol (cam; 20 μg ml-1), isopropyl-β-D-thiogalactopyranoside (IPTG; 0.2 mM) or arabinose (0.2%). M9 medium was prepared as described elsewhere [13]. Permanent stocks of cultures were stored at -80°C in dark-stored LB broth supplemented with 10% dimethylsulfoxide.
Production of recombinant X. nematophila Tdk forms in E. coli
DNA manipulations and transformation of E. coli were performed using standard protocols ([14] and product literature). To construct plasmids expressing various forms of X. nematophila ATCC19061 tdk [GenBank:AY363171] [15], the BamHItdk primer (Table 1) was combined with each primer listed in Table 2 in a PCR using either X. nematophila (HGB007; laboratory stock of ATCC19061) chromosomal DNA or a cloned copy of X. nematophila tdk as template DNA. Where a proline residue was desired ahead of other tested residues, an extra alanine codon was engineered between the methionine and proline codons to allow the NcoI restriction site (which includes the translational start site) to be used for cloning. The amplified products from each primer set were cloned into plasmid pET-28a(+) (Novagen, Madison, WI) using the engineered NcoI and BamHI sites. The resulting plasmids, listed in Table 2, in addition to plasmid pTara (arabinose-inducible expression of the T7 RNA polymerase; [16]), were transformed into E. coli KY895, BL21 Star™ (DE3), KY895 clpP and BL21 (DE3). As a control, pET-28a(+) with no insert was transformed with plasmid pTara into these strain backgrounds.
Production of truncated X. nematophila NilC (NilC-trunc) forms in E. coli
A portion of X. nematophila nilC was PCR-amplified from X. nematophila genomic DNA using primers NcoINilC or NcoIMDGPNilC and BamHINilC (Table 1), resulting in nilC production starting 3', at codon 22, of its predicted signal sequence-encoding region and with an additional 5' alanine codon (GCU). NcoIMDGPNilC encodes an additional four codons, for the amino acid series MDGP, before the added alanine codon, to create long-form NilC. Primers NcoIMAGPNilC, NcoIMDAPNilC and NcoIMDGANilC (Table 1) were used to encode the additional amino acid series MAGP, MDAP, and MDGA, respectively, before the added alanine codon. Products from the amplification reactions were cloned into pET-28a(+) as described above. The resulting plasmids, pETnilC, pETMDGPNilC, pETMAGPNilC, pETMDAPNilC and pETMDGANilC, were transformed with pTara into KY895 for protein production analysis.
Immunoblot detection of Tdk and NilC
For immunoblot detections, samples from overnight cultures of KY895 carrying pTara and the appropriate pET-28a(+)-derived vector were electrophoresed and transferred to 0.2 μm PVDF membrane (Bio-Rad) using standard protocols. Immunoblots were performed using the ECL Plus Western Blotting Kit (Amersham Biosciences, Piscataway, NJ) and a goat anti-rabbit IgG horse radish peroxidase-conjugated secondary antibody (Pierce, Rockford, IL). Anti-Tdk serum was obtained from a bleed of a New Zealand White rabbit at the University of Wisconsin Laboratory Animal Resources polyclonal antibody facility following a series of injections of purified 6× his-tagged X. nematophila Tdk [17]. Anti-Tdk and anti-NilC sera [5] were used at a final dilution of 1:5,000. Fluorescence was detected on a Storm860 Phosphorimager (Amersham Biosciences, Piscataway, NJ).
Primer extension mapping of short-and long-form transcripts
Total-cell RNA was isolated from KY895 carrying plasmids pTara and pETSTdk or pETLTdk following overnight growth in LB/kan/cam broth to OD600 = 2.0. A PAGE-purified primer, tdkprext2 (Table 1), was end-labeled using T4 Polynucleotide Kinase and [γ32-P]ATP (Perkin-Elmer Biosciences, Wellesley, MA) for 10 min at 37°C and the labeled primer used in cycle sequencing reactions with components of the fmol® DNA Cycle Sequencing System kit (Promega, Madison, WI) and either plasmid pETSTdk or pETLTdk, as indicated in Figure 3. Labeled primer was also hybridized to 5 μg of total-cell RNA by dissociation at 80°C for 10 min, followed by a slow cooling to 37°C. The primer was extended by avian myeloblastosis virus reverse transcriptase enzyme (AMV-RT, Promega, Madison, WI) at 37°C and the extension products and completed sequencing reactions were resolved on a 12% SDS-polyacrylamide gel containing 8 M urea. The resulting gel was dried and visualized by exposure to a phosphor screen overnight, which was scanned on a Storm 860 phoshorimager (Amersham Biosciences, Piscataway, NJ), and the data analyzed with ImageQuant software (Amersham Biosciences, Piscataway, NJ).
Quantitative PCR to measure relative transcript levels
E. coli KY895 carrying plasmids pTara and either pET-28a(+), pETSTdk or pETLTdk was subcultured from an overnight culture and grown to OD600~0.85 in LB broth with 0.2% glucose. Two independent cultures started from individual colonies were used for each strain. Total RNA was isolated using TRIzol (Invitrogen, Carlsbad, CA) and the RNA was DNase treated and used to make cDNA with random hexamer primers (Integrated DNA Technologies, Coralville, IA) and AMV-RT. As a control to detect DNA contamination of the DNased RNA, samples with no added AMV-RT were analyzed by PCR for amplification of E. coli tdk using primers EctdkQPfor and EctdkQPrev (Table 1) and, as expected, none showed product. Reactions for real-time PCR were performed in duplicate in a total volume of 25 μl with iQ™ SYBR® Green Supermix (Bio-Rad, Hercules, CA), cDNA template, appropriate primers and a three-step cycling protocol on a Bio-Rad iCycler and analyzed with Bio-Rad iCycler iQ™ software. The amount of X. nematophila tdk transcript was measured by amplification with XntdkQPfor2 and XntdkQPrev2 primers (Table 1). As a negative control, water was used in place of cDNA template. Cycle threshold results for each sample were adjusted according to E. coli recA levels (amplified with Ecrecaminfor and Ecrecaminrev primers, Table 1, designed from published E. coli recA sequence) and then converted to arbitrary units factoring in a two-fold change in PCR product per cycle.
Pulse labeling and immunoprecipitation
Cells were grown to OD600 = 0.3 in LB/kan/cam, washed and resuspended in M9 containing 0.4% arabinose, 1 μg ml-1 thiamine, kan and cam and incubated for 1 h before being pulse labeled with 20 mCi/ml [35S]-L-methionine (PerkinElmer Life Sciences, Boston, MA) for 3 min at 37°C. Unlabeled L-methionine was added to 300 μM and 50 μl samples were removed and added to 50 μl of a 125 mM Tris-Cl, 4% SDS, pH 6.8 buffer at 20 s, 2 min and 5 min after addition of unlabeled methionine. After immediate freezing in dry ice/ethanol, samples were boiled for 4 min and to each was added 1 ml of immunoprecipitation buffer (50 mM Tris-Cl, 150 mM NaCl, 5 mM EDTA, 1% Triton X-100, pH7.5). The samples were centrifuged and supernatants incubated with 2 μl rabbit anti-Tdk serum with rocking for 1 h at 4°C. The samples were then incubated with Protein A immobilized on sepharose CL-4B (Sigma, St. Louis, MO) for 1 h at 4°C, with rocking. The pelleted beads were washed 3 times in immunoprecipitation buffer containing 0.1% SDS, separated on a 12% acrylamide denaturing gel and the gel was dried and exposed to a phosphor screen and analyzed as for the primer extension reactions.
Authors' contributions
SSO performed the experiments and participated in experimental design and manuscript drafting and editing. HGB participated in experimental design and manuscript drafting and editing. Both authors read and approved the final manuscript.
Acknowledgements
The authors gratefully acknowledge Charles E. Cowles for providing anti-NilC antibody. This work was supported by National Institutes of Health RO1 grant GM59776, the Investigators in Pathogenesis of Infectious Disease Award from the Burroughs Wellcome Foundation and by USDA/CREES grant CRHF-0-6055 (awarded to HGB and used to support SSO in part). SSO received additional support through the National Institutes of Health Predoctoral Training Grant T32 GM07215 in Molecular Biosciences and through the National Science Foundation Graduate Teaching Fellows in K-12 Education award DUE-9979628 to the K-Through-Infinity program.
==== Refs
Sati SP Singh SK Kumar N Sharma A Extra terminal residues have a profound effect on the folding and solubility of a Plasmodium falciparum sexual stage-specific protein over-expressed in Escherichia coli Eur J Biochem 2002 269 5259 5263 12392558 10.1046/j.1432-1033.2002.03237.x
Gottesman S Minimizing proteolysis in Escherichia coli: genetic solutions Methods Enzymol 1990 185 119 129 2199774
Orchard SS Goodrich-Blair H Pyrimidine nucleoside salvage confers an advantage to Xenorhabdus nematophila in its host interactions Appl and Environ Microbiol
Neuhard J Munch-Peterson A Utilization of preformed pyrimidine bases and nucleosides Metabolism of Nucleotides, Nucleosides and Nucleobases in Microorganisms 1983 London, Academic press 95 148
Miller JH Experiments in Molecular Genetics 1972 Cold Spring Harbor, NY, Cold Spring Harbor Laboratory Press 466
Cowles CE Goodrich-Blair H Characterization of a lipoprotein, NilC, required by Xenorhabdus nematophila for mutualism with its nematode host Mol Microbiol 2004 54 464 477 15469517 10.1111/j.1365-2958.2004.04271.x
Grunberg-Manago M Messenger RNA stability and its role in control of gene expression in bacteria and phages Annu Rev Gen 1999 33 193 227 10.1146/annurev.genet.33.1.193
Sherman F Stewart JW Tsunasawa S Methionine or not methionine at the beginning of a protein BioEssays 1985 3 27 31 3024631 10.1002/bies.950030108
Tobias JW Shrader TE Rocap G Varshavsky A The N-end rule in bacteria Science 1991 254 1374 1377 1962196
Hiraga S Igarashi K Yura T A deoxythymidine kinase-deficient mutant of Escherichia coli. I. Isolation and some properties Biochim Biophys Acta 1967 145 41 51 4228236
Yu D Ellis HM Lee EC Jenkins NA Copeland NG Court DL An efficient recombination system for chromosome engineering in Escherichia coli Proc Natl Acad Sci USA 2000 97 5978 5983 10811905 10.1073/pnas.100127597
Summers WC Raksin P A method for selection of mutations at the tdk locus in Escherichia coli J Bacteriol 1993 175 6049 6051 8376351
Sambrook J Fritsch EF Maniatis T Molecular Cloning: a Laboratory Manual 1989 2nd Cold Spring Harbor, NY, Cold Spring Harbor Laboratory Press 1076
Ausubel FA Brent R Kingston RE Moore DD Seidman JG Smith JA Struhl K Current Protocols in Molecular Biology 1998 New York, Wiley and Sons
Orchard SS Goodrich-Blair H Identification and functional characterization of a Xenorhabdus nematophila oligopeptide permease Appl Env Microbiol 2004 70 5621 5627 15345451 10.1128/AEM.70.9.5621-5627.2004
Wycuff DR Matthews KS Generation of an AraC-araBAD promoter-regulated T7 expression system Anal Biochem 2000 277 67 73 10610690 10.1006/abio.1999.4385
Orchard SS Oligopeptide and Nucleoside Salvage by the Bacterium Xenorhabdus nematophila Bacteriology 2005 Madison, University of Wisconsin
|
16042769
|
PMC1190211
|
CC BY
|
2021-01-04 16:24:35
|
no
|
Microb Cell Fact. 2005 Jul 21; 4:22
|
utf-8
|
Microb Cell Fact
| 2,005 |
10.1186/1475-2859-4-22
|
oa_comm
|
==== Front
J NanobiotechnologyJournal of Nanobiotechnology1477-3155BioMed Central London 1477-3155-3-61598751610.1186/1477-3155-3-6ResearchInteraction of silver nanoparticles with HIV-1 Elechiguerra Jose Luis [email protected] Justin L [email protected] Jose R [email protected] Alejandra [email protected] Xiaoxia [email protected] Humberto H [email protected] Miguel Jose [email protected] Department of Chemical Engineering, The University of Texas at Austin, Austin, Texas 78712, USA2 Texas Materials Institute, The University of Texas at Austin, Austin, Texas 78712, USA3 Facultad de Ciencias Biológicas, Universidad Autónoma de Nuevo León, San Nicolás de los Garza, Nuevo León, Mexico2005 29 6 2005 3 6 6 28 3 2005 29 6 2005 Copyright © 2005 Elechiguerra et al; licensee BioMed Central Ltd.2005Elechiguerra et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The interaction of nanoparticles with biomolecules and microorganisms is an expanding field of research. Within this field, an area that has been largely unexplored is the interaction of metal nanoparticles with viruses. In this work, we demonstrate that silver nanoparticles undergo a size-dependent interaction with HIV-1, with nanoparticles exclusively in the range of 1–10 nm attached to the virus. The regular spatial arrangement of the attached nanoparticles, the center-to-center distance between nanoparticles, and the fact that the exposed sulfur-bearing residues of the glycoprotein knobs would be attractive sites for nanoparticle interaction suggest that silver nanoparticles interact with the HIV-1 virus via preferential binding to the gp120 glycoprotein knobs. Due to this interaction, silver nanoparticles inhibit the virus from binding to host cells, as demonstrated in vitro.
==== Body
Background
Nanotechnology provides the ability to engineer the properties of materials by controlling their size, and this has driven research toward a multitude of potential uses for nanomaterials[1]. In the biological sciences, many applications for metal nanoparticles are being explored, including biosensors[2], labels for cells and biomolecules[3], and cancer therapeutics[4].
It has been demonstrated that, in the case of noble-metal nanocrystals, the electromagnetic, optical and catalytic properties are highly influenced by shape and size [5-7]. This has driven the development of synthesis routes that allow a better control of morphology and size [8-13]. Noble-metal nanomaterials have been synthesized using a variety of methods, including hard-template[14], bio-reduction[9] and solution phase syntheses[8,10-13].
Among noble-metal nanomaterials, silver nanoparticles have received considerable attention due to their attractive physicochemical properties. The surface plasmon resonance and large effective scattering cross section of individual silver nanoparticles make them ideal candidates for molecular labeling[15], where phenomena such as surface enhance Raman scattering (SERS) can be exploited. In addition, the strong toxicity that silver exhibits in various chemical forms to a wide range of microorganisms is very well known [16-18], and silver nanoparticles have recently been shown to be a promising antimicrobial material[19].
For these reasons, and based upon our previous work regarding interactions of noble metal nanoparticles with biomolecules[20], we decided to study the interaction of silver nanoparticles with viruses. Herein, we present the first findings of our investigation, the discovery that silver nanoparticles undergo size-dependent interaction with HIV-1.
Findings
Characterization of the tested silver nanoparticle preparations
The physicochemical properties of nanoparticles are strongly dependent upon their interactions with capping agent molecules[21]. Indeed, the surface chemistry of the nanoparticles can modify their interactions with external systems. For this reason we tested silver nanoparticles with three markedly different surface chemistries: foamy carbon, poly (N-vinyl-2-pyrrolidone) (PVP), and bovine serum albumin (BSA).
Foamy carbon-coated nanoparticles were obtained from Nanotechnologies, Inc., and used without further treatment. These nanoparticles are embedded in a foamy carbon matrix which prevents coalescence during their synthesis. The as-received nanoparticle sample consists of a fine black powder. For the purposes of the present work, the as-received powder was dispersed in deionized water by ultra-sonication. TEM analysis shows that the nanoparticles tend to be agglomerated inside the foamy carbon matrix, although a significant fraction of the population is released from this matrix by the energy provided from the ultra-sonic bath (Figure 1a–1f). These released nanoparticles are mainly free-surface nanoparticles, and it was observed that only nanoparticles that have escaped from the foamy carbon matrix interact with the HIV-1 cells.
Figure 1 Transmission electron microscopy (TEM) of the foamy carbon-coated silver nanoparticles. a) TEM image of the sample prepared by dispersing the as-received powder in deionized water by ultra-sonication. The agglomeration of particles inside the foamy carbon matrix is observed. b) TEM image of nanoparticles outside of the carbon matrix. The broad distribution of shapes can be observed. c)-f) TEM images of nanoparticles with different morphologies. c) Icosahedral. d) Decahedral. e) Elongated. f) Octahedral. g) High Resolution TEM image of the foamy carbon matrix.
The interaction of the nanoparticles with the foamy carbon matrix is sufficiently weak that simply by condensing the TEM electron beam, even those nanoparticles that were not initially released by ultra-sonication are ejected from the foamy carbon agglomeration. In fact, after this experiment the complete size distribution of these nanoparticles is better observed, please refer to Additional file: 1. High resolution transmission electron microscopy (TEM) revealed that the silver nanoparticles released from the foamy carbon matrix by ultrasonication have a size distribution of 16.19 ± 8.68 nm (Figure 2a–b). By releasing the remaining nanoparticles from the foamy carbon matrix with the action of the electron beam, the average size was ~21 ± 18 nm. Additionally, TEM examination demonstrated that the sample is composed of several morphologies including multi-twinned nanoparticles with five-fold symmetry, i.e. decahedra and icosahedra, truncated pyramids, octahedral and cuboctahedral nanoparticles, among others (Figure 1c–1f).
Figure 2 Silver nanoparticle preparations. a) TEM image of free surface silver nanoparticles released from the foamy carbon matrix by dispersing the as-received powder in deionized water by ultra-sonication. b) Size distribution of free surface nanoparticles measured by TEM analysis. c) UV-Visible spectrum of carbon-coated silver nanoparticles. d) HAADF image of PVP-coated silver nanoparticles. e) Size distribution of PVP-coated nanoparticles measured by TEM analysis. f) UV-Visible spectrum of PVP-coated silver nanoparticles. g) HAADF image of BSA-coated silver nanoparticles. h) Size distribution of BSA-coated nanoparticles measured by TEM analysis. i) UV-Visible spectrum of BSA-coated silver nanoparticles.
PVP-coated nanoparticles were synthesized by the polyol method using glycerine as both reducing agent and solvent. In this method, a metal precursor is dissolved in a liquid polyol in the presence of a capping agent such as PVP[22]. PVP is a linear polymer and stabilizes the nanoparticle surface via bonding with the pyrrolidone ring. Infrared (IR) and X-ray photoelectron spectroscopy (XPS) studies have revealed that both oxygen and nitrogen atoms of the pyrrolidone ring can promote the adsorption of PVP chains onto the surface of silver[23]. The sample size distribution was obtained from high angle annular dark field (HAADF) images. The nanoparticles exhibited an average size of 6.53 nm with a standard deviation of 2.41 nm. (Figure 2d–e)
Silver nanoparticles directly conjugated to BSA protein molecules were synthesized in aqueous solution. Serum albumin is a globular protein, and is the most-abundant protein in blood plasma. Bovine serum albumin (BSA) is a single polypeptide chain composed of 583 amino acid residues [24]. Several residues of BSA have sulfur-, oxygen-, and nitrogen-bearing groups that can stabilize the nanoparticle surface. The strongest interactions with silver likely involve the 35 thiol-bearing cysteine residues. By using sodium borohydride, a strong reducing agent, BSA stabilizes nanoparticles via direct bonding with these thiol-bearing cysteine residues, and provides steric protection due to the bulkiness of the protein. The sample size distribution was obtained from HAADF images. Nearly 75% of the BSA-conjugated silver nanoparticles were 2.08 ± 0.42 nm in diameter, but a substantial fraction of larger particles was also observed, bringing the total average to 3.12 ± 2.00 nm (Figure 2g–h).
UV-visible spectroscopy is a valuable tool for structural characterization of silver nanoparticles. It is well known that the optical absorption spectra of metal nanoparticles are dominated by surface plasmon resonances (SPR), which shift to longer wavelengths with increasing particle size [25]. Also, it is well recognized that the absorbance of silver nanoparticles depends mainly upon size and shape [26,27]. In general, the number of SPR peaks decrease as the symmetry of the nanoparticle increases [27]. Recently, Schultz and coworkers[28] correlated the absorption spectra of individual silver nanoparticles with their size (40–120 nm) and shape (spheres, decahedrons, triangular truncated pyramids and platelets) determined by TEM. They found that spherical and roughly spherical nanoparticles, decahedral or pentagonal nanoparticles, and triangular truncated pyramids and platelets absorb in the blue, green and red part of the spectrum, respectively. In all the cases the SPR peak wavelength increases with size, as expected.
The UV-Visible spectra for all the nanoparticle preparations are shown in Figure 2. All samples presented a minimum at ~320 nm that corresponds to the wavelength at which the real and imaginary parts of the dielectric function of silver almost vanish [27]. The sample with carbon-coated silver nanoparticles exhibits four peaks at ~400, ~490, ~560 and ~680 nm, as shown in Figure 2c. The optical signature of this sample can be better understood in terms of the distribution of sizes and shapes observed in the TEM. As we previously mentioned, the distribution of shapes in the sample is broad, and a significant amount of nanoparticles are not spherical such as multi-twinned with five-fold symmetries. The presence of nanoparticles with pentagonal and triangular cross-sections could be responsible for the absorption at longer wavelengths. Thus, it is clear that the characteristic absorption of these nanoparticles arises from the contribution of different shapes and sizes, which agrees with the TEM observations.
On the other hand, the PVP-coated and BSA-coated silver nanoparticles present only one peak at ~450 and ~390 nm, respectively. These results indicate that both preparations are mainly composed by small spherical silver nanoparticles. It is also well know that for small particles a broadening of the plasmon absorption bands is expected, since there is a linear dependence of the full-width at half maximum (FWHM) with the reciprocal of the particle diameter[29]. The results for BSA-coated nanoparticles agree with the last statement, presenting just one broad symmetric peak at ~390 nm. On the other hand, the spectrum for the PVP-coated silver nanoparticles is not symmetric around the maximum absorption wavelength. In fact, this spectrum can be deconvoluted in two different curves, one centered at ~430 and another one at ~520 nm. The peak at ~430 nm could be assigned to the out-of-plane dipole resonance of the silver nanoparticles indicating the presence of spherical particles with small diameters. In addition, the synthesis of silver nanoparticles by the polyol method in presence of PVP promotes also the formation of multi-twinned nanoparticles (MTPs), being decahedra nanoparticles the most thermodynamically stable MTPs [23]. Therefore, the observed read shift is a consequence of both nanoparticles of larger size and the presence of decahedral nanoparticles with pentagonal cross sections.
Interaction with HIV-1
High angle annular dark field (HAADF) scanning transmission electron microscopy was employed to study the interaction of silver nanoparticles with HIV-1. In our previous works, HAADF has proven to be a powerful technique for analysis of biological samples, such as proteins[20] and bacteria[30], interfaced with inorganic nanoparticles. HAADF images are primarily formed by electrons that have undergone Rutherford backscattering. As a result, image contrast is related to differences in atomic number [31] with intensity varying as ~Z2. Therefore, image contrast is strongly related to composition. As a good approximation, lighter elements appear dark and heavier elements appear bright. Due to a large difference in atomic number, silver nanoparticles are easily distinguished from the organic matter that composes the virus.
In Figure 3, we present HAADF images of the HIV-1 virus with (3a) and without (3b) silver nanoparticles. For complete experimental details, refer to Methods Section. The presence of silver was independently confirmed by Energy Dispersive X-ray Spectroscopy (EDS), shown in Figure 3c. Interestingly, the sizes of nanoparticles bound to the virus (Figure 3d) were exclusively within the range of 1–10 nm. In the case of the silver nanoparticles released from the carbon matrix, the fact that no nanoparticles greater than 10 nm in diameter were observed to interact with the virus is significant, since the size of ~40% of the overall population is beyond this range. This provides strong evidence for the size-dependence of interaction.
Figure 3 HAADF images of the HIV-1 virus. a) HAADF image of an HIV-1 virus exposed to BSA-conjugated silver nanoparticles. Inset shows the regular spatial arrangement between groups of three nanoparticles. b) HAADF image of HIV-1 viruses without silver nanoparticle treatment. Inset highlight the regular spatial arrangement observed on the surface of the untreated HIV-1 virus. c) EDS analysis of image a) confirming the presence of Ag. The C signal comes from both the TEM grid and the virus, O, and P are from the virus, and Na, Cl, and K are present in the culture medium. Ni and Si come from the TEM grid, while Cu is attributed to the sample holder. d) Composite size distribution of silver nanoparticles bound to the HIV-1 virus, derived from all tested preparations.
Additionally, the nanoparticles seen in Figure 3a are not randomly attached to the virus, as regular spatial relationships are observed among groups of three particles. Both the spatial arrangement of nanoparticles and the size dependence of interaction can be explained in terms of the HIV-1 viral envelope, and can provide insight into the mode of interaction between the virus and nanoparticles.
The exterior of the HIV-1 virus is comprised of a lipid membrane interspersed with protruding glycoprotein knobs, formed by trimers consisting of two subunits: the gp120 surface glycoprotein subunit is exposed to the exterior, and the gp41 transmembrane glycoprotein subunit spans the viral membrane and connects the exterior gp120 glycoprotein with the interior p17 matrix protein[32]. The main function of these protruding gp120 glycoprotein knobs is to bind with CD4 receptor sites on host cells. Numerous cellular proteins are also embedded within the viral envelope[33]. However, the protruding gp120 glycoprotein knobs are more exposed to the exterior, and should be more accessible for potential nanoparticle interactions.
Leonard and coworkers[34] reported that the gp120 subunit has nine disulfide bonds, three of which are located in the vicinity of the CD4 binding domain. These exposed disulfide bonds would be the most attractive sites for nanoparticles to interact with the virus. As mentioned previously, the nanoparticles in Figure 1a appear to be located at specific positions, with regular spatial relationships observed among groups of three particles. The observed spatial arrangements correlate with the positions of the gp120 glycoprotein knobs in the structural model for HIV-1 proposed by Nermut and coworkers[32].
Regular spatial relationships are also found on the surface of the untreated virus, as seen in the inset of Figure 1b. The observed darker contrast at these sites could indicate the locations of the glycoprotein knobs. As mentioned previously, contrast in HAADF images is strongly dependent on differences in atomic number. However, this is not the only factor in determining the image contrast. If the material is composed of elements of similar atomic numbers, as is the case for the organic constituents of the pure virus, local variations in sample density will provide noticeable contrast. The majority of the viral envelope consists of a densely-packed lipid membrane. However, for the glycoprotein knobs, we would expect a localized region of lower density due to the presence of membrane-spanning gp41 glycoproteins rather than the densely-packed lipids. Hence, these areas should appear darker than the rest of the viral envelope.
It has previously been determined that the center-to-center spacing between glycoprotein knobs is ~22 nm[35]. In the inset of Figure 3a, the average measured center-to-center spacing between silver nanoparticles is ~28 nm, which correlates with the expected spacing between glycoprotein knobs. The average center-to-center spacing between the small darker regions on the untreated virus is ~22 nm, which again suggests these sites are the gp120 glycoprotein knobs. Thus, the observed spatial arrangement of nanoparticles, the center-to-center distance between nanoparticles, and the fact that the exposed sulfur-bearing residues of the glycoprotein knobs would be attractive sites for nanoparticle interaction strongly suggest that silver nanoparticles interact with the HIV-1 virus via preferential binding to the gp120 glycoprotein knobs.
Presuming that the most attractive sites for interaction are the sulfur-bearing residues of the gp120 glycoprotein knobs, there are only a limited number of bonds that a nanoparticle can form. This limited number of stabilizing sites can explain why larger nanoparticles are not observed to attach to the virus. Assuming that each nanoparticle interacts with a single glycoprotein knob, and that each nearest-neighbor knob is occupied by another nanoparticle, from geometric considerations the theoretical upper limit for the diameter for these nanoparticles would be ~20 nm. However, if a nanoparticle larger than the diameter of one knob (~14 nm[35]) were to be attached, only a small fraction of the total nanoparticle surface would be anchored, resulting in a less stable interaction. Thus, if the nanoparticles are interacting with HIV-1 via preferential binding at gp120 glycoprotein knobs, we would expect to find mostly nanoparticles less than 14 nm in diameter, as particles in this size range would have the most stable surface interactions. This corresponds closely with our experimental observation that particles greater than 10 nm were not attached to the viral envelope.
Although the mechanism by which HIV infects host cells is not yet fully understood, there are two steps that are broadly agreed to be critical. The first step involves binding of gp120 to the CD4 receptor site on the host cell. Then, upon binding to CD4, a conformational change is induced in gp120, resulting in exposure of new binding sites for a chemokine receptor, i.e. CCR5 or CXCR4 [36-38]. An agent that preferentially interacts with the gp120 glycoprotein would block the virus from binding with host cells. Therefore, we measured the inhibitory effects of silver nanoparticles against HIV-1 in vitro.
The capacity of silver nanoparticles to inhibit HIV-1 infectivity was determined by testing against CD4+ MT-2 cells and cMAGI HIV-1 reporter cells. For complete experimental details, refer to Methods Section. The cytopathic effects of CD4+ MT-2 infection were analysed by optical microscopy assessment of syncytium formation as described elsewhere[39,40], as well as by the HIV-1 infection of cMAGI cells using the Blue Cell Assay[41,42]. The cytotoxicity of all the nanoparticle preparations against MT-2 cells was determined using the Trypan Blue exclusion assay [43]. For all three nanoparticle preparations, at silver concentrations above 25 μg/mL, viral infectivity was reduced to an extent that it could not be detected by syncytium formation, as shown graphically in Figure 4. For each nanoparticle preparation, we found a dose-dependant inhibition of HIV-1 infectivity, with an IC50 where only moderate cell toxicity was observed, as seen in Figure 4.
Figure 4 Inhibition of HIV-1 and toxicity data. a) Assessment of HIV-1 mediated syncytium formation in MT-2 cells. b) Percentage of HIV-1 transmission in cMAGI cells. The toxicity of the nanoparticle preparations against MT-2 cells was determined using the Trypan Blue exclusion assay. The samples were incubated at 37°C, and the cells were evaluated via optical microscopy after c) 3 h and d) 24 h of exposure to silver nanoparticles.
Although the findings regarding interaction with HIV-1 were congruent among nanoparticles with markedly different surface chemistry, the toxicity and inhibition results were not the same. The differences in the observed trends in HIV-1 inhibition can be explained in terms of the capping agents employed for each nanoparticle preparation. BSA- and PVP-protected nanoparticles exhibit slightly lower inhibition because the nanoparticle surface is directly bound to and encapsulated by the capping agent. In contrast, the silver nanoparticles released from the carbon matrix have a greater inhibitory effect due to their essentially free surface area. The fact that the carbon-coated nanoparticles present higher cytotoxicity can also be explained in terms of surface chemistry. Since a significant amount of these silver nanoparticles possess nearly free surfaces, they are able to interact stronger with the host cells, thus increasing their toxicity. Clearly, selection of capping agents will be crucial for future research on the interaction of nanoparticles with viruses, microorganisms, and more complex biosystems, and many more variables require further testing, including the long-term effects of the presence of nanoparticles, and the impact of traces of precursor molecules and reaction by-products.
In conclusion, we have found that silver nanoparticles undergo size-dependent interaction with HIV-1, and that the bound particles exhibit regular spatial relationships. These observations lead us to suggest that the nanoparticles undergo preferential binding with the gp120 subunit of the viral envelope glycoprotein. Silver nanoparticles inhibit the HIV-1 virus infectivity in vitro, which also supports our proposal regarding preferential interaction with gp120. These findings only provide indirect evidence for our proposed mode of interaction, and we are currently undertaking testing to determine conclusively if direct conjugation between gp120 and silver nanoparticles exist.
The interactions of inorganic nanoparticles with biosystems are just beginning to be understood, and potential applications are being discovered at an increasing rate. However, in order to realize the future promise of nanoscience, it is imperative that the toxicity and long-term health effects of exposure to nanomaterials be fully explored. The flexibility of nanoparticle preparation methods, the multitude of functionalization techniques, and facile incorporation of nanoparticles into a variety of media provide the incentive for further research on the interaction of metal nanoparticles with viruses.
Methods
a) HIV-1 strains and cell lines
HIV-1IIIB laboratory strain of HIV-1 an X4 wild type (wt) virus was obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH. CD4+ MT-2 cell line was obtained from the American Type Culture Collection. The cMAGI HIV-1 reporter cells were a generous gift from Dr. Phalguni Gupta from the University of Pittsburgh. All other reagents used were of the highest quality available.
cMAGI cells were cultured in DMEM Dulbecco's Modified Eagle Medium (DMEM) (1X) liquid without sodium phosphate and sodium pyruvate. The medium contained 4,500 mg/L D-glucose and L-glutamine (Invitrogen, Paisley, UK), with 10% fetal calf serum (FCS), 0.2 mg/mL geneticin (G418), and 0.1 μg/mL puromycin. MT-2 cells were cultured in RPMI 1640, containing 10% fetal calf serum (FCS) and antibiotics.
HIV-1IIIB primary clinical isolates were propagated by subculture in MT-2 and cMAGI cells. HIV-1IIIB was reproduced according to the DAIDS Virology Manual for HIV Laboratories, version 1997, compiled by Division of AIDS of the National Institute of Allergies and Infectious Diseases and the National Institute of Health, and Collaborators. Aliquots of cell-free culture viral supernatants were used as viral inocula.
All the work related to HIV-1 cells, except for TEM analysis, was done in a Biosafety Level 3 (BSL-3) Laboratory.
b) Synthesis of the three different silver nanoparticle preparations
Carbon coated silver nanoparticles tested in this study were obtained from Nanotechnologies, Inc. and used without further treatment. For more information about the synthesis of these nanoparticles, please visit
PVP-coated silver nanoparticles were synthesized by the polyol method using glycerine as both reducing agent and solvent. Silver sulfate (Ag2SO4, reagent grade) and poly (N-vinyl-2-pyrrolidone) (PVP-K30, MW = 40,000) were purchased from Sigma Aldrich and 1,2,3-Propanetriol (Glycerin, >99%) was purchased from Fischer Chemicals, all the materials were used without any further treatment. Briefly, we added 0.2 g of PVP to a round bottom flask following by the addition of 30 mL of glycerin. Once PVP was dissolved, we increased the temperature to 140°C. After 30 minutes we added 2 mL of 0.015 M Ag2SO4 and left to react for 1 h.
Silver nanoparticles directly conjugated to bovine serum albumin (BSA) protein molecules were produced as following described. Silver nitrate (AgNO3, 0.945 M), sodium borohydride (NaBH4, 99%) and 200 proof spectrophotometric-grade ethanol were purchased from Aldrich. Bovine serum albumin (BSA) was purchased from Fisher and was used without further treatment. Briefly, sodium borohydride was added to an aqueous solution of silver nitrate and BSA under vigorous stirring. The molar ratio of Ag+:BSA was 28:1, and the molar ratio of Ag+:BH4- was 1:1. The reaction volume was 40 mL, and contained 13.50 μmol BSA. The reaction was allowed to proceed for 1 h, and the product was purified by precipitation at -5°C, followed by cold ethanol filtration.
c) Characterization of the different silver nanoparticle preparations
Transmission electron microscope was conducted in a HRTEM JEOL 2010F microscope equipped with Schottky-type field emission gun, ultra-high resolution pole piece (Cs = 0.5 mm), and a scanning transmission electron microscope (STEM) unit with high angle annular dark field (HAADF) detector operating at 200 kV. Briefly, a droplet of each different solution of silver nanoparticles was placed on a Cu grid with lacey carbon film (Ted Pella), and allowed to evaporate. Size distributions for each nanoparticle preparation were obtained from TEM analysis based on the measurement of 400 particles, and 600 particles in the case of BSA-coated nanoparticles.
UV-visible spectra were obtained at room temperature using a 10 mm path length quartz cuvette in a Cary 5000 spectrometer. All the solutions were diluted 30 × in deionized water before acquiring the spectra.
d) Electron microscopy of HIV-1 and silver nanoparticles
Samples were prepared for electron microscopy as follows: 105 TCID50 samples of HIV-1IIIB cell free virus were treated with solutions of the different silver nanoparticles at a concentration of 100 μg/mL. After 30 seconds, a 10 μL droplet was deposit on a carbon coated nickel TEM grid and exposed to a 2.5% solution of PBS/glutaraldehyde vapors for 30 minutes. Microscopy was done using a JEOL 2010-F TEM equipped with an Oxford EDS unit, at an accelerating voltage of 200 kV and operated in scanning mode using an HAADF detector.
e) Inhibition of HIV-1 with silver nanoparticles
RPMI medium only or containing varying concentrations of silver nanoparticles were mixed with samples 105 TCID50 of HIV-1IIIB cell free virus. The highest concentration of silver nanoparticles used was 100 μg/mL. After 30 seconds, sequential 2-fold dilutions of the solutions were added to cultures of target cells (2 × 105 MT-2 and 2 × 105 cMAGI HIV-1 reporter cells with 0.2–0.5 multiplicity of infection (m.o.i) of HIV-1IIIB virus) prepared as previously mentioned. Each dilution was exposed to four replicate wells. After that, the cells were incubated in a 5% CO2 humidified incubator at 37°C for 3–5 days. Assessment of HIV-1 mediated syncytium formation was performed for the MT-2 cells, while for cMAGI cells, the percentage of transmission was estimated as follows: the number of blue-stained cells obtained from the supernatant of each of the tested wells was divided by the number of blue-stained cells obtained from the culture supernatant in the well of the positive control.
f) Cytotoxicity of silver nanoparticles against MT-2 cells
The cytotoxicity of the nanoparticle preparations against MT-2 cells was determined using the Trypan Blue exclusion assay. In all cases, the initial concentration of silver nanoparticles was 50 μg/mL and sequential two-fold dilutions were made and mixed with 2 × 105 MT-2 cells. The samples were incubated at 37°C, and the cells were evaluated via optical microscopy after 3 h and 24 h of exposure to silver nanoparticles. Briefly, an aliquot of the cell suspension was diluted 1:1 (v/v) with 0.4% Trypan Blue and the cells were counted using a haemocytometer. Viability was expressed as the percentage of number of unstained treated cells to that of the total number of cells.
Supplementary Material
Additional File 1
Supporting information. The file is a word document that contains the complete size distribution of the carbon-coated silver nanoparticles evaluated by TEM
Click here for file
Acknowledgements
The authors want to thank Nanotechnologies, Inc. for providing their silver nanoparticles. J. L. Elechiguerra, J. R. Morones, and A. Camacho-Bragado acknowledge the support received from CONACYT- México. J.L.Burt thanks the University of Texas at Austin College of Engineering and Mr. Robert L. Mitchell for their financial support through the Thrust 2000 Robert L. and Jane G. Mitchell Endowed Graduate Fellowship in Engineering. This material is based upon work supported under a National Science Foundation Graduate Research Fellowship (to JLB).
==== Refs
Bonnemann H Richards RM Nanoscopic metal particles- Synthetic methods and potential applications Eur J Inorg Chem 2001 10 2455 2480 10.1002/1099-0682(200109)2001:10<2455::AID-EJIC2455>3.3.CO;2-Q
Nam JM Thaxton CS Mirkin CA Nanoparticle-based bio-bar codes for the ultrasensitive detection of proteins Science 2003 301 1884 1886 14512622 10.1126/science.1088755
Tkachenko AG Xie H Coleman D Glomm W Ryan J Anderson MF Franzen S Feldheim DL Multifunctional Gold Nanoparticle-Peptide Complexes for Nuclear Targeting J Am Chem Soc 2003 125 4700 4701 12696875 10.1021/ja0296935
Hirsch LR Stafford RJ Bankson JA Sershen SR Rivera B Price RE Hazle JD Halas NJ West JL Nanoshell-mediated near-infrared thermal therapy of tumors under magnetic resonance guidance PNAS 2003 100 13549 13554 14597719 10.1073/pnas.2232479100
Liz-Marzan LM Nanometals: Formation and color Materials Today 2004 7 26 31 10.1016/S1369-7021(04)00080-X
Mulvaney P Surface Plasmon Spectroscopy of Nanosized Metal Particles Langmuir 1996 12 788 800 10.1021/la9502711
Burda C Chen X Narayanan R El-Sayed MA Chemistry and Properties of Nanocrystals of Different Shapes Chemical Reviews (Washington, DC, United States) 2005 105 1025 1102
Yu YY Chang SS Lee CL Wang CRC Gold nanorods: electrochemical synthesis and optical properties Journal of Physical Chemistry B 1997 101 6661 6664 10.1021/jp971656q
Canizal G Ascencio JA Gardea-Torresday J Jose-Yacaman M Multiple twinned gold nanorods grown by bio-reduction techniques Journal of Nanoparticle Research 2001 3 475 481 10.1023/A:1012578821566
Jana NR Gearheart L Murphy CJ Seed-mediated growth approach for shape-controlled synthesis of spheroidal and rod-like gold nanoparticles using a surfactant template Advanced Materials (Weinheim, Germany) 2001 13 1389 1393 10.1002/1521-4095(200109)13:18<1389::AID-ADMA1389>3.0.CO;2-F
Jana NR Gearheart L Murphy CJ Wet chemical synthesis of high aspect ratio cylindrical gold nanorods Journal of Physical Chemistry B 2001 105 4065 4067 10.1021/jp0107964
Sun Y Mayers B Herricks T Xia Y Polyol Synthesis of Uniform Silver Nanowires: A Plausible Growth Mechanism and the Supporting Evidence Nano Letters 2003 3 955 960 10.1021/nl034312m
Lisiecki I Filankembo A Sack-Kongehl H Weiss K Pileni MP Urban J Structural investigations of copper nanorods by high-resolution TEM Physical Review B: Condensed Matter and Materials Physics 2000 61 4968 4974
Zhou Y Yu SH Cui XP Wang CY Chen ZY Formation of Silver Nanowires by a Novel Solid-Liquid Phase Arc Discharge Method Chemistry of Materials 1999 11 545 546 10.1021/cm981122h
Schultz S Smith DR Mock JJ Schultz DA Single-target molecule detection with nonbleaching multicolor optical immunolabels Proceedings of the National Academy of Sciences of the United States of America FIELD Publication Date:2000 Feb 1 97 996 1001. FIELD Reference Number: FIELD Journal Code:7505876 FIELD Call Number: 10.1073/pnas.97.3.996
Liau SY Read DC Pugh WJ Furr JR Russell AD Interaction of silver nitrate with readily identifiable groups: ralationship to the antibacterial action of silver ions Lett Appl Microbiol 1997 25 279 283 9351278 10.1046/j.1472-765X.1997.00219.x
Gupta A Silver S Silver as biocide:Will resistance become a problem? Nat Biotechnol 1998 16 888 9788326 10.1038/nbt1098-888
Nomiya K Yoshizawa A Tsukagoshi K Kasuga NC Hirakawa S Watanabe J Synthesis and structural characterization of silver (I), aluminium (III) and cobalt(II) complexes with 4-isopropyltropolone (hinokitiol) showing noteworthy biological activities. Action of silver(I)-oxygen bonding complexes on the antimicrobial activities J Inorg Biochem 2004 98 46 60 14659632 10.1016/j.jinorgbio.2003.07.002
Sondi I Salopek-Sondi B Silver nanoparticles as antimicrobial agent: A case study on E. coli as a model for Gram-negative bacteria J Colloid Interface Sci 2004 275 177 182 15158396 10.1016/j.jcis.2004.02.012
Burt JL Gutierrez-Wing C Miki-Yoshida M Jose-Yacaman MJ Noble-Metal Nanoparticles Directly Conjugated to Globular Proteins Langmuir 2004 20 11778 11783 15595811 10.1021/la048287r
Bradley JS Schmid GE Clusters and Colloids: From Theory to Applications Clusters and Colloids: From Theory to Applications 1994 Weinheim, VCH 459 536
Bonet F Guery C Guyomard D Urbina RH Tekaia-Elhsissen K Tarascon JM Electrochemical reduction of noble metal compounds in ethylene glycol International Journal of Inorganic Materials 1999 1 47 51 10.1016/S1463-0176(99)00007-1
Wiley B Sun Y Mayers B Xia Y Shape-controlled synthesis of metal nanostructures: The case of silver Chemistry - A European Journal 2005 11 454 463 10.1002/chem.200400927
Peters TJ All About Albumin: Biochemistry, Genetics, and Medical Applications All About Albumin: Biochemistry, Genetics, and Medical Applications 1996 San Diego, Academic Press 9 75
Brause R Moeltgen H Kleinermanns K Characterization of laser-ablated and chemically reduced silver colloids in aqueous solution by UV/VIS spectroscopy and STM/SEM microscopy Applied Physics B: Lasers and Optics 2002 75 711 716 10.1007/s00340-002-1024-3
Kerker M The optics of colloidal silver: something old and something new Journal of Colloid and Interface Science 1985 105 297 314 10.1016/0021-9797(85)90304-2
Sosa IO Noguez C Barrera RG Optical Properties of Metal Nanoparticles with Arbitrary Shapes Journal of Physical Chemistry B 2003 107 6269 6275 10.1021/jp0274076
Mock JJ Barbic M Smith DR Schultz DA Schultz S Shape effects in plasmon resonance of individual colloidal silver nanoparticles Journal of Chemical Physics 2002 116 6755 6759 10.1063/1.1462610
Petit C Lixon P Pileni MP In situ synthesis of silver nanocluster in AOT reverse micelles Journal of Physical Chemistry 1993 97 12974 12983 10.1021/j100151a054
Sweeney RY Mao C Gao X Burt JL Belcher AM Georgiou G Iverson BL Bacterial Biosynthesis of Cadmium Sulfide Nanocrystals Chemistry & Biology 2004 11 1553 1559 15556006 10.1016/j.chembiol.2004.08.022
James EM Browning ND Practical aspects of atomic resolution imaging and analysis in STEM Ultramicroscopy 1999 78 125 139 10.1016/S0304-3991(99)00018-2
Forster MJ Mulloy B Nermut MV Molecular modelling study of HIV p17gag (MA) protein shell utilising data from electron microscopy and X-ray crystallography J Mol Biol 2000 298 841 857 10801353 10.1006/jmbi.2000.3715
Arthur LO Bess JW Sowder RC Benveniste R Mann D Chermann J Henderson L Cellular proteins bound to immunodeficiency viruses: implications for pathogenesis and vaccines Science 1992 258 1935 1938 1470916
Leonard CK Spellman MW Riddle L Harris RJ Thomas JN Gregory TJ Assignment of intrachain disulfide bonds and characterization of potential glycosylation sites of the type 1 recombinant human immunodeficiency virus envelope glycoprotein (gp120) expressed in Chinese hamster ovary cells J Biol Chem 1990 265 10373 10382 2355006
Gelderblom HR Hausmann EHS Ozel M Pauli G Koch MA Fine structure of human immunodeficiency virus (HIV) and immunolocalization of structural proteins Virology 1987 156 171 176 3643678 10.1016/0042-6822(87)90449-1
Dalgleish AG Beverley PCL Clapham PR Crawford DH Greaves MF Weiss RA The CD4 (T4) antigen is an essential component of the receptor for the AIDS retrovirus Nature 1984 312 763 767 6096719 10.1038/312763a0
Klatzmann D Champagne E Chamaret S Gruest J Guetard D Hercend T Gluckman JC Montagnier L T-lymphocyte T4 molecule behaves as the receptor for human retrovirus LAV Nature 1984 312 767 768 6083454 10.1038/312767a0
Feng Y Broder CC Kennedy PE Berger E HIV-1 entry cofactor: Functional cDNA cloning of a seven-transmembrane, G protein-coupled receptor. Science 1996 272 872 877 8629022
Harada S Koyanagi Y Yamamoto N Infection of HTLV-III/LAV in HTLV-I-carrying cells MT-2 and MT-4 and application in a plaque assay Science 1985 229 563 566 2992081
Sodroski J Goh WC Rosen C Campbell K Haseltine WA Role of the HTLV-III/LAV envelope in syncytium formation and cytopathicity Nature 1986 322 470 474 3016552 10.1038/322470a0
Deng H Liu R Ellmeier W Choe S Unutmaz D Burkhart M Di Marzio P Marmon S Sutton RE Hill CM Davis CB Peiper SC Schall TJ Littman DR Landau NR Identification of a major co-receptor for primary isolates of HIV-1 Nature 1996 381 661 8649511 10.1038/381661a0
Chackerian B Long EM Luciw PA Overbaugh J Human immunodeficiency virus type 1 coreceptors participate in postentry stages in the virus replication cycle and function in simian immunodeficiency virus infection J Virol 1997 71 3932 3939 9094670
Kaltenbach JP Kaltenbach MH Lyons WB Nigrosin as a dye for differentiating live and dead ascites cells*1 Exp Cell Res 1958 15 112 117 13574164 10.1016/0014-4827(58)90067-3
|
15987516
|
PMC1190212
|
CC BY
|
2021-01-04 16:38:05
|
no
|
J Nanobiotechnology. 2005 Jun 29; 3:6
|
utf-8
|
J Nanobiotechnology
| 2,005 |
10.1186/1477-3155-3-6
|
oa_comm
|
==== Front
Part Fibre ToxicolParticle and Fibre Toxicology1743-8977BioMed Central London 1743-8977-2-41610518410.1186/1743-8977-2-4ResearchCytokine release from alveolar macrophages exposed to ambient particulate matter: Heterogeneity in relation to size, city and season Hetland Ragna B [email protected] Flemming R [email protected]åg Marit [email protected] Magne [email protected] Erik [email protected] Per E [email protected] Division of Environmental Medicine, Norwegian Institute of Public Health, P.O. Box 4404 Nydalen, N-0403 Oslo, Norway2 Centre for Environmental Health Research, National Institute for Public Health and the Environment, P.O. Box 1, NL-3720 BA Bilthoven, the Netherlands2005 17 8 2005 2 4 4 13 4 2005 17 8 2005 Copyright © 2005 Hetland et al; licensee BioMed Central Ltd.2005Hetland et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Several studies have demonstrated an association between exposure to ambient particulate matter (PM) and respiratory and cardiovascular diseases. Inflammation seems to play an important role in the observed health effects. However, the predominant particle component(s) that drives the inflammation is still not fully clarified. In this study representative coarse (2.5–10 μm) and fine (0.1–2.5 μm) particulate samples from a western, an eastern, a northern and a southern European city (Amsterdam, Lodz, Oslo and Rome) were collected during three seasons (spring, summer and winter). All fractions were investigated with respect to cytokine-inducing potential in primary macrophages isolated from rat lung. The results were related to the physical and chemical parameters of the samples in order to disclose possible connections between inflammatory potential and specific characteristics of the particles.
Results
Compared on a gram-by gram basis, both site-specific and seasonal variations in the PM-induced cytokine responses were demonstrated. The samples collected in the eastern (Lodz) and southern (Rome) cities appeared to be the most potent. Seasonal variation was most obvious with the samples from Lodz, with the highest responses induced by the spring and summer samples. The site-specific or seasonal variation in cytokine release could not be attributed to variations in any of the chemical parameters. Coarse fractions from all cities were more potent to induce the inflammatory cytokines interleukin-6 and tumour necrosis factor-α than the corresponding fine fractions. Higher levels of specific elements such as iron and copper, some polycyclic aromatic hydrocarbons (PAHs) and endotoxin/lipopolysaccaride seemed to be prevalent in the coarse fractions. However, variations in the content of these components did not reflect the variation in cytokine release induced by the different coarse fractions. Addition of polymyxin B did not affect the particle-induced cytokine release, indicating that the variations in potency among the coarse fractions are not explained by endootoxin.
Conclusion
The inflammatory potential of ambient PM demonstrated heterogeneity in relation to city and season. The coarse particle fractions were consistently more potent than the respective fine fractions. Though a higher level of some elements, PAH and endotoxin was found in the coarse fractions, the presence of specific components was not sufficient to explain all variations in PM-induced cytokine release.
==== Body
Background
Previous epidemiological studies in Europe have demonstrated a heterogeneity in respiratory and cardiovascular diseases after exposure to ambient particulate matter (PM) [1,2]. Inflammation plays a crucial part in the long-term development of lung diseases, and possibly also cardiovascular diseases induced by particles. Exacerbation of inflammatory responses also seems to be involved in the acute effects triggered by short-term exposure to particles [3-5].
The composition of PM is influenced by emissions from different sources such as traffic, industrial activities, residential heating and long distance transported air pollution. The PM may in certain areas contain considerable amounts of mineral particles generated by road surface abrasion, due to the use of studded tires during winter season. Occasionally, a significant part of the long distance transported PM also consists of mineral particles from the dry southern parts of Europe and Africa. Furthermore, temporal variations in source emissions and/or seasonal variations in temperature and other meteorological conditions may influence the composition of ambient PM. As a result, considerable site-specific and seasonal variations in the physical and chemical characteristics of particulate air pollution may occur.
Particle size is a critical parameter due to differential deposition in the respiratory tract, but also due to differential effects on the lung cells per se. Epidemiological data suggest that fine ambient particles may be more important than coarse in PM-associated mortality and adverse respiratory health effects [6-8]. Conversely, associations between PM and daily mortality in an environment in which PM is dominated by the coarse fraction have also been reported [9,10]. Our previous studies have shown that coarse and fine mineral particles induce differential responses in lung cells [11,12]. Furthermore, several in vivo and in vitro studies have demonstrated larger inflammatory responses to coarse fractions of PM compared to fine [13-15].
Various chemical components may influence the inflammatory potential of ambient particles. With respect to inorganic components, the importance of metals has been demonstrated. The ability of particles collected in the Utah Valley to induce cytokine production correlated with their metal content [16]. Transition metals in the soluble fraction of residual oil fly ash (ROFA) were found to be responsible for increased release of inflammatory cytokines [17,18]. However, such concentrations of metals are rather unusual and cannot explain health outcomes of all epidemiological studies. In other in vitro studies with rat lung macrophages, however, inflammatory responses to PM could not be explained by variations in the concentrations of soluble metals, but seemed attributable to the insoluble components of the particles [19,20]. Organic components of PM may also elicit inflammatory responses [21]. Microbial components bound to the particles, e.g. endotoxin, may contribute to the inflammatory potential of ambient PM. In vitro studies have demonstrated stronger pro-inflammatory effect of the coarse (2.5 – 10 μm) than the fine (< 2.5 μm) fractions of PM10 and attributed these effects to the endotoxin content, even though some of the endotoxin was found in the fine fractions [14,22-24].
Alveolar macrophages and different types of epithelial cells constitute the primary targets of inhaled lung toxicants and are therefore particularly important in the induction of inflammatory responses in the lung. Distinct particle properties involved in the activation of various uptake mechanisms or the triggering of responses by interactions between particles and receptors on the plasma membrane may in turn lead to different cytokine responses. Macrophages release a variety of inflammatory cytokines upon particle exposure, such as tumour necrosis factor (TNF)-α and interleukin (IL)-6 [25]. TNF-α is known to stimulate, among other effects, the epithelial cells to release various cytokines involved in the recruitment and activation of inflammatory cells, and also enhance the response to subsequent treatment with particles [26]. IL-6 serves as a chemoattractant for lymphocytes [27].
This work was performed within the scope of a European Union sponsored project entitled "Respiratory Allergy and Inflammation Due to Ambient Particles" (RAIAP). The overall objective was to assess the role of ambient suspended particles in causing inflammation in the respiratory tract and induction and elicitation of respiratory allergies. Representative samples of ambient particulate matter (PM2.5–10 and PM0.1–2.5) were collected in cities across Europe with expected differences in PM composition and traffic intensity (Amsterdam, Lodz, Oslo and Rome) during spring, summer and winter seasons. In the study presented here, we investigated the ability of the particle samples to induce release of the pro-inflammatory cytokines TNF-α and IL-6 from primary alveolar macrophages isolated from rats. Data from these in vitro studies and data from the particle characterisation studies [28] were then related in order to disclose possible relations between the potential to induce inflammatory markers and the presence of specific components of the various particle samples.
Results and discussion
Particle samples
Particles were sampled in four major cities (Rome, Amsterdam, Lodz and Oslo) representing the southern, western, eastern and northern part of Europe using a high-volume cascade impactor (HVCI) [29]. The cities were selected based on the expected differences in chemical composition rather than on known differences in health status due to PM exposure. In this light it should be mentioned that we did not aim to compare the cities explicitly. Both seasonal and regional differences in PM mass were observed during the collection period (2001 – 2002) [28]. Generally, the highest levels of PM were observed during winter (Lodz and Rome), whereas the lowest levels were observed in the spring (Oslo and Amsterdam). Variations in the relative contribution of pollution sources within, as well as between the cities, may be illustrated by variations in the observed ratio between fine and coarse particles collected during the different seasons (Table 1). In summer, the fine fraction constituted 50 – 60% of the total amount in each city based on the HVCI data. During winter, however, fine particles represented 77% of the total mass in Lodz compared to 51% in Oslo. This higher percentage of fine PM in the winter particulate air pollution in Lodz may be explained by the greatly increased use of fossil fuels for heating. In Oslo, a relatively low amount of fine particles compared to coarse was found during winter (51%) and spring (42%). This may partly be explained by the generation of mineral particles due to extensive road surface abrasion by cars with studded tires and sand sprinkling on icy roads. It should be mentioned that the technique of high-volume sampling is not ideal to determine actual ambient particle concentrations over time and these values might deviate from reference methods.
Table 1 Proportion of fine and coarse fraction of the collected PM estimated for each city and each season. The proportion is expressed as mass of the fine fraction in percent of the total mass (coarse + fine fraction) collected by high-volume cascade impactors in each sampling period.
Spring (percent fine of total) Summer (percent fine of total) Winter (percent fine of total)
Lodz 58 54 77
Rome 53 52 68
Oslo 42 60 51
Amsterdam 55 60 56
Cytokine release in relation to localization and season
Figure 1 displays the effect of increasing concentrations of PM on IL-6 release from alveolar macrophages. The coarse samples collected in Lodz during spring and summer appeared to be the most potent, reaching 440% and 460% increase, respectively, compared to control. The coarse summer fractions collected in Rome and Oslo induced higher levels of IL-6 than the corresponding sample from Amsterdam (370% and 310% versus 190% in maximal increase) (Figure 1, lane A). The coarse fractions of the winter season samples, however, exhibited a different order of potency compared to spring and summer season samples. Both the samples from Rome and Amsterdam induced higher levels of IL-6 than the samples from Lodz and Oslo (340% and 300% versus 165% and 160%, respectively). Generally the fine fractions did not induce any significant release of cytokines, even though a slight dose-dependent increase was observed after exposure to all fine summer samples and the fine winter sample from Lodz (Figure 1, lane B).
Figure 1 Release of IL-6 after exposure of alveolar rat macrophages to increasing concentrations of coarse (lane A) and fine (lane B) fractions of ambient PM collected in Oslo, Rome, Lodz and Amsterdam during spring (upper), summer (middle) and winter (lower) seasons. Values are mean ± SEM (n = 3).
A distinct seasonal variation in potential to induce IL-6 was demonstrated among the coarse fractions collected in Lodz (Figure 2, lane A). A notably lower level of IL-6 was induced by the winter sample compared to the spring and summer samples. In contrast to the Lodz results, the coarse fractions collected during the different seasons in Rome seemed equally potent. The coarse summer fraction from Oslo induced a higher level of IL-6 release compared to the spring and winter samples, though not statistically significant. The Amsterdam coarse fraction collected during winter induced the highest level of IL-6, which is in contrast to the other three cities. However, the differences in Amsterdam did not reach statistical significance. In general, the highest pro-inflammatory potential seemed to be found in particles collected during spring and summer, the seasons with the highest prevalence of allergens.
Figure 2 Seasonal variations in release of IL-6 (lane A) and TNF-α (lane B) from alveolar rat macrophages after exposure to the coarse fractions of ambient PM collected in Oslo, Rome, Lodz and Amsterdam (top to bottom). Values are shown as fold increase compared to control levels (IL-6ctr.: 455 – 477 pg/ml, SEM: 172 – 230; TNF-αctr.: 382 – 679 pg/ml, SEM: 13 – 23). Values are mean ± SEM (n = 3).
Both particle fractions induced a relatively similar pattern of responses with TNF-α as with IL-6, although the responses in Oslo, Rome and Amsterdam were lower than for IL-6. The results from the coarse fractions from each city are presented as fold increase in Figure 2, lane B. The coarse fractions collected during spring in Rome and Lodz demonstrated a significant increase in TNF-α release compared to control. The fine fractions did not induce a marked increase in TNF-α release in any season or city (data not shown). The macrophages used in this study were from healthy animals. A different result may have been demonstrated with primed or activated macrophages. In vitro studies have demonstrated increased levels of PM-induced inflammatory responses in macrophages primed with LPS when compared to unprimed [26]. Furthermore, primed epithelial cells have also been shown to have augmented proinflammatory responses to particles [30]. This may imply that inflammatory potent particles could be even more potent to persons with preexisting inflammation.
Cytokine release in relation to coarse and fine fractions
The cytokine release induced by the coarse fractions was notably higher compared to the cytokine release induced by the fine fractions (Figure 1). In a related study, in which rat type 2 cells were exposed to the coarse and fine fractions of the collected samples, the particle-induced release of MIP-2 demonstrated a relatively similar pattern of responses as in the macrophages [31]. These results are in accordance with other studies who report that the coarse fractions of ambient PM appear to be more potent than the fine fractions to induce inflammatory responses [14,20,24,32]. The authors attributed the greater inflammatory potential of the coarse particles to higher levels of bioactive biological components bound primarily to the coarse fraction. In a study in mice, however, inflammatory responses after exposure to fine and ultrafine fractions of ambient particles were similar or even greater compared to coarse [13]. The very low responses after exposure to the fine fractions observed in our study could be influenced by a certain loss of ultrafine particles due to the mode of operation of the high volume sampler [29,33]. However, we exposed A549 cells to the similarly collected fine particle samples as used with the macrophages and a dose-dependent increase in IL-8 was observed (unpublished results). Similar levels of IL-8 release were also demonstrated when A549 cells were exposed to low concentrations of coarse, fine and ultrafine fractions of urban particles collected with a similar sampler as used in this study [34]. The previous explanation appears therefore less likely. It has, however, been shown that diesel particles may have a suppressive effect on the cytokine release from alveolar macrophages [35]. Thus, if diesel particles constitute a considerable part of the fine particle fractions in our study, a corresponding suppression might be an explanation.
Cytokine release in relation to chemical components
Major differences in chemical composition were observed in the collected RAIAP-samples [28]. The predominance of combustion particles in Lodz is reflected in a high content of zinc, PAH and other organic components compared to the other samples. In contrast, ambient PM in Rome seems to be influenced by long distance transboundary particles from the African continent, resulting in relatively high levels of metals and mineral components. Amsterdam is close to the North Sea, and the ambient PM will therefore contain sea salt, as well as long distance transported particles. Typical for Oslo, especially in winter, is a relatively high level of PAH in addition to inorganic components. The PAH is most likely a result of extensive wood burning for heating. Temporal and spatial variation in chemical composition of ambient PM and a corresponding variation in biological activity have previously been reported by other investigators [20,36-39].
Elements
The elements iron (Fe), aluminium (Al) and copper (Cu) are typical for both crustal material and other sources, whereas zinc (Zn) and vanadium (V) are elements more related to various combustion processes. The relationship between the content of selected elements and release of IL-6 after exposure of the macrophages to non-toxic levels of particles (20 μg/ml) is presented in Figure 3. The coarse fractions were characterised by a higher content of Fe, and to a certain extent also Cu and Al, than their respective fine fractions. Higher levels of these metals may therefore be related to the higher cytokine responses induced by the coarse particles. In contrast, the fine fractions, which induced no or only small cytokine responses, seemed to contain the highest levels of Zn and V. Especially, the fine winter sample collected in Lodz stands out with a much higher content of Zn than any of the other samples.
Figure 3 The concentrations of selected elements in the PM samples presented together with the respective PM-induced IL-6 from alveolar rat macrophages after exposure to 20 μg/ml of coarse (C) and fine (F) fraction collected in Oslo, Rome, Lodz and Amsterdam in spring (upper), summer (middle) and winter (lower) seasons. IL-6 release is shown on the left ordinate, values in diagram are mean ± SEM (n = 3). Concentration of elements is shown on the right ordinate (ng/mg PM). Indicated values in the diagram for Fe and Al should be multiplied by 103, values for Zn and Cu by 102 and values for V by 101.
Several studies have focused on an importance for redox-active transition metals such as Fe and Cu in lung inflammatory responses, partly due to their ability to participate in Fenton chemistry and the production of reactive oxygen species (ROS) [40]. Furthermore, both V and Zn have been suggested to contribute to inflammatory reactions through their effect as inhibitors of protein phosphatases [41]. Aluminium salts, on the other side, have been associated with a decreased inflammatory potential of particles (quartz) by modifying the particle surface [42,43]. With respect to the importance of metals for PM-induced responses, an effect of both soluble and insoluble fractions has been reported [17-19,44,45]. This supports that the metal content may contribute to the observed higher levels of cytokines induced by the coarse fractions in our study. Apparently, the concentrations of V and Zn found in our samples are not high enough to induce any cytokine release in the alveolar macrophages. An exception might be a relation between the high content of Zn in the fine winter fraction from Lodz and the slight dose-dependent increase in IL-6 release (Figure 1, lane B). Another aspect in this study is that the cytokine responses are compared to the total amounts of metals in the particle samples and not the soluble fractions.
Although Fe, Cu and Al seemed to be prevalent in the coarse particle samples and might contribute to the cytokine-releasing potential of the coarse versus the fine fractions, the occurrence of any single element was not sufficient to explain the regional or seasonal heterogeneity between the coarse particles. This is illustrated by the fact that the cytokine release induced by the coarse spring sample from Lodz, in which the content of Fe was relatively low, equalled the response induced by the coarse samples from Rome, the most Fe-rich samples of all. The potential of the different particle samples may therefore be influenced by combinations of various elements or components. An increased inflammatory response has been shown after exposure to combinations of Cu/Zn compared to exposure to the single elements [46].
PAH
A wide range of PAH is associated with PM derived from combustion of materials such as diesel, gasoline, coal and wood. The different size fractions of ambient PM may contain various amounts of these compounds. In Figure 4, the content of subgroups of PAH associated with particles resulting from combustion of carbonaceous material is presented in relation to the observed IL-6 responses. The coarse fractions contained equal or higher levels of some PAH typically found in emission from combustion processes dominated by diesel. One exception was the very high level of these PAH found in the fine winter fraction collected in Lodz. Moreover, the fine fractions demonstrated equal or higher concentration of the other two subgroups of PAH (PAH from gasoline and woodburning and other PAH) than the respective coarse fractions. However, variations in neither the total PAH content nor the PAH content of the subgroups could explain the higher potency of the coarse fractions, or the observed seasonal and geographical differences in IL-6 release. This is best illustrated by the samples from Lodz and Rome, in which the coarse fractions collected in summer, with relatively low PAH content, induced the highest cytokine release. In contrast, exposure to the most PAH rich samples, the coarse and fine winter fractions from Lodz, induced much lower levels of IL-6. It has been shown that organic compounds that induced CYP1A1 expression were critical for the inflammatory response induced by diesel exhaust in airway epithelial cells [21]. In our studies, it should not be excluded, however, that the apparent lack of association between cytokine release and PAH content might be due to low metabolism of PAH to reactive metabolites in the alveolar macrophages studied. This notion is supported by a recent study, in which CYP1A1 protein levels were found very low, and hence the toxicity of the PAH minimal in alveolar rat macrophages [47]. Furthermore, another study indicated that the total soluble organic fractions, rather than specific PAH, were involved in the inflammatory responses [48].
Figure 4 The concentrations of subgroups of PAH in the collected samples presented together with the respective PM-induced IL-6 from alveolar rat macrophages after exposure to 20 μg/ml of coarse (C) and fine (F) fraction collected in Oslo, Rome, Lodz and Amsterdam in spring (upper), summer (middle) and winter (lower) seasons. IL-6 release is shown on the left ordinate, values in the diagram are mean ± SEM (n = 3). Concentration of PAH are shown on the right ordinate (ng/mg PM). Values of the subgroup of PAH representing emissions from combustion of diesel are the summarised concentrations of phenantrene, 1-methylphenantrene, fluoranthene and pyrene in each PM-sample. Values of the subgroup of PAH representing emissions from gasoline engines and wood burning are the summarised concentrations of naphthalene, benz[a]anthracene, chrysene, benzo[b]fluoranthene, benzo[k]fluoranthene and benzo[g,h,i]perylene in each sample. Values of other PAH represent the sum of acenaphthene and benzo[a]pyrene in each sample.
Inorganics
In Figure 5, the relative amounts of some inorganic compounds in relation to the particle-induced release of IL-6 are shown. The coarse fractions contained equal or higher levels of chloride compared to the fine fractions, in which higher levels of ammonium and sulphate were demonstrated. The various levels of inorganic components measured in the PM samples did not seem to be associated with the higher levels of cytokine release induced by the coarse fractions. Furthermore, no relation was revealed between seasonal and geographical variations and content of the inorganics. This is in accordance with conclusions from the current toxicological database regarding health effects and inorganic components of ambient air particles, which concluded that these components had a low potential for health effects [49]. The observed high levels of sulphate in the fine summer samples may in part be due to more warm and humid conditions favourable for more rapid oxidation of sulphur dioxide generated from combustion of fossil fuels.
Figure 5 The concentrations of selected inorganic components in the collected samples presented together with the PM-induced IL-6 from alveolar rat macrophages after exposure to 20 μg/ml of coarse (C) and fine (F) fractions collected in Oslo, Rome, Lodz and Amsterdam in spring (upper), summer (middle) and winter (lower) seasons. IL-6 is shown on the left ordinate, values are mean ± SEM (n = 3). The concentration of inorganics is shown on the right ordinate (μg/mg PM).
Endotoxin
In Figure 6, the endotoxin levels measured by the Limulus amoebocyte lysate (LAL)-test and the Endotoxin Pyrogen Test (EPT) are presented in relation to the particle-induced release of IL-6. The coarse fractions showed a higher content of endotoxin than the fine fractions when related to the results from the LAL-test. This is in accordance with several other studies [14,20,22,24,32]. Also, the coarse summer samples from Oslo, Lodz and Amsterdam contained higher levels of endotoxin than the spring and winter samples when analysed by this method. A higher level of endotoxin in particles collected during the warmer season has also been reported by others [50]. Thus, there may be a trend in our study that the observed higher IL-6 levels induced by the coarse fractions are associated with higher endotoxin levels. However, there is no direct proportionality, as illustrated by the differing results from Lodz and Amsterdam. Spring and summer particles from Lodz induced notably higher cytokine release compared to the particles from Amsterdam, whereas the particles from Amsterdam contained similar or higher levels of endotoxin. The notion that the presence of endotoxin may contribute to the observed higher potential of the coarse fractions than the fine, is to a certain extent supported when endotoxin was analysed by the more quantitative EPT-assay. Higher levels of endotoxin in the coarse than the fine samples were found. Compared to the LA assay, consistently lower levels, as well as less variation between samples were measured by this method. Furthermore, no relation between the seasonal or site-specific variation in cytokine release and the EPT-results on endotoxin content was observed. The discrepancy in endotoxin content measured by these methods may be explained by an overestimation of the levels in the LAL-assay. Other organic components, such as β-glucans, were detected (but not quantified) in PM from all seasons and cities [51]. The presence of these components may have influenced the LAL-result, since glucans, proteins and other agents have demonstrated a capability to activate the LAL-assay [52].
Figure 6 The concentrations of endotoxins in the collected samples presented together with PM-induced IL-6 from alveolar rat macrophages after exposure to 20 μg/ml of coarse (C) and fine (F) fractions collected in Oslo, Rome, Lodz and Amsterdam in spring (upper), summer (middle) and winter (lower) seasons. IL-6 release is shown on the left ordinate, values in diagram are mean ± SEM (n = 3). Concentration of endotoxin is shown on the right ordinate (ng/mg PM). Endotoxin content in PM samples measured by the LAL-method and the EPT-method are shown. LAL-analysis of the Oslo spring samples is missing due to shortage of sample material.
To further study the involvement of endotoxin, the particle samples were treated with the endotoxin-binding protein polymyxin B sulphate (polymyxin) before addition to the cell cultures. Lipopolysaccaride (LPS), an endotoxin from Gram negative bacteria, was included as a positive control. The LPS-binding property of polymyxin was demonstrated by the complete reduction in IL-6 release induced by LPS and a marked reduction in IL-6 induced by the urban standard PM EHC-93 (Figure 7). However, no significant reduction in the cytokine release was observed after polymyxin-treatment of the collected particle samples. This indicated that the observed differences in potential of the coarse fractions to induce IL-6 release could not be attributed to their content of endotoxin. Synergistic interactions have also been suggested for endotoxin and other proinflammatory components [53]. Such possibilities, as well as a role of other constituents of biological origin cannot be excluded for the observed results in our study.
Figure 7 Release of IL-6 from alveolar rat macrophages after exposure to 20 μg/ml of the coarse (C) PM samples collected during the summer season, EHC-93 or LPS with or without treatment of polymyxin B sulphate. Values are mean ± SEM (n = 3).
In summary, the present studies demonstrated that ambient air particles collected in a western, an eastern, a northern and a southern European city differed in their potential to induce pro-inflammatory cytokines in primary macrophages isolated from rat lung. On a gram-by-gram basis significant differences in the levels of IL-6 and TNF-α release were observed after exposure to particle samples collected in the different cities and different seasons. The coarse fractions induced significant higher levels of cytokines than the fine fractions. Though specific metals and endotoxin seem to be prevalent in the coarse particle samples, the observed variations in potency were not related to variations in concentrations of any component in the particle samples. Interaction(s) between constituents in the different particle samples might be a probable explanation for the observed effects. Both interactions between different metals and metals and organic compounds are conceivable. The strongest cytokine responses were also observed of particles collected during spring and summer, the seasons with the highest prevalence of allergens. Given that a preceding or side-by-side non-specific PM-induced inflammation in the airways might influence the susceptibility of persons with respiratory allergies, seasonal heterogeneity in inflammatory potential of ambient PM may influence the onset or severity of allergic responses in the respiratory system.
Conclusion
The cytokine-inducing potential of collected ambient PM varied between sampling-site and -season. Clear differences were also observed between the coarse and the fine PM fractions: the coarse fractions were consistently more potent than the corresponding fine fractions. A simple relationship between the presence of a specific component in the particle samples and their potential to induce cytokines could not be demonstrated.
Methods
Chemicals
Lipopolysaccharide (LPS) and polymyxin B sulphate were obtained from Sigma-Aldrich, St. Louis, MO, USA. Foetal bovine serum (FBS) was obtained from Gibco BRL, Paisley, Scotland. The antibiotics ampicillin and fungizone were purchased from Bristol-Myers Squibb AB, Denmark; penicillin/streptomycin and the culture medium RPMI 1640 were from BIO Whittaker, Walkersville, MD, USA. The enzyme-linked immunosorbent assays (ELISA) for analysis of rat TNF-α and IL-6 were obtained from R&D Systems Europe, Oxon, UK. Ottawa dust (EHC-93) was kindly provided by Dr. Renauld Vincent, Environmental Health Directorate, Health Canada, Ottawa, Ontario, Canada.
Particle sampling
The PM sampling campaign is described in detail in the final project report [49]. In short, ambient air particles were collected in the cities Amsterdam, Lodz, Rome and Oslo during spring, summer and winter 2001/2002. A high-volume cascade impactor with a multi-stage slit nozzle impactor has been used to collect coarse (PM2.5–10) and fine (PM0.1–2.5) fractions on polyurethane foam (PUF) by impaction. The fractions of dry particle samples extracted with methanol from the PUFs were provided for the particle characterisation and biological experiments [28].
Particle characterisation
Characterisation of the collected samples was performed at the National Institute for Public Health and the Environment (RIVM), the Netherlands, and is described in detail by Cassee et al. [28].
Elements and other inorganics
In summary, elemental composition of the particles was analysed with ICPMS. Secondary aerosols were detected with ion-chromatography (Cl, NO3 and SO4) or photometry (NH4), and anions were analysed using a Dionex guard column (AG-4A), Separation column (Dionex AS-A4) and pulsed electrochemical detector (Dionex-PED). The highest content of Fe was found in the coarse fractions when the coarse and fine fractions within each city were compared. The content of Cu in the coarse fractions generally exceeded the content in the fine fractions. In contrast, the highest levels of Zn and V were found in the fine fractions [28].
PAH
PAHs were analysed on a 30 m 0.25 mm WCOT DB-5MS column in a Fisons 8000 series gas chromatograph equipped with an Interscience MD800 mass spectrometer with EI in the SIR mode. The coarse fractions contained equal or higher levels of PAH typically found in emissions from combustion processes dominated by diesel fuelled engines (phenantrene, 1-methylphenantrene, fluoranthene and pyrene) compared to the fine fractions. The fine fractions demonstrated an equal or higher concentration of PAH typically generated by combustion of gasoline, as well as from wood burning (naphthalene, benz[a]anthracene, chrysene, benzo[b]fluoranthene, benzo[k]fluoranthene and benzo[g,h,i]perylene) and also other PAH (acenaphthene and benzo[a]pyrene) when compared to the coarse fractions.
Inorganics
The highest level of NH4 was found in the coarse sample collected in Oslo during spring. In all other samples, NH4 was higher in the fine than in the coarse fractions. A higher level in the fine compared to the coarse fractions was also observed for SO4. Cl was equal or higher in the coarse fractions, whereas the highest levels of NO3 were found in the fine fractions.
Endotoxins
Endotoxin concentrations were determined using the Limulus amoebocyte lysate (LAL) test (Limusate, Sigma-Aldrich Chemie BV, Zwinrecht, The Netherlands), as described by the manufacturer (detection level < 0.125 ng endotoxin/ml). Endotoxin content in the PM suspensions was also determined using the U.S. Pharmacopeia (USP) Endotoxin Pyrogen Test (EPT) (USP 23 NF 18, 1994, U.S. Pharmacopeial Convention, INC., Rockville, MD, USA). Generally, higher levels of endotoxin were found in the coarse fractions compared to the fine fractions within each city.
Preparation of particles for biological studies
The collected dry particle samples were suspended in 0.9% NaCl to a concentration of 20 mg/ml and stirred overnight. The suspensions were further diluted in culture medium to stock solutions of 2 mg/ml. Before use in the exposure studies, the stock solutions were stirred on a magnetic stirrer overnight. Dry particle samples, as well as particle suspensions, were stored at -20°C. In experiments aimed to study if endotoxins were involved in the observed responses, stock solutions of particles were treated with the LPS-binding polymyxin B sulphate (10 μg/ml) for 1 hour before addition to the cells.
Primary rat alveolar macrophages
Male rats (Crl/Wky) were purchased from Harlan, UK. Alveolar macrophages were collected by airway lavage, suspended in RPMI medium with supplements and added to 35 mm 6 well culture dishes (1.5 × 106/well). Non-attached cells were removed after 1 hour, whereas the attached macrophages were used for exposure.
Exposure of cell cultures to particles
Cells were cultured in FBS-free medium from the start of exposure and the subsequent 6 hours, then 5% FBS was added. After 20 hours of exposure in a total of 1 ml/well of culture medium, the medium was collected and centrifuged for 10 minutes to remove cells (250 × g) and to remove particles (2500 × g). Supernatants were stored at -70°C until further analysis of inflammatory cytokines. Responses after exposure to coarse and fine fractions collected within each season were studied in the same experiment and repeated at least three times.
Cytokine assays
Analysis of the inflammatory cytokines (IL-6 and TNF-α) was performed using enzyme-linked immunosorbent assay (ELISA) according to the manufacturer's manual. The increase in colour intensity was quantified using a plate reader with software (TECAN Sunrise with Magellan V 1.10, Tecan Austria, Salzburg, Austria).
Statistical analysis
The data were analysed for significance by One Way Analysis of Variance (ANOVA) (Tukey Test). Kruskal Wallis ANOVA on Ranks was used when Normality or Equal Variance Test failed (Dunn's Method). P < 0.05 was judged to be statistically significant.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
RBH performed the experimental studies, the biochemical and statistical analysis, and prepared the manuscript. ML, MR and PES performed the isolation of the macrophages and contributed to the writing of the manuscript. FRC was responsible for collection and physical-chemical characterisation of the particulate samples and contributed to the writing of the manuscript. ED coordinated the RAIAP project. All authors have read, reviewed, commented and approved the final manuscript.
Acknowledgements
We gratefully acknowledge E. Lilleaas, T. Skuland, H. Hopen and H.J. Dahlman, Norwegian Institute of Public Health, Oslo, Norway, for valuable technical assistance. We thank A.J.F. Boere, D.L.A.C. Leseman and P.H.B. Fokkens, National Institute for Public Health and the Environment, Bilthoven, the Netherlands, for sampling and handling the PM samples. This work was supported by the RAIAP project (Respiratory Allergy and Inflammation due to Ambient Particles) – A European Commission Shared-Cost Research Project, QLK-CT-2000-00792.
==== Refs
Atkinson RW Anderson HR Sunyer J Ayres JG Baccini M Vonk JM Boumghar A Forastiere F Forsberg B Touloumi G Schwartz J Katsouyanni K Acute effects of particulate air pollution on respiratory admissions. Results from APHEA 2 project Am J Respir Crit Care Med 2001 164 1860 1866 11734437
Le Tertre A Medina S Samoli E Forsberg B Michelozzi P Boumghar A Vonk JM Bellini A Atkinson R Ayres JG Sunyer J Schwartz J Katsouyanni K Short-term effects of particulate air pollution on cardiovascular diseases in eight European cities J Epidemiol Community Health 2002 56 773 779 12239204 10.1136/jech.56.10.773
Gong HJ Linn WS Sioutas C Terrell SL Clark KW Anderson KR Terrell LL Controlled exposures of healthy and asthmatic volunteers to concentrated ambient fine particles in Los Angeles Inhal Toxicol 2003 15 305 325 12635001 10.1080/08958370304455
Stenfors N Nordenhall C Salvi SS Mudway I Soderberg M Blomberg A Helleday R Levin JO Holgate ST Kelly FJ Frew AJ Sandstrom T Different airway inflammatory responses in asthmatic and healthy humans exposed to diesel Eur Respir J 2004 23 82 86 14738236 10.1183/09031936.03.00004603
Pope CA IIIBurnett RT Thurston GD Thun MJ Calle EE Krewski D Godleski JJ Cardiovascular mortality and long-term exposure to particulate air pollution: epidemiological evidence of general pathophysiological pathways of disease Circulation 2004 109 71 77 14676145 10.1161/01.CIR.0000108927.80044.7F
Schwartz J Dockery DW Neas LM Is daily mortality associated specifically with fine particles? J Air Waste Manag Assoc 1996 46 927 939 8875828
Anderson HR Bremner SA Atkinson RW Harrison RM Walters S Particulate matter and daily mortality and hospital admissions in the west midlands conurbation of the United Kingdom: associations with fine and coarse particles, black smoke and sulphate Occup Environ Med 2001 58 504 510 11452044 10.1136/oem.58.8.504
McDonnell WF Nishino-Ishikawa N Petersen FF Chen LH Abbey DE Relationships of mortality with the fine and coarse fractions of long-term ambient PM10 concentrations in nonsmokers J Exp Anal Environ Epidemiol 2000 10 427 436 10.1038/sj.jea.7500095
Ostro BD Broadwin R Lipsett MJ Coarse and fine particles and daily mortality in the Coachella Valley, California: a follow-up study J Exp Anal Environ Epidemiol 2000 10 412 419 10.1038/sj.jea.7500094
Castillejos M Borja-Aburto VH Dockery D Gold DR Loomis D Airborne coarse particles and mortality Inhal Toxicol 2000 12 61 72 10.1080/089583700196392
Hetland RB Schwarze PE Johansen BV Myran T Uthus N Refsnes M Silica-induced cytokine release from A549 cells: importance of surface area versus size Hum Exp Toxicol 2001 20 46 55 11339625 10.1191/096032701676225130
Ovrevik J Myran T Refsnes M Lag M Becher R Hetland RB Schwarze PE Mineral particles of varying composition induce differential chemokine release from epithelial lung cells: importance of physico-chemical characteristics Ann Occup Hyg 2005 49 219 231 15640311 10.1093/annhyg/meh087
Dick CA Singh P Daniels M Evansky P Becker S Gilmour MI Murine pulmonary inflammatory responses following instillation of size-fractionated ambient particulate matter J Toxicol Environ Health A 2003 66 2193 2207 14669776
Becker S Soukup JM Sioutas C Cassee FR Response of human alveolar macrophages to ultrafine, fine and coarse urban air pollution particles Exp Lung Res 2003 29 29 44 12652814 10.1080/01902140303762
Becker S Soukup JM Gilmour MI Devlin RB Stimulation of human and rat alveolar macrophages by urban air particulates: effects on oxidant radical generation and cytokine production Toxicol Appl Pharmacol 1996 141 637 648 8975789 10.1006/taap.1996.0330
Frampton MW Ghio AJ Samet JM Carson JL Carter JD Devlin RB Effects of aqueous extracts of PM(10) filters from the Utah Valley on human airway epithelial cells Am J Physiol 1999 277 L960 L967 10564181
Dreher KL Jaskot RH Lehmann JR Richards JH McGee JK Ghio AJ Costa DL Soluble transition metals mediate residual oil fly ash induced acute lung injury J Toxicol Environ Health 1997 50 285 305 9055877 10.1080/009841097160492
Kodavanti UP Hauser R Christiani DC Meng ZH McGee J Ledbetter A Richards J Costa DL Pulmonary responses to oil fly ash particles in the rat differ by virtue of their specific soluble metals Toxicol Sci 1998 43 204 212 9710962 10.1006/toxs.1998.2460
Imrich A Ning Y Kobzik L Insoluble components of concentrated air particles mediate alveolar macrophage responses in vitro Toxicol Appl Pharmacol 2000 167 140 150 10964765 10.1006/taap.2000.9002
Schins RPF Lightbody JH Borm PJA Shi T Donaldson K Stone V Inflammatory effects of coarse and fine particulate matter in relation to chemical and biological constituents Toxicol Appl Pharmacol 2004 195 1 11 14962500 10.1016/j.taap.2003.10.002
Bonvallot V Baeza-Squiban A Baulig A Brulant S Boland S Muzeau F Barouki R Marano F Organic compounds from diesel exhaust particles elicit a proinflammatory response in human airway epithelial cells and induce cytochrome P450 1A1 expression Am J Respir Cell Mol Biol 2001 25 515 521 11694458
Monn C Becker S Cytotoxicity and induction of proinflammatory cytokines from human monocytes exposed to fine (PM2.5) and coarse particles (PM10-2.5) in outdoor and indoor air Toxicol Appl Pharmacol 1999 155 245 252 10079210 10.1006/taap.1998.8591
Soukup JM Becker S Human alveolar macrophage responses to air pollution particulates are associated with insoluble components of coarse material, including particulate endotoxin Toxicol Appl Pharmacol 2001 171 20 26 11181108 10.1006/taap.2000.9096
Huang S Cheng WL Lee CT Huang HC Chan CC Contribution of endotoxin in macrophage cytokine response to ambient particles in vitro J Toxicol Environ Health A 2002 65 1261 1272 12167209 10.1080/152873902760125741
Becher R Hetland R Refsnes M Dahl J Dahlman H Schwarze P Rat lung inflammatory responses after in vivo and in vitro exposure to various particles Inhal Toxicol 2001 13 789 805 11498806 10.1080/089583701316941311
Imrich A Ning YY Koziel H Coull B Kobzik L Lipopolysaccharide priming amplifies lung macrophage tumor necrosis factor production in response to air particles Toxicol Appl Pharmacol 1999 159 117 124 10495775 10.1006/taap.1999.8731
Bacon K Gearing A Camp R Induction of in vitro human lymphocyte migration by interleukin 3, interleukin 4, and interleukin 6 Cytokine 1990 2 100 105 1716485 10.1016/1043-4666(90)90003-C
Cassee FR Fokkens PHB Leseman DLAC Bloemen HJT Boere AJF Respiratory Allergy and Inflammation Due to Ambient Particles (RAIAP) – collection and characterisation of particulate matter samples from 5 European sites RIVM report 2003
Demokritou P Gupta T Koutrakis P A high-volume apparatus for the condensational growth of ultrafine particles for inhalation toxicological studies Aerosol Sci Technol 2002 36 1061 1072 10.1080/02786820290092230
Ning Y Tao F Guozhong Q Imrich A Goldsmith CA Yang Z Kobzic L Particle-epithelial interaction – effect of priming and bystander neutrophils on interleukin-8 release Am J Respir Cell Mol Biol 2004 30 744 750 14630611 10.1165/rcmb.2003-0123OC
Dybing E Lovdal T Hetland RB Lovik M Schwarze PE Respiratory allergy adjuvant and inflammatory effects of urban ambient particles Toxicology 2004 198 307 314 15138057 10.1016/j.tox.2004.02.007
Pozzi R De Berardis B Paoletti L Guastadisegni C Inflammatory mediators induced by coarse (PM2.5-10) and fine (PM2.5) urban air particles in RAW 264.7 cells Toxicology 2003 183 243 254 12504355 10.1016/S0300-483X(02)00545-0
Sillanpää M Hillamo R Mäkelä T Pennanen AS Salonen RO Field and laboratory tests of a high-volume cascade impactor J Aerosol Sci 2003 34 485 500 10.1016/S0021-8502(02)00214-8
Hetland RB Cassee FR Refsnes M Schwarze PE Lag M Boere AJF Dybing E Release of inflammatory cytokines, cell toxicity and apoptosis in epithelial lung cells after exposure to ambient air particles of different size fractions Toxicol in Vitro 2004 18 203 212 14757111 10.1016/S0887-2333(03)00142-5
Amakawa K Terashima T Matsuzaki T Matsumaru A Sagai M Yamaguchi K Suppressive effects of diesel exhaust particles on cytokine release from human and murine alveolar macrophages Exp Lung Res 2003 29 149 164 12637227 10.1080/01902140303770
Salonen RO Halinen AI Pennanen AS Hirvonen MR Sillanpaa M Hillamo R Shi T Borm P Sandell E Koskentalo T Aarnio P Chemical and in vitro toxicologic characterization of wintertime and springtime urban-air particles with an aerodynamic diameter below 10 microm in Helsinki Scand J Work Environ Health 2004 30 80 90 15487689
Shi T Knaapen AM Begerow J Birmili W Borm PJA Schins RPF Temporal variation of hydroxyl radical generation and 8-hydroxy-2'-deoxyguanosine formation by coarse and fine particulate matter Occup Environ Med 2003 60 315 321 12709515 10.1136/oem.60.5.315
Alfaro-Moreno E Martinez L Garcia-Cuellar C Bonner JC Murray JC Rosas I Rosales SP Osornio-Vargas AR Biologic effects induced in vitro by PM10 from three different zones of Mexico City Environ Health Perspect 2002 110 715 720 12117649
Salonen RO Pennanen AS Halinen AI Hirvonen MR Sillanpaa M Karlson V Koskentalo T Aarnio P Ferguson S Koutrakis P A chemical and toxicological comparison of urban air PM10 collected during winter and spring in Finland Inhal Toxicol 2000 12 95 103 10.1080/08958370050029806
Aust AE Ball JC Hu AA Lighty JS Smith KR Straccia AM Veranth JM Young WC Particle characteristics responsible for effects on human lung epithelial cells Res Rep Health Eff Inst 2002 1 65 12578113
Samet JM Silbajoris R Wu W Graves LM Tyrosine phosphatases as targets in metal-induced signaling in human airway epithelial cells Am J Respir Cell Mol Biol 1999 21 357 364 10460753
Albrecht C Schins RPF Hohr D Becker A Shi T Knaapen AM Borm PJA Inflammatory time course after quartz instillation: role of tumor necrosis factor-alpha and particle surface Am J Respir Cell Mol Biol 2004 31 292 301 15191911 10.1165/rcmb.2003-0300OC
Donaldson K Stone V Duffin R Clouter A Schins R Borm P The quartz hazard: effects of surface and matrix on inflammogenic activity J Environ Pathol Toxicol Oncol 2001 20 109 118 11570668
Adamson IYR Prieditis H Vincent R Soluble and insoluble air particle fractions induce differential production of tumor necrosis factor alpha in rat lung Exp Lung Res 2004 30 355 368 15204828 10.1080/01902140490438933
Prieditis H Adamson YR Comparative pulmonary toxicity of various soluble metals found in urban particulate dusts Exp Lung Res 2002 28 563 576 12396249 10.1080/01902140290096782
Huang YC Ghio AJ Stonehuerner J McGee J Carter JD Grambow SC Devlin RB The role of soluble components in ambient fine particles-induced changes in human lungs and blood Inhal Toxicol 2003 15 327 342 12635002 10.1080/08958370304460
Solhaug A Refsnes M Lag M Schwarze PE Husoy T Holme JA Polycyclic aromatic hydrocarbons induce both apoptotic and anti-apoptotic signals in Hepa1c1c7 cells Carcinogenesis 2004 25 809 819 14729587 10.1093/carcin/bgh069
Baulig A Poirault JJ Ausset P Schins R Shi T Baralle D Dorlhene P Meyer M Lefevre R Baeza-Squiban A Marano F Physicochemical characteristics and biological activities of seasonal atmospheric particulate matter sampling in two locations of Paris Environ Sci Technol 2004 38 5985 5992 15573597 10.1021/es049476z
Schlesinger RB Cassee F Atmospheric secondary inorganic particulate matter: the toxicological perspective as a basis for health effects risk assessment Inhal Toxicol 2003 15 197 235 12579454 10.1080/08958370304503
Carty CL Gehring U Cyrys J Bischof W Heinrich J Seasonal variability of endotoxin in ambient fine particulate matter J Environ Monit 2003 5 953 958 14710938 10.1039/b308488d
RAIAP Respiratory Allergy an Inflammation Due to Ambient Air Particles – A European-wide Assessment Final Report ISBN: 82-88082-089-2 2004
Hollander A Heederik D Versloot P Douwes J Inhibition and enhancement in the analysis of airborne endotoxin levels in various occupational environments Am Ind Hyg Assoc J 1993 54 647 653 8256688
Long CM Suh HH Kobzik L Catalano PJ Ning YY Koutrakis P A pilot investigation of the relative toxicity of indoor and outdoor fine particles: in vitro effects of endotoxin and other particulate properties Environ Health Perspect 2001 109 1019 1026 11689347
|
16105184
|
PMC1190213
|
CC BY
|
2021-01-04 16:38:30
|
no
|
Part Fibre Toxicol. 2005 Aug 17; 2:4
|
utf-8
|
Part Fibre Toxicol
| 2,005 |
10.1186/1743-8977-2-4
|
oa_comm
|
==== Front
Popul Health MetrPopulation Health Metrics1478-7954BioMed Central London 1478-7954-3-61599846510.1186/1478-7954-3-6ResearchMeasurement of vaccination coverage at age 24 and 19–35 months: a case study of multiple imputation in public health Santibanez Tammy A [email protected] Lawrence E [email protected] Kate M [email protected] Centers for Disease Control and Prevention, National Immunization Program, Atlanta, Georgia, USA2005 5 7 2005 3 6 6 14 7 2004 5 7 2005 Copyright © 2005 Santibanez et al; licensee BioMed Central Ltd.2005Santibanez et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Aim
Childhood immunization coverage in the United States (U.S.) is often measured at age 24 months or, in the National Immunization Survey (NIS) at age of interview, which is between 19 and 35 months. This paper compares these standards.
Methods
Data from the NIS is used to compare immunization coverage at time of interview, retrospectively among all children aged 24 or more months at time of interview, and obtained via multiple imputation (with 10 imputations) for all children, both nationally, by state, and by demographic groups.
Results
At the national level, the difference between the 19–35 month estimate and the 24 month complete-case estimate was 1.9 percentage points. For most but not all states and subgroups, the 19–35 month estimate was higher than the 24 month complete-case estimate. The difference between vaccination coverage measured at 19–35 months and 24 months ranged from -2.3 to 7.5 percentage points among states. For three states, the difference between the 19–35 month and 24 month complete-case estimate was more than 6 percentage points, in twelve states there was a 4–6 percentage point difference, and in sixteen states a 2–4 percentage point difference. Conversely, five states had higher 24 month complete-case estimates than 19–35 month estimates.
Conclusion
We found that the coverages at 19–35 and 24 months differ such that they would rarely be adequate surrogates for one another, particularly at a state level. Multiple imputation, which is easily implemented, increases precision of estimates of coverage at age 24 months.
==== Body
1. Introduction
Multiple imputation is a well-established statistical practice [1-6]; however it is rarely used in health care surveys. For example, the National Health Interview Survey limits reports to complete case analyses [7]. Here, we demonstrate how multiple imputation can be applied to public health surveys, using data from the National Immunization Survey (NIS).
The NIS is the primary means by which immunization coverage is measured among pre-school aged children in the U.S. The NIS, conducted annually since 1995 by the Centers for Disease Control and Prevention (CDC), provides annual estimates of vaccination coverage among children aged 19–35 months for the nation and for each of the 50 states and 28 selected urban areas. Coverage is measured among children aged 19–35 months, as of the time of household interview. Each year the CDC publishes these NIS estimates of coverage at 19–35 months of age [8,9].
Vaccination coverage is measured by many others at 24 months of age. For example, the Health Plan Employer Data and Information Set (HEDIS®), of the National Committee for Quality Assurance, has a Childhood Immunization Status measure to estimate vaccination coverage among managed care enrollees who turned two years old during the measurement year [10,11]. Second, the Retrospective Surveys of School Enterers' Immunization Records measured coverage at 24 months of age[12,13]. In addition, many state, county, and local surveys assess childhood vaccination coverage at 24 months of age [14-18]. Finally, CDC also provides coverage estimates at 24 months (and other benchmark ages) from the NIS on its website ; these estimates include only those children in the NIS who were 24–35 months at the time of interview, thus representing a subset of children in the complete NIS. Other benchmark ages are at ages younger than 19 months, so all children are included in these analyses.
It is plausible that coverage measured at 19–35 months might be higher than that measured at 24 months; some children have up to 12 months extra to obtain vaccinations and some have up to 6 months less time to obtain vaccinations [19]. Is 19–35 month coverage higher than 24 month coverage? How different are 19–35 month and 24 month coverage measures, and is the difference uniform across states and demographic subgroups? Here, we compare vaccination coverage measured at 19–35 months with 24 month coverage, obtained both through complete case analysis and multiple imputation. Further, we demonstrate how multiple imputation could improve precision of estimates of 24 months immunization coverage.
Methods
Survey Design and Collection of Data
The NIS is conducted annually by CDC to obtain national, state, and selected urban-area estimates of vaccination coverage for the U.S. non-institutionalized population of children aged of 19–35 months. The NIS is a random-digit-dialing survey of households with age-eligible children followed by a mail survey of the eligible children's vaccination providers to obtain the children's vaccination information [20]. Interviews are conducted with the household member with best knowledge of the child's immunizations (a female guardian in about 85% of the cases). Analyses are restricted to children whose vaccination history was verified by their providers.
This study uses data from the 2001 and 2002 NIS. The 2001 NIS included children born February 1998 – June 2000 while the 2002 NIS included children born February 1999 – June 2001. Adequate provider data were obtained from 23,642 children in the 2001 NIS and 21,410 children in the 2002 NIS. A complex weighting methodology is used in the NIS to adjust estimates for household non-response, households with multiple phone lines, and vaccination history non-response. Details of NIS methodology have been published previously, and are not repeated here [20,21].
Definitions
We evaluated coverage with the 4:3:1:3:3 series. Being 4:3:1:3:3 up-to-date (UTD) was defined as having: 4 or more doses of diphtheria and tetanus toxoids and pertussis vaccine (DTP), 3 or more doses of poliovirus vaccine, 1 or more dose of any measles-containing vaccine (MCV), 3 or more doses of Haemophilus influenzae type b vaccine (Hib), and 3 or more doses of hepatitis B vaccine.
Race/ethnicity of the child was respondent-reported during the household interview. The ratio of household income to the poverty level ratio was calculated based upon reported household income, number of persons in the household and official U.S. Census Bureau thresholds for poverty. Children were categorized into one of four groups: living "above poverty" for household income/poverty ratios ≥ 125%, living "near poverty" for ratios 100%-124%, "intermediate poverty" for ratios 50%-99%, and "severe poverty" if the household income/poverty ratio was less than 50%. This four-class categorization of poverty better reflects the impact of poverty on vaccination status than the simple above/below poverty level dichotomization [22]. For 2001 and 2002 respectively, 86.2 and 87.4 % of the respondents provided household income data. For the same years, of those providing household income data, useable provider information was obtained from 73.6 and 70.2 % of the respondents; adjustments for provider response propensity are included in the NIS weights [23].
Three Measures of 4:3:1:3:3 Vaccination Coverage
19–35 Month Coverage
All vaccinations received up to the day that the NIS interview was conducted are included in this measure. This is the usually quoted coverage [8,9].
24 Month Complete-Case Coverage
For the 24 month complete-case measure, all vaccinations received up to and including the child's second birthday are included. This measure is calculated only among the children in the NIS who were at least 24 months of age at household interview; children aged 19–23 months at interview are excluded from this measure. This commonly used measure for 24 month coverage is statistically inefficient; information about children aged 19–23 months at time of survey is gathered but not used.
24 Month Imputed Coverage
For this third measure of 4:3:1:3:3 coverage, we modeled 24-month coverage for children aged 19–23 months at the time of interview and used previously calculated 24 month coverage for children who were 24–35 months at the time of interview. Modeled immunizations were multiply imputed using 10 imputations. We divided the children into classes based on age at interview and vaccine doses received by age 19 months. Some children's immunization coverage at age 24 months could be determined either retrospectively (aged 24 months or more at time of interview) or prospectively (aged 19–23 months and: had all doses so will still have them at age 24 months; or had so few doses at interview that being fully immunized at age 24 months was implausible). The 24 month immunization status of some children aged 19–23 months at interview was uncertain. For these children, we constructed a logistic regression model based on demographics. See Appendix 1 (Additional file 1 and Table 2)for details.
Statistical Methods
We compared 19–35 month coverage and 24 month coverage, obtained by both methods, nationally, by race/ethnicity, by poverty status, and by state for both years of NIS data. Percentages are reported with 95% confidence intervals. All analyses were conducted using SAS release 8.02 and SAS-callable SUDAAN release 8.0.0, a software package designed for the analysis of complex survey data.
Results
Table 1 presents 4:3:1:3:3 vaccination coverage estimates, using the three different measures, nationally and by race/ethnicity, poverty status, and state (only 2002 results are presented; 2001 results were similar). The percent of imputed cases (unweighted) also appears in Table 1.
Table 1 Comparison of estimated 4:3:1:3:3* vaccination coverage among children using three different measures, United States, National Immunization Survey, Q1/2002-Q4/2002†
Level of analysis (percent of children aged 19–23 months at time of household interview, unweighted) Estimates for 4:3:1:3:3 Coverage (95% Confidence Interval)
19–35 Month (n = 21,410) 24 Month (Complete case) (n = 14,910) 24 Month (Imputed)‡ (n = 21,410)
Overall (30.4%) 74.8 (73.8, 75.8) 72.9 (71.7, 74.1) 72.7 (71.6, 73.8)
Race/Ethnicity:
Hispanic (31.3%) 72.8 (70.4, 75.1) 72.4 (69.6, 75.2) 71.1 (68.6, 73.6)
White, non-Hispanic (29.7%) 77.5 (76.4, 78.8) 75.3 (73.9, 76.7) 75.2 (74.0, 76.5)
Black, non-Hispanic (31.6%) 67.5 (64.6, 70.5) 63.8 (60.0, 67.5) 65.1 (61.8, 68.3)
All others, non-Hispanic (31.0%) 75.4 (70.8, 80.0) 74.4 (69.3, 79.5) 74.1 (69.3, 78.9)
Poverty Status:
Above poverty (30.2%) 77.0 (75.9, 78.2) 74.7 (73.3, 76.1) 74.9 (73.6, 76.2)
Near poverty (28.4%) 67.6 (62.4, 72.8) 68.2 (62.7, 73.6) 65.9 (60.6, 71.2)
Intermediate poverty (31.7%) 70.4 (67.1, 73.8) 68.1 (64.0, 72.2) 68.1 (64.6, 71.6)
Severe poverty (31.2%) 67.7 (64.1, 71.4) 68.1 (63.8, 72.3) 66.4 (62.5, 70.3)
State:
Alabama (33.3%) 76.8 (71.5, 82.1) 72.6 (65.6, 79.6) 73.6 (67.9, 79.3)
Alaska (31.3%) 75.3 (69.4, 81.2) 67.8 (60.2, 75.4) 68.1 (61.4, 74.8)
Arizona (28.7%) 67.9 (63.2, 72.7) 68.8 (63.3, 74.4) 66.9 (61.7, 72.0)
Arkansas (27.3%) 71.0 (64.9, 77.1) 69.9 (62.9, 76.9) 70.5 (64.0, 76.9)
California (33.4%) 73.2 (69.4, 77.0) 73.4 (68.9, 77.9) 72.3 (68.2, 76.5)
Colorado (26.3%) 62.8 (56.1, 69.4) 60.1 (52.3, 67.9) 58.7 (51.5, 66.0)
Connecticut (27.3%) 81.9 (76.7, 87.1) 77.9 (71.6, 84.2) 78.3 (72.8, 83.8)
Delaware (31.3%) 78.7 (73.2, 84.2) 78.2 (71.6, 84.8) 77.7 (71.7, 83.7)
District of Columbia (27.4%) 69.7 (62.2, 77.2) 66.7 (58.2, 75.2) 67.3 (59.2, 75.5)
Florida (30.2%) 74.5 (69.8, 79.2) 71.3 (65.4, 77.2) 71.1 (65.7, 76.4)
Georgia (32.8%) 80.4 (76.1, 84.6) 80.3 (75.4, 85.2) 78.7 (74.3, 83.1)
Hawaii (34.6%) 78.7 (73.2, 84.3) 73.4 (66.0, 80.8) 75.8 (69.6, 82.0)
Idaho (31.0%) 69.4 (63.4, 75.3) 66.6 (59.1, 74.0) 67.5 (61.2, 73.8)
Illinois (30.1%) 78.6 (74.3, 82.9) 76.8 (71.4, 82.1) 77.2 (72.7, 81.7)
Indiana (29.8%) 76.0 (71.0, 81.0) 78.3 (72.7, 84.0) 75.1 (69.0, 81.2)
Iowa (31.5%) 78.7 (73.2, 84.2) 77.5 (71.2, 83.9) 77.1 (71.3, 82.8)
Kansas (32.0%) 66.8 (59.9, 73.7) 66.5 (58.5, 74.6) 65.8 (58.8, 72.9)
Kentucky (29.8%) 72.3 (65.9, 78.7) 67.5 (59.6, 75.5) 69.6 (62.5, 76.7)
Louisiana (29.6%) 66.8 (61.2, 72.5) 66.4 (60.1, 72.7) 66.3 (60.3, 72.3)
Maine (29.8%) 80.7 (75.6, 85.8) 76.9 (70.3, 83.5) 76.8 (71.1, 82.5)
Maryland (34.4%) 78.7 (73.1, 84.4) 76.3 (69.7, 82.9) 76.5 (70.4, 82.7)
Massachusetts (28.5%) 86.2 (82.5, 90.0) 81.8 (76.4, 87.1) 82.8 (78.5, 87.1)
Michigan (29.9%) 81.6 (77.2, 86.0) 75.8 (69.5, 82.0) 77.0 (71.8, 82.3)
Minnesota (27.7%) 76.8 (70.3, 83.4) 78.5 (71.8, 85.2) 76.1 (69.2, 83.0)
Mississippi (31.1%) 75.7 (69.2, 82.3) 74.4 (65.9, 82.9) 73.3 (65.9, 80.8)
Missouri (28.8%) 73.0 (66.5, 79.6) 66.9 (58.8, 75.0) 69.2 (62.4, 76.1)
Montana (29.7%) 66.6 (59.8, 73.5) 62.6 (54.2, 71.1) 63.6 (56.5, 70.7)
Nebraska (24.8%) 78.2 (72.6, 83.8) 77.5 (71.0, 84.0) 75.8 (69.5, 82.1)
Nevada (28.3%) 76.4 (70.3, 82.4) 74.4 (67.2, 81.6) 75.9 (69.5, 82.2)
New Hampshire (28.8%) 83.5 (78.6, 88.5) 78.4 (72.0, 84.8) 80.9 (75.4, 86.4)
New Jersey (30.1%) 76.1 (70.7, 81.6) 73.1 (65.8, 80.4) 73.0 (66.7, 79.3)
New Mexico (30.9%) 64.6 (57.9, 71.4) 61.2 (52.6, 69.8) 61.2 (54.0, 68.4)
New York (27.9%) 77.5 (73.2, 81.8) 77.9 (73.0, 82.9) 76.6 (72.1, 81.1)
North Carolina (32.4%) 82.4 (76.9, 88.0) 80.3 (73.5, 87.2) 79.8 (73.7, 85.9)
North Dakota (27.7%) 77.7 (71.0, 84.4) 73.5 (64.9, 82.1) 74.8 (67.7, 81.9)
Ohio (29.5%) 75.0 (70.5, 79.6) 72.2 (66.9, 77.6) 71.9 (66.9, 76.8)
Oklahoma (33.6%) 65.3 (57.9, 72.7) 63.1 (54.2, 72.0) 63.3 (55.4, 71.1)
Oregon (33.9%) 70.0 (64.1, 75.9) 64.6 (56.9, 72.3) 65.5 (59.1, 71.8)
Pennsylvania (29.6%) 74.7 (69.3, 80.2) 72.7 (65.9, 79.5) 74.7 (69.0, 80.3)
Rhode Island (32.0%) 84.5 (78.9, 90.1) 78.2 (70.7, 85.6) 81.1 (75.2, 86.9)
South Carolina (32.6%) 78.8 (72.3, 85.3) 77.2 (69.2, 85.1) 77.3 (70.4, 84.1)
South Dakota (33.7%) 79.9 (73.5, 86.3) 77.2 (68.7, 85.7) 75.8 (68.6, 82.9)
Tennessee (31.1%) 78.2 (74.1, 82.3) 75.8 (70.7, 80.9) 76.1 (71.5, 80.7)
Texas (28.6%) 67.9 (62.8, 72.9) 65.2 (59.2, 71.2) 65.3 (60.1, 70.5)
Utah (28.5%) 75.7 (69.9, 81.6) 73.8 (66.8, 80.7) 74.8 (68.6, 81.1)
Vermont (30.6%) 80.9 (76.1, 85.6) 75.2 (69.0, 81.5) 75.4 (69.9, 80.9)
Virginia (31.3%) 72.0 (65.8, 78.2) 68.6 (60.8, 76.3) 70.5 (63.9, 77.1)
Washington (29.6%) 69.2 (64.2, 74.2) 64.8 (59.0, 70.7) 65.1 (59.9, 70.3)
West Virginia (32.4%) 76.9 (70.6, 83.1) 75.8 (69.0, 82.6) 74.2 (67.4, 80.9)
Wisconsin (30.6%) 80.3 (75.9, 84.6) 79.6 (75.1, 84.1) 78.1 (73.6, 82.6)
Wyoming (30.6%) 73.3 (67.0, 79.7) 72.0 (64.4, 79.6) 70.5 (64.0, 77.1)
* 4:3:1:3:3 Four or more doses of any diphtheria and tetanus toxoids and pertussis vaccines including diphtheria and tetanus toxoids, and any acellular pertussis vaccine (DTP/DTaP/DT), three or more doses of poliovirus vaccine, one or more doses of any measles containing vaccine (MCV), three or more doses of Haemophilus influenzae type b (Hib), and three or more doses of hepatitis B vaccine. † Children in the Q1/2002-Q4/2002 National Immunization Survey were born between February 1999 and June 2001; n = 21,410. ‡ 10 imputations were performed.
Table 2 Logistic regression model used for the multiple imputation analysis
Variable Odds for intercept, odds ratio otherwise 95% confidence interval
Intercept 0.06 0.03 – 0.10
Race: Hispanic 0.79 0.61 – 1.03
White, non-Hispanic REF REF
Black, non-Hispanic 0.73 0.53 – 0.99
All others, non-Hispanic 0.95 0.61 – 1.46
Census Region: Northeast 1.90 1.36 – 2.65
Midwest 1.42 1.04 – 1.94
South 1.28 0.97 – 1.69
West REF REF
MSA*: MSA, central city REF REF
MSA, non-central city 1.20 0.95 – 1.51
Non-MSA 0.91 0.68 – 1.23
Number of DTP† doses at 19 months: 4+ doses‡ 2.94 2.37 – 3.65
3 doses REF REF
Number of MCV§ doses at 19 months: 1+ doses|| 2.82 2.11 – 3.80
0 doses REF REF
Number of hepatitis B doses at 19 months: 3+ doses¶ 3.86 2.82 – 5.28
2 doses 0 0
* Metropolitan Statistical Area (MSA)
† Diphtheria and tetanus toxoids, and any acellular pertussis vaccine (DTP)
‡ Four or more doses (4+)
§Measles-containing vaccine (MCV)
|| One or more doses (1+)
¶Three or more doses (3+)
At the national level, the difference between the 19–35 month estimate and the 24 month complete-case estimate was 1.9 percentage points. For most states and subgroups, the 19–35 month estimate was higher than the 24 month complete-case estimate. This can also be seen in Figure 1, which displays the difference between the coverage measures by state. The difference between vaccination coverage measured at 19–35 months and 24 months ranged from -2.3 to 7.5 percentage points among states. For three states, the difference between the 19–35 month and 24 month complete-case estimate was more than 6 percentage points, in twelve states there was a 4–6 percentage point difference, and in sixteen states a 2–4 percentage point difference. Conversely, five states had higher 24 month complete-case estimates than 19–35 month estimates; these were: Indiana, Minnesota, Arizona, New York, and California (Table 1 and Figure 1). Two income subgroups, those near poverty and those in severe poverty, also had higher 24 month complete-case coverage than 19–35 month coverage (Table 1).
Figure 1 Difference Between the 19–35 month, 24 month complete-case, and the 24-month multiple-imputation* estimates, National Immunization Survey, 2002.
Comparing the 24 month complete-case estimates to the 24 month imputed estimates, nationally the imputed measure was 0.2 percentage points lower than the complete-case estimate. The imputed estimates were about equally divided between being greater than or less than the 24 month complete-case estimates (Table 1). As expected, the 24 month imputed estimates had, with the exception of two states (Indiana and Minnesota), narrower confidence intervals than the 24 month complete case estimates. The authors found no explanation of why this was so; in 2001 (results not presented), only one state (North Carolina) had a complete case confidence interval that was narrower than the imputed confidence interval.
The percentage decrease in standard error from using the multiple imputation measure of 4:3:1:3:3 coverage over using the complete case estimate was 8% ([SE complete case-SE imputed case]/ [SE complete case]= [0.61–0.56]/ [0.61]= 0.08). When the imputation model was applied to those for which 24 month coverage was known, the immunization status was correctly assigned for approximately 60%. Note that random variation among imputations was added as an additional component to the model.
Discussion
The problem of determining immunization coverage at age 24 months is well-suited to multiple imputation. Data are missing at random. The NIS telephone interviews are conducted on a national basis throughout the year, making the age of child at time of telephone interview independent of demographics. Age of child at household interview alone determines if a child's immunization coverage at age 24 months is missing.
In comparing coverages, we found that, at the national level, 19–35 month 4:3:1:3:3 coverage was slightly higher than both the 24 month complete-case and imputed estimates (Table 1). However, at the state level there were much larger differences. For five states the 24 month coverage estimates were larger than the 19–35 month estimates. For 2001 (data not shown), 11 states, all different from those of 2002, had larger 24 month coverage than 19–35 month coverage. The authors found no common characteristics of these states; perhaps larger 24 month coverage is simply random variation, and is not explainable by state characteristics.
How can 24 month coverage estimates be larger than 19–35 month coverage? Approximately 29% (percent varies slightly among years) of the children in the NIS are younger than 24 months at time of household interview. In the 24 month complete-case estimate, these children are excluded from both numerator and denominator. If, because of lower incidence of delayed vaccination or increased incidence of very delayed vaccination (age 24–35 months) or by chance, coverage among 24–35 month old children was substantially lower than among those aged 19–23 months, the estimate for 19–35 month olds would be lower than coverage measured at 24 months of age.
We found that our multiple imputation resulted in slightly narrower confidence intervals for all except two states. One could narrow the confidence intervals for the complete case analysis by increasing the sample size. If we can assume that the hypothetical increased sample size would not change the design effect, then to achieve on a national scale the same precision from the complete case analysis that multiple imputation yielded would require a sample size 1.19 times as large ([width of complete case confidence interval/width of imputed case confidence interval]2). To increase the NIS sample size by this magnitude would require an addition of approximately $2.6 million to the survey budget. The improvement in precision is more substantial for some states; as an arbitrarily chosen example, we would need a sample 1.51 times as large in Alabama to get the same improvement in precision that we get from imputation by increasing sample size. Thus, use of multiple imputation produces an essentially cost-free increase in precision of the 24 month vaccination coverage estimates that would be costly and difficult to equal through increased sample size alone. Additionally, because the data are not missing completely at random, but are missing at random, generally there is bias in complete-case methods that is mostly removed (depending on the quality of the imputation model) by multiple imputation.
The potential precision gains to be achieved must be considered in light of drawbacks inherent in this approach. One potential drawback of using multiple imputation is that a predictive model must be specified. However, a body of both analytic and simulation-based work supports the claim that the model used in multiple imputation can be quite far off and still result in substantially better answers than other methods; thus the importance of the "multiple" part of multiple imputation.[24] In this study our model performed only modestly in classifying UTD status of those for whom we knew UTD status; vaccination behaviors involve a complex, and not fully understood interaction of personal beliefs, attitudes, and knowledge, of health provider variables, and system variables.
Conclusion
Many published results concern vaccination coverage among 19–35 month old children [8,9,22]. This is a moving cohort, consisting of children whose age is between 19 and 35 months at any point during a given calendar year; many children in one year's cohort are also in the next year's. Coverage for 19–35 month old children reflects vaccination status at time of household interview; a child's age at time of interview impacts coverage, with older children more likely to be vaccinated. In short, coverage among 19–35 month old children is difficult to interpret while coverage at age 24 months is readily interpreted, and is a common standard [10-18].
Can one directly compare 19–35 month old coverage with 24 month coverage? We have shown that 19–35 month old coverage can be either smaller or greater than 24 month coverage. Thus, neither coverage measure bounds the other. We have shown that coverage at 19–35 months and at 24 months can, at best, serve as crude surrogates for one another, particularly at a state level. We would, for most purposes, caution against direct comparison (although, for those who wish to make direct comparison, our methods provide a measure of the likely differences between these immunization coverages).
We have shown that multiple imputation can yield more precise estimates of coverage at 24 months than is obtained by a complete case analysis, at essentially no increase in cost. Similar methods are applicable to other public health surveys. With today's limited resources for public health surveys, we recommend that multiple imputation be used, where appropriate, to improve precision.
Competing interests
The author(s) declare that they have no competing interests.
Authors' Contributions
All authors participated in the analyses and drafted the manuscript. All authors read and approved the final manuscript.
Supplementary Material
Additional File 1
Appendix 1.
Click here for file
Acknowledgements
We gratefully acknowledge Mary McCauley for her helpful editorial review.
==== Refs
Barnard J Meng XL Applications of multiple imputation in medical studies: from AIDS to NHANES Stat Methods Med Res 1999 8 17 36 10347858 10.1191/096228099666230705
Rubin DB Multiple Imputation for Nonresponse in Surveys 1987 New York: John Wiley & Sons
Rubin DB Schenker N Multiple imputation in health-care databases: an overview and some applications Stat Med 1991 10 585 598 2057657
Schafer JL Multiple imputation: a primer Stat Methods Med Res 1999 8 3 15 10347857 10.1191/096228099671525676
Sinharay S Stern HS Russell D The use of multiple imputation for the analysis of missing data Psychol Methods 2001 6 317 329 11778675 10.1037//1082-989X.6.4.317
Zhou XH Eckert GJ Tierney WM Multiple imputation in public health research Stat Med 2001 20 1541 1549 11343373 10.1002/sim.689
Lethbridge-Çejku M Schiller JS Bernadel L Summary health statistics for U.S. adults National Health Interview Survey, 2002 2004 10 Vital and Health Statistics. National Center for Health Statistics 1 151
CDC National, state, and urban area vaccination coverage levels among children aged 19–35 months – United States, 2001 MMWR Morb Mortal Wkly Rep 2002 51 664 666 12197212
CDC National, state, and urban area vaccination coverage levels among children aged 19–35 months – United States, 2002 MMWR Morb Mortal Wkly Rep 2003 52
NCQA The Health Plan Employer Data and Information Set (HEDIS®) 2003 1-5-2004
NCQA Childhood immunization status 2003 11-26-2003
Simpson DM Ezzati-Rice T Zell ER Forty years and four surveys: how does our measuring measure up? Am J Prev Med 2001 20 6 14 11331125 10.1016/S0749-3797(01)00286-0
CDC Retrospective assessment of vaccination coverage among school-aged children – selected U.S. cities, 1991 MMWR Morb Mortal Wkly Rep 1992 41 103 107 1734227
Texas Retrospective Immunization Survey (TRIS) 2003 12-18-2003
2003 Kindergarten Retrospective Survey Report 10-1-2003. 12-18-2003
Retrospective immunization coverage survey of kindergarten students, capturing coverage levels at the age of 2 years 2003 12-18-2003
Retrospective School Survey 1997-1999 2003 12-18-2003
2002 Retrospective Immunization Survey 2003 12-18-2003
Fairbrother G Freed GL Thompson JW Measuring immunization coverage Am J Prev Med 2000 19 78 88 11024332 10.1016/S0749-3797(00)00208-7
Smith PJ Battaglia MP Huggins VJ Hoaglin DC Rodén A Khare M Overview of the sampling design and statistical methods used in the National Immunization Survey Am J Prev Med 2001 20 17 24 11331127 10.1016/S0749-3797(01)00285-9
Zell ER Ezzati-Rice T Battaglia MP National Immunization Survey: the methodology of a vaccination surveillance system Public Health Rep 2000 115 65 77 10968587 10.1093/phr/115.1.65
Klevens RM Luman ET U.S. children living in and near poverty: risk of vaccine-preventable diseases Am J Prev Med 2001 20 41 46 11331131 10.1016/S0749-3797(01)00281-1
Smith PJ Rao JNK Battaglia MP Ezzati-Rice TM Daniels D Khare M Compensating for provider nonresponse using response propensities to form adjustment cells: the National Immunization Survey Vital Statistics 2001 2 Hyattsville, Maryland. Vital and Health Statistics. National Center for Health Statistics 1 17
Rubin DB Multiple imputation after 18+ years Journal of the American Statistical Association 1996 91 473 489
|
15998465
|
PMC1190214
|
CC BY
|
2021-01-04 16:37:42
|
no
|
Popul Health Metr. 2005 Jul 5; 3:6
|
utf-8
|
Popul Health Metr
| 2,005 |
10.1186/1478-7954-3-6
|
oa_comm
|
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-251596975610.1186/1477-7827-3-25ResearchA link between high serum levels of human chorionic gonadotrophin and chorionic expression of its mature functional receptor (LHCGR) in Down's syndrome pregnancies Banerjee Subhasis [email protected] Alan [email protected] Anne E [email protected] Aris [email protected] Hugues [email protected] Kevin [email protected] Stuart [email protected] Kypros [email protected] Harris Birthright Research Centre for Fetal Medicine, King's College Hospital Medical School, Denmark Hill, London SE5 9RS, UK2 INSERM U135, Biochimie hormonale, CHU de Kremlin-Bicêtre, Bat Paul Broca, 3e niveau 78 avenue du général Leclerc, 94275 Le Kremlin-Bicêtre, France3 Endocrine Unit, Clinical Biochemistry Department, Harold Wood Hospital, Gubbins Lane, Romford RM3 0BE, UK2005 21 6 2005 3 25 25 2 5 2005 21 6 2005 Copyright © 2005 Banerjee et al; licensee BioMed Central Ltd.2005Banerjee et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Human chorionic gonadotrophin (hCG) is released from placental trophoblasts and is involved in establishing pregnancy by maintaining progesterone secretion from the corpus luteum. Serum hCG is detected in the maternal circulation within the first 2–3 wks of gestation and peaks at the end of the first trimester before declining. In Down's syndrome (DS) pregnancies, serum hCG remains significantly high compared to gestation age-matched uncompromised pregnancies. It has been proposed that increased serum hCG levels could be due to transcriptional hyper-activation of the CGB (hCG beta) gene, or an increased half life of glycosylated hCG hormone, or both. Another possibility is that serum hCG levels remain high due to reduced availability of the hormone's cognate receptor, LHCGR, leading to lack of hormone utilization. We have tested this hypothesis by quantifying the expression of the hCG beta (CGB) RNA, LHCGR RNA and LHCGR proteins in chorionic villous samples. We demonstrate that chorionic expression of hCG beta (CGB) mRNA directly correlates with high serum hCG levels. The steady-state synthesis of LHCGR mRNA (exons 1–5) in DS pregnancies was significantly higher than that of controls, but the expression of full-length LHCGR mRNA (exons 1–11) in DS was comparable to that of uncompromised pregnancies. However, the synthesis of high molecular weight mature LHCGR proteins was significantly reduced in DS compared to uncompromised pregnancies, suggesting a lack of utilization of circulating hCG in DS pregnancies.
==== Body
Introduction
The incidence of aneuploidy in human pregnancies is unusually high (1–2%) compared to other mammals [1]. Monosomies and trisomies together account for 35% of clinically detected spontaneous abortions (6–20 wks of gestation), stillbirth (4%) and most importantly, are the leading cause of developmental disability and mental retardation of those surviving such pregnancies [2-4]. Of all the genetically compromised pregnancies, Down's syndrome (Trisomy 21, T21) is the most frequent (1/700 live births [5]). The Edward's (Trisomy 18, T18) and Pautau's (Trisomy 13, T13) syndromes are considered relatively rare pregnancy disorders with a prevalence at birth of 1 in 7000 and 29000, respectively [6,7].
Genetically, 89–95% of Down's syndrome (DS) patients carry an extra chromosome 21 (chr 21) which arises due to meiotic nondysjunction and is usually inherited from the mother [1]. About 1–2% of DS patients have genetic mosaicism (nondysjunction following fertilisation in early embryos), while 3–4% of cases are due to translocation of chr 21 to another autosome, usually chr 14 [8]. In addition to the characteristic variability in mental retardation, physical and facial features, congenital heart and gastro-intestinal defects, the DS patients are also susceptible to leukaemia and Alzheimer's-like dementia [9-11].
The chromosomal abnormalities in DS and other trisomic pregnancies are very often associated with increased or reduced levels of proteins, growth factors and hormones in the maternal blood compared to those of normal pregnancies. For example, in DS pregnancies (11–14 wks of gestation), the serum human chorionic gonadotrophin beta (hCG-β) and pregnancy-associated plasma protein-A (PAPP-A) concentrations tend to be high and low, respectively [12].
Human chorionic gonadotrophin (hCG) is the key reproductive hormone regulating human pregnancy. It is a member of the family of glycoprotein hormones that includes luteinizing hormone (LH), follicle stimulating and thyroid stimulating hormones, each member of which functions through the formation of a non-covalent heterodimer from two subunits, α and β.
In human placenta hCG is primarily produced by syncytotrophoblasts and to a certain extent by extravillous cytotrophoblasts [13]. One of the earliest endocrine roles of hCG is to sustain the corpus luteum which must produce enough progesterone to establish pregnancy at the outset. In addition, hCG facilitates trophoblast differentiation, remodeling of the uterine epithelium and stroma (decidualization) and endometrium for implantation, invasion of the maternal spiral arterioles, and angiogenesis by acting on vascular smooth muscle and endothelial cells [14]. In normal pregnancies, detectable levels of hCG begin to appear in the maternal circulation at about 2–3 wks after conception, and reach their peak at ~11–13 wks before declining significantly in the later stages of pregnancy. Indeed, high serum hCG levels at mid-late pregnancy have been associated with pre-eclampsia, intra-uterine growth restriction and Down's syndrome (DS) [15-18].
The hCG hormone transduces signals by binding to its specific LH/hCG receptor (LHCGR) expressed on surface of the cell. Since hCG and LH receptors are identical, it is often referred to as the LH/hCG receptor (LHCGR) and is encoded by a single copy ~70 Kb LHCGR gene, located on human chromosome 2p21 [19]. This receptor is structurally very similar to two other hormone receptors (thyroid stimulating and follicle stimulating hormone receptors). The LHCGR gene has 11 exons and codes for multiple alternatively spliced species (at least 6) of mRNA. These different mRNA transcripts are initiated at multiple sites spanning a region more than a kilobase upstream of the first exon [20].
On the basis of structure and topology, LHCGR is a member of the rhodopsin/β-adrenergic receptor superfamily of G protein-coupled receptors. Agonist (hormone) binding to LHCGR allows dissociation of membrane-bound cognate G proteins that regulate phospholipase C, adenylyl cyclase and ion channels which in turn control cellular inositol phosphates, cAMP, Ca+2 and other secondary messengers [21,22].
LHCGR is a 701 amino acid residue protein containing three distinct domains: an unusually large (340 residues) N-terminal extracellular domain which binds hCG, a serpentine transmembrane (TM) region containing seven TM repeats connected by three extra- and intracellular loops, and a C-terminal tail. The predicted relative molecular mass (Mr) is ~75 K, or higher, depending upon the level of glycosylation [23].
Moreover, alternatively spliced mRNAs produce several truncated intra-cellular protein isoforms which have ligand binding capacity but are ineffective in transducing signals [24]. The functional significance of all isoforms remains to be established. However, the accessibility of the Mr 85–95 K species to surface biotinylation, protease and glycosidase (neuraminidase), suggests that they have ligand binding and signal transduction capacities. On the other hand, the Mr 65–75K proteins contain high-mannose type side chains which are susceptible to endoglycosidase H, and are immature and intracellular [25-27]. The high relative molecular mass 165–200K group is thought to be a dimer of the mature functional receptor [28]. Interestingly, smaller species of LHCGR proteins (Mr 45–51K) can be detected in tissues or cells transfected with cDNAs [25-27,29].
Natural missense [30] and deletion mutations [31] of the human LH receptor have been reported to be associated with elevated serum LH levels in these patients. Similarly, the circulating LH concentration remains high [32,33] in mice carrying a homozygous deletion of Lhr gene (Lhr-/-). Moreover, the lack of functional cytokine receptor expression, due to natural mutations of the IFN-γ receptors 1 and 2, has been directly linked to high serum IFN-γ levels in patients suffering from infectious diseases [34,35]. These reports prompted us to investigate whether increased serum hCG levels in DS pregnancies could be linked to the expression of its functional receptor in chorionic villi.
Materials and methods
Placental tissues and chorionic villous samples
This study was approved by the local ethics committee of King's College Hospitals, London, UK and written consent was obtained from patients before the collection of samples. Placental tissue was obtained from patients undergoing termination of pregnancy with gestational age range of 7–12 wks, after vaginal delivery or caesarian section. Chorionic villous samples (CVS) were collected in a Petri dish and material not required for genetic analysis was washed in Ca+2/Mg+2-free PBS (Invitrogen, Carlsbad, CA, USA) and stored in 500 μl RNA Later (Ambion, Huntingdon, UK), held at 4 C overnight before long term storage at -20°C.
RNA extraction and cDNA synthesis
Depending upon the availability, up to 30 mg of CVS was extensively washed in PBS, homogenized in 500 μl Trizol (Invitrogen) with a tissue grinding pestle (Anachem, UK). Subsequent RNA extraction, DNase I treatment (Sigma, St Louis, MO, USA), cDNA synthesis were exactly as described previously [36,37]. In some cases, a Qiagen RNA extraction kit (Qiagen, West Sussex, UK) was used and RNA was stored at -70°C in water or ethanol.
Quantitative PCR
Quantitative PCR was performed using the Light Cycler RNA amplification system in glass capillaries and a fluorescence-based hybridization detection format (Roche Diagnostics GmbH, Mannheim, Germany) as described [38]. Briefly, the assays were carried out in duplex where both the experimental sample and an internal control (ACTB and HPRT) were run in the same reaction. The reporters LC-Red 640 and LC-Red 705 were employed to generate hybridization probes for experimental and internal controls and the amplification for each cDNA was recorded by dual color detection. A color compensation file was created according to the manufacturer instructions (Roche Diagnostics) and was compensated for PCR cycles in duplex during each run. In some experiments, the control reactions were run at the same time as the test samples under the same reaction conditions, but as single reactions. In order to establish that there was no cross contamination, negative controls (a full reaction without cDNA) were run in each experiment.
Cross-contamination was avoided by sequentially adding water, reaction master-mix (containing enzyme, Mg+2 and PCR buffer; Light Cycler RNA amplification Kit, Roche Diagnostics), and cDNA to a final volume of 10 μl. The analysis mode set to quantification included initial denaturation at 95°C for 10 min followed by 40 cycles consisting of the following parameters for segments 1–3: Target temp, 95°C, 55°C and 72°C respectively; incubation time 10, 7 and 12 sec respectively; Transition rate 20°C/sec, 20°C/sec and 10°C/sec respectively, and a single acquisition mode at segment 2. PCR primers, hybridization probes and amplicon lengths are shown in Table 1. Data were automatically collected and filtered to remove background by the Light Cycler software which set the crossing point for all the different reactions against the standard curve. Data were transferred to Microsoft Excel for further analysis.
Table 1 Quantitative PCR primers, HUGO approved gene names, GenBank accession numbers, hybridization probe sequences and the length of respective amplicons.
Gene Acc. No PCR Primer Hybridization Probes bp
ACTB XM_037235 F 5'-agc ctc gcc ttt gcc ga-3'
R 5'-ctg gtg cct ggg gcg-3' 5'-ttg cac atg ccg gag ccg ttg-FL-3'
5'-LC Red 705-cga cga cga cgc cgg cga tat c-Ph-3' 178
HPRT M31642 F 5'-atc aga ctg aag agc tat tgt aat gac ca-3'
R 5'-tgg ctt ata tcc aac act tcg tg-3' 5'-aga ctt tgc ttt cct tgg tca ggc agt-FL-3'
5'-LC Red 705-aat cca aag atg gtc aag gtc gca agc-Ph-3' 230
CGB (hCG β) NM_000737 F 5'-gac gca cca agg atg gag at-3'
R 5'-gcg gta gtt gca cac cac ct-3' 5'-gtg tgc atc acc gtc aac acc acc-FL-3'
5'-LC Red 640-tct gtg ccg gct act gcc cca c-Ph-3' 251
LHCGR Ex 1–5 NM_000233 F 5'-tcg act atc act tgc cta cc-3'
R 5'-gga gaa gac ctt cgt aac at-3' 5'-ttt gtc tga aat act gat cca gaa cac ca-FL-3'
5'-LC Red 640-aat ctg aga tac att gag ccc gc-Ph-3' 291
LHCGR Ex 11 F 5'-act tcc tta ggg tcc tg-3'
R 5'-gtg atg acg gtg agg g-3' 5'-ggc tct atc tgc tgc tca tag c-FL-3'
5'-LC Red 640-cag ttg att ccc aaa cca agg g-Ph-3 303
Cell culture, protein extraction from placenta and CVS, gel electrophoresis and Western blots
The HEK-293 cell line expressing N-terminal 362 amino acid residues of human LHCGR was kindly provided by Professor Axel Themmen, Erasmus Universty, Rotterdam, The Netherlands. The expression vector contained LHR extracellular domain (ECD, 1–362) fused to a tag peptide (YPYDVPDYA) from the hemagglutinin 1 (HA1) epitope of influenza virus and tetracycline (tet)-inducible promoter. Cells were grown exponentially in tet-free fetal bovine serum prior to overnight induction with tetracycline as recommended.
Protein extraction from cultured cells, placental villous tissues with T-PER (Perbio, Helsinborg, Sweden) and from CVS, following Trizol lysis were exactly as described previously [36] except the 50 mM Tris-HCl, pH 8.0 buffer was replaced by 25 mM HEPES-OH, 8.0. The total protein concentration in each extract was measured in duplicate (Lowry assay; BioRad DC substrates, BioRad, Hemel Hempstead, UK). Based on this estimate, approximately, 10–20 μg of total protein was loaded in each lane; for each CVS sample it was 10 μg per lane. The separation of proteins by SDS-polyacrylamide gels and Western blot analysis were as described [36-38]. Both 1% casein and 1% non-fat milk were equally effective blocking agents in Western blots.
The primary and secondary antibodies used were as follows: murine control IgG (Sigma) at a concentration of 1 μg/ml, protein A-sepharose purified anti-human LHCGR mouse monoclonal antibody (LHR-29) at a concentration 1 μg/ml, anti -β Actin, clone AC-15 (Sigma) and goat anti-mouse IgG (H+L) HRP-conjugated (Chemicon International Inc., CA, USA) at dilutions of 1 in 2000 and 1 in 5000, respectively.
Hormone assays
Patient information and consent forms were given to each patient who came for ultrasound scan at HBRC, King's College Hospital. The venous blood (5 + 5 ml) was collected from those who consented (at 12–14 wks of gestation) with and without anticoagulant. Sera and plasma obtained by centrifugation (1500 rpm, 10 min at 4 °C) were aliquoted and stored at -20 °C.
Free hCG β and intact hCG were measured using the Brahms Kryptor (Brahms AG, Berlin, Germany) random continuous access immunoassay analyzer by a time resolved amplified cryptate fluorescence emission method. The performance of these methods have been described previously [39,40].
Densitometry and data analysis
Densitometry of autoradiograms was carried out using a 1D-Multi Lane Densitometry program in an AlphaImager (1220v5.5, Alpha Innotech Corp. San Leandro, CA, USA) as described [36-38]. Scan data (experimental and β-actin control) were transferred to Microsoft Excel where the pixel density of each experimental lane was normalized to its corresponding β-actin value. Each experiment was repeated at least twice and average values for each data point were plotted. The means, standard deviations, variance (anova) for each data-set were computed using Analysis ToolPak (ATP) software. Values are shown as mean +/- SEM. A value for the level of significance (P-value) was calculated using the Poisson statistic. P<0.05 was considered significant. In experiments, where mRNA expression and the serum hormone concentration data were not normally distributed, the median values and 95% confidence intervals were calculated and the Mann-Whitney non-parametric U-test was employed to establish statistical significance.
Results
The goal of this study was to examine the placental expression of the LHCGR mRNA and functional receptor protein expression with respect to serum hCG concentrations in Down's syndrome pregnancies. To achieve this, 1,152 CVS from high-risk pregnancies were collected. Of these, 58 were Down syndrome (DS, trisomy 21 [T21]), 22 Edwards's syndrome (trisomy 18 [T18]) and 12 were Patau's syndrome (trisomy 13 [T13]) confirmed by biochemical, molecular and cytogenetic analyses. The number of samples that contained sufficient tissue for RNA analysis was 41 for T21, 14 for T18 and 7 for T13.
CGB (hCG β) and LHCGR genes are hyperactivated in Down's syndrome pregnancies
The hCG α subunit is synthesized in excess and is common to all members of this hormone family, whereas the hCG β subunit, which recognizes the cognate receptor, is specific for the hormone. Therefore, in order to evaluate the chorionic regulation of the hormone, the expression of hCG β mRNA synthesis was measured.
The expression of hCG β (CGB) mRNA in CVS were assayed by quantitative real-time PCR (Q-PCR) amplification of cDNA. Since mRNA expression values in trisomic pregnancies exhibited a wide range of distribution, the 95% confidence interval limits and the median values in each pregnancy conditions were determined. Such analysis revealed that CGB (hCG β) gene expression in Down's syndrome CVS was significantly higher (P <0.001) compared to that of controls (Fig. 1). Moreover, hCG β (CGB) mRNA expression levels in T18 and T13 pregnancies were comparable to those of controls (Fig. 1).
Figure 1 CGB (hCG β) expression in chorionic villous samples (CVS) from DS and other trisomic pregnancies. Quantitative PCR analysis (Table 1) of the chorionic hCG β (CGB) mRNA expression in control (N = 24), T21 [DS, N = 41], T18 (N = 14) and T13 (N = 7) pregnancies. **P<0.01
The structure of the full-length (FL) human LHCGR cDNA (exons 1–11), the coding region for extracellular domain (ECD), hinge region, transmembrane (TM), and intracellular domains (ICD) of the receptor, together with multiple alternatively spliced isoforms [41-43] are shown in Fig. 2a. The ECD alone has high ligand affinity [44] whereas, the TM and ICD are necessary for signal transduction [24]. The majority of isoforms exhibit deletion of exon 9 and 11 as observed in sheep [45], pig [21,27] and rat [46]. Deletion of coding sequences due to alternative splicing of isoforms 1–5 maintains an open reading frame, but in isoforms 6 and 7 frame-shift mutations are introduced resulting in stop codons at exon 11 and the potential production of soluble truncated receptor [42,43].
Figure 2 The structure of LHCGR mRNA and expression in chorionic villous samples (CVS) from DS and other trisomic pregnancies. a) the organization of exons 1–11 in full-length (FL) LHCGR mRNA (open boxes) and possible alternatively spliced isoforms 1–7 [41-43]. The sequences deleted from the isoforms are indicated by closed boxes. Regions of mRNA encoding the extracellular domain (ECD), transmembrane (TM) domain, the hinge region and the intra-cellular domain (ICD) are shown. The regions of cDNA (exons 1–5 and exon 11) amplified by Q-PCR are indicated by bidirectional arrows. b) chorionic LHCGR mRNA expression (exons 1–5) in control (N = 24), T21 [DS, N = 23], T18 (N = 8) and T13 (N = 3) pregnancies; c) chorionic LHCGR mRNA expression (exon 11) in control (N = 15) and T21 (N = 18) pregnancies. The median values and 95% confidence ranges of RNA expression in each pregnancy condition are shown; N = number of patient samples analyzed. *P < 0.05, **P < 0.01.
The LHCGR mRNA (exons 1–5) expression in DS pregnancies was significantly higher (P= 0.0501) compared to that of T18, T13 and uncompromised pregnancies (Fig. 2b). The CGB gene expression (Fig. 1) positively correlated with LHCGR exons 1–5 mRNA synthesis (correlation coefficient, r= 0.61). These results suggested that both the CGB and LHCGR genes were hyperactivated in DS placenta compared to expression levels in normal and other trisomic (T18 and T13) pregnancies. Notably, the quantitative increase in hCG β (CGB) /LHCGR mRNA production in the T21 group of pregnancies (Figs. 1 and 2b) was at least 3-fold higher than that observed for the T18 and T13 CVS.
The results described above showed a significant increase in LHCGR mRNA transcription in DS placenta compared to that of control pregnancies. However, as noted above, alternative splicing could give rise to mRNA variants that may not encode the full-length functional receptor. As a critical test of whether the quantitative increase in LHCGR (exons 1–5) mRNA in DS placenta truly reflects full-length receptor mRNA synthesis, the transcription of the 3' end of the gene (representing exon 11) was measured by Q-PCR using the same set of cDNA samples. The results (Fig. 2c) demonstrate that the chorionic expression of exon 11 in DS is comparable to that of control pregnancies, indicating that a significant population of LHCGR mRNA in DS placenta does not contain parts of exon 11.
In further attempts to measure the full-length LHCGR transcripts, we have tested three sets of custom-designed primer and probe to amplify exons 10–11 by Q-PCR. None of these were capable of amplification of LHCGR-specific cDNA whereas β-actin, CGB,IFNGR1,IFNGR2, LIFR and syncytin could be amplified by semiquantitative and Q-PCR. The LHCGR exons 7–9 could be amplified by Q-PCR from placental cDNAs obtained from early and late pregnancies. However, the signal intensity was reduced by at least 100-fold compared to that of exons 1–5 or exon 11. Therefore, the amount of cDNA in CVS samples was not sufficient for either semiquantitative or Q-PCR. To ensure that the light-cycler signals during exon 11 amplification were not due to DNA contamination, equivalent amount of the DNase-treated mRNA (not reverse transcribed) corresponding to experimental samples were tested. Amplification only occurred when cDNA was added to the Q-PCR.
Serum hCG β and intact hCG (α and β) are significantly elevated at 11–14 wks gestation in Down's syndrome pregnancies
Consistent with data from previous studies [12,17,39,40], serum hCG β levels in DS pregnancies were significantly higher (P <0.01) compared to that of uncompromised pregnancies. Moreover, serum hCG β levels were lowest in T18 (P <0.01) and were significantly low in T13 (P <0.05) compared to control serum samples (Fig. 3a). When adjusted for gestational age, the mean serum hCG β concentrations in DS were between 2 and 3.6-fold higher than those of normal pregnancies (12–14 wks). Serum hCG β levels positively correlated (r = 0.39) with hCG β (CGB) mRNA expression.
Figure 3 Serum hCG β and hCG heterodimer concentrations in DS and other trisomic pregnancy conditions. a) serum hCG β hormone concentrations in control (N = 18), T21 (N = 23), T18 (N = 7) and T13 (N = 4) pregnancies b) serum hCG heterodimer concentrations in control (N = 19), T21 (N = 15), T18 (N = 12) and T13 (N = 4) pregnancies; c) serum hCG heterodimer concentrations at 11 wk (control, N = 417; T21 N = 68), 12 wk (control, N = 417; T21, N = 148), 13 wk (control, N = 161 ; T21, N = 73), and 14 wk (control, N = 65 ; T21, N = 14) pregnancies; The median hormone concentrations and 95% confidence intervals for each data set are shown. N = number of patient samples analyzed. *P < 0.05, **P < 0.01.
As noted above, the circulating hCG which transduces signals by binding to its receptor LHCGR is intact hCG heterodimer composed of both α and β subunits. In order to establish the relation between the circulating ligand hCG and its receptor expression, we next measured the hCG heterodimer concentrations in serum from normal and trisomic pregnancies. Such analysis revealed that serum hCG heterodimer concentrations in DS pregnancies (12–14 wks) was significantly increased (P <0.01) compared to that of control sera (Fig. 3b). The hCG heterodimer concentration in T13 pregnancies was comparable to uncompromised controls, but was significantly reduced (P <0.01) in T18 pregnancies (Fig. 3b).
These results suggest that circulating free β hCG and intact hCG hetrodimers are abundant at 11–14 wks of pregnancy in DS. The data shown in Fig. 3b were obtained from a limited number of serum samples from control (n, 18) and DS (n, 24) pregnancies predominantly at 12–14 wks of gestation. In order to further verify these data, the hCG heterodimer concentrations at 11, 12, 13 and 14 wks of pregnancy measured in sera from a large number of control and DS pregnancies as part of the previous studies [39,40] were compared. Such analysis revealed that hCG heterodimer concentrations in DS pregnancies were significantly higher than that of control sera at each time point tested (Fig. 3c).
Western blot analysis to establish the specificity of the mouse monoclonal anti human LHR29 antibody
LHR29 monoclonal antibody was originally obtained by immunizing mice with the purified recombinant human receptor extracellular domain (amino acids 75–406) expressed in Escherichia coli. The specificity of the antibody was verified by its ability to immunoprecipitate recombinant receptor and immunopurify 125I-hCG-receptor complexes from transfected cells, western-blot analysis with the immunogen, immunocytochemistry in cells transfected with either the cloned receptor or a mock vector. Patterns obtained by immunohistochemistry of human testis matched with results expected for a transmembrane receptor specific to Leydig cells ([47], and Axel Themmen, personal communication). In order to further verify the antigenic specificity of this antibody, the HEK 293 cell line expressing LHR ECD (1–362) was grown in the presence and absence of tetracycline. Extracted proteins were resolved via 8% SDS-PAGE, Western blotted, and blots were reacted with control mouse IgG or LHR 29 monoclonal antibody. In order to ensure that equal quantities of protein were transferred, the blots were stained with coomassie brilliant blue following chemiluminescence detection. The results (Fig. 4a, and 4b) demonstrate that the LHR29 monoclonal antibody specifically recognizes at least three (Mr 44–48K) tet-inducible species of LHCGR expressed in vitro (two bands appear to migrate as doublet). These three variants possibly reflect different levels of glycosylation. Moreover, these species (Mr 44–48K) are also recognized by anti-HA1 antibody in Western blots (Axel Themmen, personal communication), providing a further line of evidence that the LHR 29 antibody reaction is specific.
Figure 4 The LHCGR extracellular domain (ECD) expressed in HEK-293 specifically reacted with human LHCGR mouse monoclonal antibody, LHR29. HEK 293 cells (expressing human LHCGR ECD, amino acid residues 1–362) were grown exponentially in tetracycline-free fetal bovine sera in the absence (lanes 1 and 3 in a and b, respectively) and in the presence of tetracycline (lanes 2 and 4 in a and b, respectively). Each lane contains 25 μg of total protein separated via electrophoresis through 8% polyacrlamide SDS gels. Blots were immuno-reacted with antibodies (lanes 3 and 4 of a and b, respectively). A shorter exposure of lane 4 (a) is shown on the right hand side of the Fig. 4a. Following chemiluminescence detection, blots were stained with coomassie brilliant blue (lanes 1 and 2 of a and b, respectively).
Human placenta expresses at least six LHCGR protein variants
In order to examine the LHCGR proteins produced in human placenta, the villous tissues obtained from 7 wk- and 10 wk-gestational age placenta were detergent extracted, reacted with non-specific control mouse IgG (not shown) and LHR29 in Western blots. To further control the experiment, the extracts from HEK293 (LHR ECD) were also incorporated and the blots were stained with coomassie blue following chemiluminescence detection. We detected (Fig. 5a) at least six major LHCGR variants ranging in molecular mass (Mr) from 44K-95K (44K, 48K, 52K, 62–68K, 80K and 95K). These bands were also detected by LHR74 which recognizes LHCGR epitopes different from that of LHR29, but not when the primary antibody was murine IgG (data not shown). The Western blot patterns of human placental tissue obtained with LHR29 antibody were very similar to those described by VuHai-LuuThi et al in porcine testis [27,47]. Additionally, our results are fully consistent with the data more recently reported by Bukovsky et al in human tissues [28] using independent mouse monoclonal antibodies (anti-LHR mAb clone 3B5) and in LH-induced human M17 neuroblastoma cells [48] using a rabbit polyclonal antibody (raised against the N-terminal peptide sequence 15–38 of the rat LH/CG receptor).
Figure 5 At lease six LHCGR protein isoforms are expressed in human placenta. a) The HEK 293 (LHCGR ECD 1–362) cells grown in the absence (lane 1), in the presence of tetracycline (lane 2), 7 wk (lane 3) and 10 wk (lane 4) of gestation placental tissues were lysed with detergent (T-Per) and extracts (10 μg of protein in each lane) were immuno-reacted with LHR 29 antibody; b) Placental tissues (10–14 wks gestation) were extracted with Tri-zol reagents, proteins (10 μ/lane) were separated by extended electrophoresis in 8% SDS-PAGE and the blot was reacted with LHR29 antibody.
The chorionic villous samples (20–30 mg tissues) are not sufficient for separate RNA and protein analysis. We have recently described a method where RNA, DNA and proteins can be quantitatively recovered from the same sample by Trizol extraction [36]. The results shown in Fig 5b demonstrate that the detergent extracted LHCGR protein variants from placenta (Fig. 5a) were identical to those observed with Trizol extracted proteins from human placenta. Indeed, the Trizol-extracted bands were somewhat sharper than the corresponding detergent extracted LHCGR variants in Western blot analyses.
Expression of high molecular weight full-length LGCGR proteins is reduced in Down's syndrome pregnancies
In order to compare the LHCGR mRNA and protein expression both mRNA and protein samples were extracted from the same Trizol lysate. The LHCGR protein expression in control and genetically compromised CVS was examined by Western blot analysis and representative data from such analyses are shown in Fig. 6.
Figure 6 The production of mature LHCGR isoforms in chorinic villi from Down's syndrome pregnancies are significantly reduced compared to that of controls. The CVS samples were Tri-zol extracted to recover mRNA as well as proteins. Approximately 10 μg of total protein was loaded in each lane. The proteins extracted from DS (T21) CVS (a) and control CVS (b) pregnancies were resolved in 8% polyacrylamide-SDS gels, Western blotted and immunoreacted with anti-human LHCGR (LHR-29) monoclonal antibody. Blots were stripped prior to immunostaining with anti-β-Actin monoclonal antibody. The data shown in a) and b) were from the same experiment except that the control and DS proteins were separated in two gels at the same time. In order to compare the band intensity in different experiments, two known CVS samples in duplicate were incorporated in each experiment. The density of the 80K LHCGR and 42K β-actin bands served as references for quantitative analysis of the experimental samples. The relative migration of the isoforms is indicated by an arrow. The Mr 80K protein band (LHCGR p80), indicated by * in a) and b), well separated from the neighboring variants were scanned and c) the relative densities of the LHCGR p80 with respect to β-Actin in normal and trisomic pregnancies, n = total number of experiments carried out on protein samples in each condition. **P < 0.01.
The Mr 80–110K LHCGR protein isoforms are thought to be the full-length functional receptor which is expressed on the cell surface and has ligand binding and signal transduction capacity [24,27]. In order to distinguish between the expression of full-length and other LHCGR isoforms, proteins in trisomic and normal CVS were separated by extended electrophoresis and LHCGR variants were detected by Western blot (Fig. 6a and 6b). A visual examination of the blots shows that that the mature isofoms (≥80K) were less abundant in DS CVS compared to control pregnancies. Moreover, the stoichiometric yield of the variants (Mr 44K-52K) in DS CVS (Fig. 6a) appears to be distinctly different from that of control CVS (Fig. 6b). For direct comparison, the full-length functional isoform was quantified by densitometry of the Mr 80K bands with respect to β-Actin expression in each lane. Each sample including the reference in duplicate (control CVS extracts) was analysed in at least two independent experiments and the mean band density was used as a measure of expression for each sample. Placental expression of LHCGR (Mr 80 K) was lowest in DS (P<0.01), remained unchanged in T13 and was marginally reduced (P<0.06) in T18 CVS compared to that of control averages (Fig. 6c). Serum hCG heterodimer levels negatively correlated (r = -0.37) with LHCGR Mr 80 K protein expression in DS pregnancies.
Discussion
In this study we have demonstrated that despite the high concentrations of serum hCG heterodimer and hCG β in DS pregnancies, their autocrine/paracrine effects on the placenta may be severely impaired due to a reduced expression of the hormone's cognate functional receptor. The accumulation of high levels of serum hormone or cytokine as a result of inadequate receptor-mediated signaling is not unprecedented [34,35]. Serum concentrations of IFN-γ are significantly higher in children suffering from innocuous mycobacterial infection due to the inheritance of non-functional IFN-γ R1 and 2 receptors [34,35]. Partial or complete inactivation of the LHCGR gene due to a naturally occurring somatic mutation within the coding sequence could be responsible for the increase in serum LH concentrations in leydig cell hypoplasia, male hypogonadism, and primary amenorrhea [49,50].
There are conflicting reports on the transcriptional regulation of human hCG α and β mRNA in DS pregnancies. For example, some studies suggest that the steady-state RNA synthesis from the CGB gene in Down's syndrome and gestation age-matched control pregnancies are comparable [51], while others show that the CGB gene is activated [52,53] or repressed [54] in comparisons of in vitro cultured trophoblasts from DS and control placenta.
The work presented here shows that when mRNA synthesis is quantified for a large number of CVS samples (41), the CGB (hCG β) gene in DS placenta is upregulated. This suggests that the conflict between previous reports may be due to sample size. In addition to this, the gestational age of the DS placenta, as well as the methodology employed by different laboratories to measure mRNA (northern blotting and PCR), and to purify and in vitro culture of cytotrophoblasts (CT) (discussed by Goshen [55]), could have contributed to the conflicting results. Purified CTs or placental explants have limited ability to differentiate under normal O2 tension and exhibit an invasive phenotype in vitro [54,56-58]. Purified CTs cultured under 2% O2 tension undergo a change into an invasive phenotype [59]. Notably, differentiation also leads to the formation of multinucleated giant cells instead of polarized epithelial layers of ST (the major source of hCG β) with typical microvillous structure and special antigen repertoire (Susan Fisher, personal communication).
The steady state level of transcription (as measured by Q- PCR of exons 1–5) from the LHCGR gene in DS was significantly higher than that of control CVS, whereas, the receptor expression in T18 and T13 pregnancies did not significantly differ from that of gestation age-matched control pregnancies (Fig. 2b). Nevertheless, a comparison of the expression of exon 11 from control and DS CVS indicated that a large proportion of the LHCGR transcripts in DS CVS may not be full-length, since they did not contain parts of exon 11. This might explain why up-regulation of LHCGR mRNA (exon 1–5) did not correlate with the expression of mature LHCGR p80 and other high Mr isoforms which are less abundant in DS placenta (Figs. 2 and 6). These results also provide an explanation for previously reported increase in LHCGR mRNAs from T21 and T18 pregnancies compared to controls where cDNA common to all spliced variants of LHCGR mRNAs was used as an in situ hybridization probe on placental sections [60]. Interestingly, semi-quantitative PCR amplification of placental cDNA (LHCGR exons 9–11) and agarose gel analysis revealed that truncated products were highly abundant in late, compared to early, pregnancies (unpublished data).
Our protein data differ from those of others [60] who have demonstrated a significant increase in LHCGR protein in T21 and T18 placentas. However, this apparent contradiction is reconciled by considering that immunohistochemical staining of tissue sections using a polyclonal primary antibody raised against the common amino-terminal 15–38 residues of the LHCGR peptide [60] would stain both mature LHCGR and non-transducing LHCGR isoforms produced from alternatively spliced LHCGR mRNAs that have common N-terminal sequences (Figure 2, and [20,21,42,43]). Therefore, immunohistochemistry might be insufficient to distinguish mature LHCGR from its truncated isoforms (Fig. 6).
Alternative promoter use [20,61], and differential splicing of mRNA to produce various mRNA species [21,41-43,62] and multiple protein isoforms (Figs. 5 and 6, and [27,28,48]) are the hallmarks of molecular regulation of the LHCGR gene. While our work implies that there could be an increase in transcriptional initiation from the LHCGR gene in DS placenta, further work is needed to establish whether different promoters are utilized in DS compared to physiologically normal pregnancies. This could be important because in the transgenic mouse model developed by Huhtaniemi's group [61] there appears to be a link between alternative promoter utilization and differential splicing in transgenes. It is interesting to note that the multiple LHCGR protein isoforms (Figs. 5 and 6, and [28]) detected in placental extracts by Western blot are also expressed in LH-induced human neuroblastoma cells [48]. How many of these isoforms are capable of sequestering hCG by ligand binding is currently under investigation.
The clinical relevance of this report stems from the significance of hCG in establishing and maintaining placental/fetal development in human pregnancy. Given the recent discovery of an unanticipated role for LH/hCG in the distribution of cerebral blood flow [63], neurosteroidogenesis and fetal development of sensory and autonomic functions [64], the reduced expression of functional LHCGR protein in placenta may have far-reaching consequences. Indeed, the pathological activation of cerebral microglial cells abundantly expressing LHCGR, has been linked to Alzheimer's and other neurodegenerative diseases with high circulating LH [48,65]. Some outstanding questions remain to be answered, including whether the reduced expression of the functional LHCGR isoforms in placenta described here reflects a similar reduced expression in the fetal brain that might affect sensory and autonomic development in DS babies, and whether the reduction in functional LHCGR expression can be attributed to somatic mutations or an extra copy of chromosome 21.
Acknowledgements
We are grateful to the numerous patients at King's for appreciating this research and kindly providing consent to obtain CVS, to Dr. Vanessa Sangala at King's for providing placental tissues from very early pregnancies, to Dr. Axel Themmen for providing LHR expressing HEK 293 cell line, to Dr. Alan Hardy at King's for providing laboratory space at the initial stage of this study. We appreciate the generous financial support of the Fetal Medicine Foundation, UK.
==== Refs
Hassold TJ Jacobs PA Trisomy in man Ann Rev Genet 1983 18 69 97 6241455 10.1146/annurev.ge.18.120184.000441
Warburton D Fraser FC Spontaneous abortion risks in man: data from reproductive histories collected in a medical genetics unit Am J Hum Genet 1964 16 1 25 14131871
Hassold T Chen N Funkhouser J Jooss T Manuel B Matsuura J Matsuyama A Wilson C Yamane JA Jacobs PA A cytogenetic study of 1000 spontaneous abortions Ann Hum Genet 1980 44 151 178 7316468
Gras L McBain J Trounson A Kola I The incidence of chromosomal aneuploidy in stimulated and unstimulated (natural) uninseminated human oocytes Hum Reprod 1992 7 1396 1401 1291565
Hook EB Cross PK Schreinemachers DM Chromosomal abnormality rates at amniocentesis and in live-born infants JAMA 1983 249 2034 2038 6220164 10.1001/jama.249.15.2034
Goldstein H Nielsen KG Rates and survival of individuals with trisomy 13 and 18. Data from a 10-year period in Denmark Clin Genet 1988 34 366 372 3233784
Baty BJ Blackburn BL Carey JC Natural history of trisomy 18 and trisomy 13: I. Growth, physical assessment, medical histories, survival, and recurrence risk Am J Med Genet 1994 49 175 188 8116665 10.1002/ajmg.1320490204
Nicolaidis P Petersen MB Origin and mechanisms of non-disjunction in human autosomal trisomies Hum Reprod 1998 13 313 319 9557829 10.1093/humrep/13.2.313
Wisniewski KE Wisniewski HM Wen GY Occurrence of neuropathological changes and dementia of Alzheimer's disease in Down's syndrome Ann Neurol 1985 17 278 282 3158266 10.1002/ana.410170310
Hitzler JK Zipursky A Origins of leukaemia in children with Down syndrome Nat Rev Cancer 2005 5 11 20 15630411 10.1038/nrc1525
Menendez M Down syndrome, Alzheimer's disease and seizures Brain Dev 2005 27 246 252 15862185 10.1016/j.braindev.2004.07.008
Spencer K Souter V Tul N Snijders R Nicolaides KH A screening program for trisomy 21 at 10–14 weeks using fetal nuchal translucency, maternal serum free beta-human chorionic gonadotropin and pregnancy-associated plasma protein-A Ultrasound Obstet Gynecol 1999 13 231 237 10341399 10.1046/j.1469-0705.1999.13040231.x
Hoshina M Boothby M Boime I Cytological localization of chorionic gonadotropin alpha and placental lactogen mRNAs during development of the human placenta J Cell Biol 1982 93 190 198 6896058 10.1083/jcb.93.1.190
Licht P Russu V Wildt L On the role of human chorionic gonadotropin (hCG) in the embryo-endometrial microenvironment: implications for differentiation and implantation Semin Reprod Med 2001 19 37 47 11394202 10.1055/s-2001-13909
Bahado-Singh RO Oz AU Kingston JM Shahabi S Hsu CD Cole L The role of hyperglycosylated hCG in trophoblast invasion and the prediction of subsequent pre-eclampsia Prenat Diagn 2002 22 478 481 12116306 10.1002/pd.329
Muller F Savey L Le Fiblec B Bussieres L Ndayizamba G Colau JC Giraudet P Maternal serum human chorionic gonadotropin level at fifteen weeks is a predictor for preeclampsia Am J Obstet Gynecol 1996 175 37 40 8694072
Spencer K Evaluation of an assay of the free beta-subunit of choriogonadotropin and its potential value in screening for Down's syndrome Clin Chem 1991 37 809 814 1710952
Spencer K Macri JN Aitken DA Connor JM Free beta-hCG as first-trimester marker for fetal trisomy Lancet 1992 339 1480 1376387 10.1016/0140-6736(92)92073-O
Rousseau-Merck MF Misrahi M Atger M Loosfelt H Milgrom E Berger R Localization of the human luteinizing hormone/choriogonadotropin receptor gene (LHCGR) to chromosome 2p21 Cytogenet Cell Genet 1990 54 77 79 2249480
Atger M Misrahi M Sar S Le FL Dessen P Milgrom E Structure of the human luteinizing hormone-choriogonadotropin receptor gene: unusual promoter and 5' non-coding regions Mol Cell Endocrinol 1995 111 113 123 7556872 10.1016/0303-7207(95)03557-N
Loosfelt H Misrahi M Atger M Salesse R Vu Hai-Luu Thi MT Jolivet A Guiochon-Mantel A Sar S Jallal B Garnier J Cloning and sequencing of porcine LH-hCG receptor cDNA: variants lacking transmembrane domain Science 1989 245 525 528 2502844
McFarland KC Sprengel R Phillips HS Kohler M Rosemblit N Nikolics K Segaloff DL Seeburg PH Lutropin-choriogonadotropin receptor: an unusual member of the G protein-coupled receptor family Science 1989 245 494 499 2502842
Clouser CL Menon KM N-linked glycosylation facilitates processing and cell surface expression of rat luteinizing hormone receptor Mol Cell Endocrinol 2005 235 11 19 15866423 10.1016/j.mce.2005.02.005
Ascoli M Fanelli F Segaloff DL The lutropin/choriogonadotropin receptor, a 2002 perspective Endocr Rev 2002 23 141 174 11943741 10.1210/er.23.2.141
Hipkin RW Sanchez-Yague J Ascoli M Identification and characterization of a luteinizing hormone/chorionic gonadotropin (LH/CG) receptor precursor in a human kidney cell line stably transfected with the rat luteal LH/CG receptor complementary DNA Mol Endocrinol 1992 6 2210 2218 1491699 10.1210/me.6.12.2210
Min L Ascoli M Effect of activating and inactivating mutations on the phosphorylation and trafficking of the human lutropin/choriogonadotropin receptor Mol Endocrinol 2000 14 1797 1810 11075813 10.1210/me.14.11.1797
VuHai-LuuThi MT Misrahi M Houllier A Jolivet A Milgrom E Variant forms of the pig lutropin/choriogonadotropin receptor Biochemistry 1992 31 8377 8383 1326331 10.1021/bi00150a035
Bukovsky A Indrapichate K Fujiwara H Cekanova M Ayala ME Dominguez R Caudle MR Wimalsena J Elder RF Copas P Foster JS Fernando RI Henley DC Upadhyaya NB Multiple luteinizing hormone receptor (LHR) protein variants, interspecies reactivity of anti-LHR mAb clone 3B5, subcellular localization of LHR in human placenta, pelvic floor and brain, and possible role for LHR in the development of abnormal pregnancy, pelvic floor disorders and Alzheimer's disease Reprod Biol Endocrinol 2003 1 46 12816543 10.1186/1477-7827-1-46
Tsai-Morris CH Buczko E Wang W Dufau ML Intronic nature of the rat luteinizing hormone receptor gene defines a soluble receptor subspecies with hormone binding activity J Biol Chem 1990 265 19385 19388 2174034
Kremer H Kraaij R Toledo SPA Post M Fridman JB Hayashida CY van Reen M Milgrom E Ropers HH Mariman E Themmen APN Brunner HG Male pseudohermaphroditism due to a homozygous missense mutation of the luteinizing hormone receptor gene Nature Genet 1995 9 160 164 7719343 10.1038/ng0295-160
Gromoll J Eiholzer U Nieschlag E Simoni M Male hypogonadism caused by homozygous deletion of exon 10 of the luteinizing hormone (LH) receptor: differential action of human chorionic gonadotropin and LH J Clin Endocrinol Metab 2000 85 2281 2286 10852464 10.1210/jc.85.6.2281
Zhang FP Poutanen M Wilbertz J Huhtaniemi I Normal prenatal but arrested postnatal sexual development of luteinizing hormone receptor knockout (LuRKO) mice Mol Endocrinol 2001 15 172 183 11145748 10.1210/me.15.1.172
Rao CV Lei ZM Consequences of targeted inactivation of LH receptors Mol Cell Endocrinol 2002 187 57 67 11988312 10.1016/S0303-7207(01)00694-3
Dorman SE Holland SM Mutation in the signal-transducing chain of the interferon-gamma receptor and susceptibility to mycobacterial infection J Clin Invest 1998 101 2364 2369 9616207
Fieschi C Dupuis S Picard C Smith CI Holland SM Casanova JL High levels of interferon gamma in the plasma of children with complete interferon gamma receptor deficiency Pediatrics 2001 107 E48 11335769 10.1542/peds.107.4.e48
Banerjee S Smallwood A Chambers AE Nicolaides K Quantitative recovery of immunoreactive proteins from clinical samples following RNA and DNA isolation Biotechniques 2003 35 450 2, 454, 456 14513546
Smallwood A Papageorghiou A Nicolaides K Alley MK Jim A Nargund G Ojha K Campbell S Banerjee S Temporal regulation of the expression of syncytin (HERV-W), maternally imprinted PEG10, and SGCE in human placenta Biol Reprod 2003 69 286 293 12620933 10.1095/biolreprod.102.013078
Banerjee S Smallwood A Moorhead J Chambers AE Papageorghiou A Campbell S Nicolaides K Placental expression of interferon-gamma (IFN-gamma) and its receptor IFN-gamma R2 fail to switch from early hypoxic to late normotensive development in preeclampsia J Clin Endocrinol Metab 2005 90 944 952 15585559 10.1210/jc.2004-1113
Spencer K Berry E Crossley JA Aitken DA Nicolaides KH Is maternal serum total hCG a marker of trisomy 21 in the first trimester of pregnancy? Prenat Diagn 2000 20 311 317 10740204 10.1002/(SICI)1097-0223(200004)20:4<311::AID-PD809>3.0.CO;2-P
Spencer K Crossley JA Aitken DA Nix AB Dunstan FD Williams K Temporal changes in maternal serum biochemical markers of trisomy 21 across the first and second trimester of pregnancy Ann Clin Biochem 2002 39 567 576 12564838 10.1258/000456302760413342
Minegishi T Tano M Abe Y Nakamura K Ibuki Y Miyamoto K Expression of luteinizing hormone/human chorionic gonadotrophin (LH/HCG) receptor mRNA in the human ovary Mol Hum Reprod 1997 3 101 107 9239715 10.1093/molehr/3.2.101
Reinholz MM Zschunke MA Roche PC Loss of alternately spliced messenger RNA of the luteinizing hormone receptor and stability of the follicle-stimulating hormone receptor messenger RNA in granulosa cell tumors of the human ovary Gynecol Oncol 2000 79 264 271 11063655 10.1006/gyno.2000.5946
Madhra M Gay E Fraser HM Duncan WC Alternative splicing of the human luteal LH receptor during luteolysis and maternal recognition of pregnancy Mol Hum Reprod 2004 10 599 603 15169923 10.1093/molehr/gah076
Xie YB Wang H Segaloff DL Extracellular domain of lutropin/choriogonadotropin receptor expressed in transfected cells binds choriogonadotropin with high affinity J Biol Chem 1990 265 21411 21414 2254302
Bacich DJ Rohan RM Norman RJ Rodgers RJ Characterization and relative abundance of alternatively spliced luteinizing hormone receptor messenger ribonucleic acid in the ovine ovary Endocrinology 1994 135 735 744 7518389 10.1210/en.135.2.735
Koo YB Ji I Ji TH Characterization of different sizes of rat luteinizing hormone/chorionic gonadotropin receptor messenger ribonucleic acids Endocrinology 1994 134 19 26 8275933 10.1210/en.134.1.19
Meduri G Charnaux N Loosfelt H Jolivet A Spyratos F Brailly S Milgrom E Luteinizing hormone/human chorionic gonadotropin receptors in breast cancer Cancer Res 1997 57 857 864 9041186
Bowen RL Verdile G Liu T Parlow AF Perry G Smith MA Martins RN Atwood CS Luteinizing hormone, a reproductive regulator that modulates the processing of amyloid-beta precursor protein and amyloid-beta deposition J Biol Chem 2004 279 20539 20545 14871891 10.1074/jbc.M311993200
de RN Milgrom E Inherited disorders of GnRH and gonadotropin receptors Mol Cell Endocrinol 2001 179 83 87 11420132 10.1016/S0303-7207(01)00471-3
Wu SM Leschek EW Rennert OM Chan WY Luteinizing hormone receptor mutations in disorders of sexual development and cancer Front Biosci 2000 5 D343 D352 10704433
Brizot ML Jauniaux E Mckie AT Farzaneh F Nicolaides KH Placental expression of alpha and beta subunits of human chorionic gonadotrophin in early pregnancies with Down's syndrome Hum Reprod 1995 10 2506 2509 8530698
Eldar-Geva T Hochberg A deGroot N Weinstein D High maternal serum chorionic gonadotropin level in Downs' syndrome pregnancies is caused by elevation of both subunits messenger ribonucleic acid level in trophoblasts J Clin Endocrinol Metab 1995 80 3528 3531 8530594 10.1210/jc.80.12.3528
Goshen R Gonik B Ariel I Weiss Y de-Groot N Hochberg A High levels of maternal serum human chorionic gonadotropin in Down syndrome pregnancies: the possible role of a transcription factor on chromosome 21 Fetal Diagn Ther 1999 14 106 111 10085509 10.1159/000020899
Frendo JL Vidaud M Guibourdenche J Luton D Muller F Bellet D Giovagrandi Y Tarrade A Porquet D Blot P Evain-Brion D Defect of villous cytotrophoblast differentiation into syncytiotrophoblast in Down's syndrome J Clin Endocrinol Metab 2000 85 3700 3707 11061527 10.1210/jc.85.10.3700
Goshen R What factors regulate HCG production in Down's syndrome pregnancies?. Screening for Down's syndrome using HCG concentrations – a common practice but still an enigma Mol Hum Reprod 1999 5 893 895 10508214 10.1093/molehr/5.10.893
Genbacev O Zhou Y Ludlow JW Fisher SJ Regulation of human placental development by oxygen tension Science 1997 277 1669 1672 9287221 10.1126/science.277.5332.1669
Hoang VM Foulk R Clauser K Burlingame A Gibson BW Fisher SJ Functional proteomics: examining the effects of hypoxia on the cytotrophoblast protein repertoire Biochemistry 2001 40 4077 4086 11300788 10.1021/bi0023910
Kliman HJ Feinberg RF Human trophoblast-extracellular matrix (ECM) interactions in vitro: ECM thickness modulates morphology and proteolytic activity Proc Natl Acad Sci U S A 1990 87 3057 3061 2326266
Damsky CH Fisher SJ Trophoblast pseudo-vasculogenesis: faking it with endothelial adhesion receptors Curr Opin Cell Biol 1998 10 660 666 9818178 10.1016/S0955-0674(98)80043-4
Jauniaux E Bao S Eblen A Li X Lei ZM Meuris S Rao CV HCG concentration and receptor gene expression in placental tissue from trisomy 18 and 21 Mol Hum Reprod 2000 6 5 10 10611254 10.1093/molehr/6.1.5
Hamalainen T Kero J Poutanen M Huhtaniemi I Transgenic mice harboring murine luteinizing hormone receptor promoter/beta-galactosidase fusion genes: different structural and hormonal requirements of expression in the testis, ovary, and adrenal gland Endocrinology 2002 143 4096 4103 12239121 10.1210/en.2002-220162
Nakamura K Yamashita S Omori Y Minegishi T A splice variant of the human luteinizing hormone (LH) receptor modulates the expression of wild-type human LH receptor Mol Endocrinol 2004 18 1461 1470 15031322 10.1210/me.2003-0489
Hershkovitz R Erez O Sheiner E Landau D Mankuta D Mazor M Elevated maternal mid-trimester chorionic gonadotropin > or = 4 MoM is associated with fetal cerebral blood flow redistribution Acta Obstet Gynecol Scand 2003 82 22 27 12580835 10.1034/j.1600-0412.2003.820104.x
Apaja PM Harju KT Aatsinki JT Petaja-Repo UE Rajaniemi HJ Identification and structural characterization of the neuronal luteinizing hormone receptor associated with sensory systems J Biol Chem 2004 279 1899 1906 14581462 10.1074/jbc.M311395200
Liu B Hong JS Role of microglia in inflammation-mediated neurodegenerative diseases: mechanisms and strategies for therapeutic intervention J Pharmacol Exp Ther 2003 304 1 7 12490568 10.1124/jpet.102.035048
|
15969756
|
PMC1190215
|
CC BY
|
2021-01-04 16:37:11
|
no
|
Reprod Biol Endocrinol. 2005 Jun 21; 3:25
|
utf-8
|
Reprod Biol Endocrinol
| 2,005 |
10.1186/1477-7827-3-25
|
oa_comm
|
==== Front
Reprod Biol EndocrinolReproductive biology and endocrinology : RB&E1477-7827BioMed Central London 1477-7827-3-291601881710.1186/1477-7827-3-29ResearchLocal versus systemic effect of ovulation-inducing factor in the seminal plasma of alpacas Ratto Marcelo H [email protected] Wilfredo [email protected] Jaswant [email protected] Gregg P [email protected] Veterinary Biomedical Sciences, Western College of Veterinary Medicine, University of Saskatchewan, S7N 5B4, Canada2 Faculty of Veterinary Medicine, San Marcos University, Lima, Peru2005 14 7 2005 3 29 29 24 5 2005 14 7 2005 Copyright © 2005 Ratto et al; licensee BioMed Central Ltd.2005Ratto et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Camelids are induced (reflex) ovulators. We have recently documented the presence of an ovulation-inducing factor (OIF) in the seminal plasma of alpacas and llamas. The objective was to test the hypothesis that OIF exerts its effect via a systemic rather than a local route and that endometrial curettage will enhance the ovulatory response to intrauterine deposition of seminal plasma in alpacas.
Methods
Female alpacas were assigned randomly to 6 groups (n = 15 to 17 per group) in a 2 × 3 factorial design to test the effect of seminal plasma versus phosphate-buffered saline (PBS) given by intramuscular injection, by intrauterine infusion, or by intrauterine infusion after endometrial curettage. Specifically, alpacas in the respective groups were given 1) 2 ml of alpaca seminal plasma intramuscularly, 2) 2 ml of PBS intramuscularly (negative control group), 3) 2 ml of alpaca seminal plasma by intrauterine infusion, 4) 2 ml of PBS by intrauterine infusion (negative control group), 5) 2 ml of alpaca seminal plasma by intrauterine infusion after endometrial curettage, or 6) 2 ml of PBS by intrauterine infusion after endometrial curettage (negative control group). The alpacas were examined by transrectal ultrasonography to detect ovulation and measure follicular and luteal diameters.
Results
Intramuscular administration of seminal plasma resulted in a higher ovulation rate than intrauterine administration of seminal plasma (93% versus 41%; P < 0.01), while intrauterine seminal plasma after endometrial curettage was intermediate (67%). None of the saline-treated controls ovulated. The diameter of the CL after treatment-induced ovulation was not affected by the route of administration of seminal plasma.
Conclusion
We conclude that 1) OIF in seminal plasma effects ovulation via a systemic rather than a local route, 2) disruption of the endometrial mucosa by curettage facilitated the absorption of OIF and increased the ovulatory effect of seminal plasma, and 3) ovulation in alpacas is not associated with a physical stimulation of the genital tract, and 4) the alpaca represents an excellent biological model to evaluate the bioactivity of OIF.
==== Body
Background
Early studies of South American camelids documented that copulatory stimulation is responsible for inducing ovulation in these species [1,2]. The first significant increase in plasma LH concentrations occurred 15–40 minutes after the initiation of mating in llamas and alpacas [3,4]. A similar LH increase was observed in Bactrian and dromedary camels (related camelid species) beginning 20–30 min after mating [5,6]. The rapid increase in plasma LH concentration after mating in camelids resembles that observed in rabbits [7] and cats [8] – also classified as induced ovulators. A 40-fold increase in GnRH secretion from the medio-basal hypothalamus was detected within 20–60 minutes of mating in rabbits [9], followed by a preovulatory LH surge and ultimately ovulation about 10 hours after mating [10].
The primary mechanism responsible for ovulation induction in these species is thought to involve a neuro-endocrine response to physical stimulation of the vagina and cervix by the penis during mating [11]. The results of recent studies in llamas and alpacas, however, provide support for the hypothesis that a chemical substance in the semen is responsible, in whole or in part, for inducing ovulation [12]. The existence of a potent ovulation-inducing factor (OIF) was demonstrated by intramuscular administration of cell-free llama and alpaca seminal plasma to females of the respective species. Collectively, 28 of 30 (93%) females ovulated after seminal plasma treatment compared to 0 of 32 (0%) saline-treated controls [12].
The discovery of OIF in llamas and alpacas is consistent with an early study in which intrauterine or intramuscular administration of Bactrian semen induced ovulation in Bactrian females [13,5]. However, conflicting results have been reported about the effect of local versus systemic administration of semen. In female alpacas that were mounted by a male (intromission prevented) and those that were mounted (intromission prevented) followed by artificial insemination, ovulation (detected at necropsy 3 days later) occurred in 2/15 and 3/9, respectively [1]. Since ovulation occurred in 36/44 females after natural copulation, the authors concluded that the physical act of coitus was responsible for eliciting ovulation in alpacas. In contrast, ovulation was detected in 6/10 alpacas and 5/8 llamas inseminated intravaginally with conspecific semen cited in [14]. In Bactrian camels, ovulation was detected by rectal palpation after intravaginal or intrauterine infusion of whole semen or seminal plasma in ≥ 75% of females [13,5,15]. In a recent ultrasonographic study [12], ovulation was detected in 13 of 14 alpacas given seminal plasma intramuscularly, but in 0 of 12 given seminal plasma by transcervical intrauterine deposition.
The reason for the disparity in results is not clear, but authors of the latter study [12] speculated that differences may be attributed to attenuated absorption of OIF from the genital mucosa compared to the muscle. In this regard, copulation in alpacas and llamas is a prolonged event (30 to 50 minutes) [16,3] and ejaculation is intrauterine [17]. A normal sequela of copulation in these species is acute, transient inflammation of the endometrium as a result of repeated abrasion by the penis [17]. Perhaps absorption of OIF in seminal plasma subsequent to natural mating is facilitated by the hyperemia of the excoriated endometrium.
The objective of the present study was to test the hypothesis that OIF exerts its effect via a systemic rather than a local route and that endometrial curettage will enhance the ovulatory response to intrauterine deposition of seminal plasma in alpacas. A 2-by-3 factorial design was used to compare the ovulatory effects of alpaca seminal plasma versus phosphate-buffered saline (control) administered by intramuscular injection, by intrauterine deposition, or by intrauterine deposition after endometrial curettage.
Methods
Seminal plasma collection
Semen was collected from male alpacas (n = 8) by artificial vagina [18] over a period of 2 months prior to the start of the experiment (10 ejaculates per animal) and processed as previously described [12]. Briefly, ejaculates were diluted 1:1 (v/v) with phosphate buffered saline (PBS, Gibco, Grand Island, N.Y., USA) and centrifuged for 30 minutes at 1500 × g. The supernatant was decanted to remove spermatozoa and a drop was evaluated by microscopy to confirm the absence of cells. If spermatozoa were detected, the sample was centrifuged again in like manner until all spermatozoa were removed. Sperm-free seminal plasma was stored at -70°C. Upon thawing, the diluted seminal plasma was pooled and kanamycin sulfate (Sigma Chemical Co., St Louis, MO, USA) was added to a final concentration of 25 μg/ml.
Animals & Treatments
The study was conducted during February to March at the Quimsachata Research Station in the Department of Puno, Peru (15°S, 71°W, and 4,500 m above sea level) using mature non-lactating female alpacas ≥ 4 years of age and weighing an average of 75 kg. To facilitate data collection, ovarian follicular development was synchronized among females (n = 100) by giving 5 mg Armour Standard LH (Lutropin-V®, Bioniche Animal Health, Belleville, ON, Canada) to induce ovulation. We expected approximately 85% to 90% of the alpacas to ovulate after LH treatment, resulting in synchronous emergence of a new follicular wave 2 days after treatment [19]. Alpacas were examined by transrectal ultrasonography (Aloka 500 with a 7.5 MHz linear-array probe, Instruments for Science & Medicine Inc., Vancouver, BC, Canada) 12 days after LH treatment – sufficient time to permit complete luteal regression and growth of a new dominant follicle [19,20]. Alpacas with a follicle ≥ 8 mm in diameter (n = 92) were assigned randomly to 6 groups and given: 1) 2 ml of alpaca seminal plasma intramuscularly (n = 15), 2) 2 ml of PBS intramuscularly (control; n = 15), 3) 2 ml of alpaca seminal plasma by intrauterine infusion (n = 17), 4) 2 ml of PBS by intrauterine infusion (control; n = 15), 5) 2 ml of alpaca seminal plasma by intrauterine infusion after endometrial curettage (n = 15), or 6) 2 ml of PBS by intrauterine infusion after endometrial curettage (control; n = 15). Intramuscular injections were given in the semimembranosus muscle using a 20-gauge 40 mm long needle. Intrauterine infusions were accomplished by passing a plastic pipette through the cervix via transrectal manipulation and depositing 1 ml of alpaca seminal plasma or PBS into each uterine horn. To mimic the transient inflammation of the endometrium caused by the penis during natural mating [17], both uterine horns were curettaged before intrauterine infusion by repeatedly scraping the tip of the plastic infusion pipette back and forth over the surface of the endometrial of both uterine horns for 3 minutes. Curettage was accomplished by transrectal manipulation of the uterus with one hand and manipulation of the pipette with the other.
Alpacas were examined by transrectal ultrasonography on Day 2 (Day 0 = treatment) to detect ovulation. Ovulation was defined as the sudden disappearance of a large follicle (≥ 8 mm) that was detected during the previous examination [21]. To confirm ovulation and assess corpus luteum (CL) development, transrectal ultrasonography was repeated on Day 8; i.e., expected time of maximum CL diameter [21,20].
Statistical Analyses
Single-point measurements (i.e., follicle size at the time of treatment, maximum CL diameter) were compared among groups by analyses of variance. If the overall effect was significant (P < 0.05), specific comparisons were made between groups using Tukey multiple comparisons. Ovulation rates were compared among groups by chi-square analysis.
Results and Discussion
The diameter of the largest follicle at the time of treatment did not differ among groups (P = 0.9). Ovulations were observed in groups treated by intramuscular administration or intrauterine deposition of seminal plasma (Table 1). Ovulation and luteal development were not detected in females that were given PBS by intramuscular or intrauterine administration (control groups). The ovulation rate in the intramuscular group (93%) was higher (P < 0.01) than in the intrauterine group (41%), while the endometrial curettage group was intermediate (67%). Of the alpacas that ovulated, the diameter of the CL did not differ among groups.
Table 1 Effect of administration of alpaca seminal plasma administered intramuscularly or by intrauterine infusion with or without endometrial curettage on ovulation and corpus luteum formation (mean ± SEM) in female alpacas.
Intramuscular Intrauterine Intrauterine with curettage
Seminal plasma Phosphate buffered saline Seminal plasma Phosphate buffered saline Seminal plasma Phosphate buffered saline
Follicle diameter at treatment (mm)* 8.0 ± 0.3 (n = 15) 8.2 ± 0.3 (n = 15) 8.1 ± 0.3 (n = 17) 8.0 ± 0.3 (n = 15) 8.3 ± 0.2 (n = 15) 8.4 ± 0.3 (n = 15)
Ovulation rate (%) 14/15a (93%) 0/15c (0%) 7/17b (41%) 0/15c (0%) 10/15ab (67%) 0/15c (0%)
CL diameter (mm) on Day 8 (Day 0 = treatment)* 9.3 ± 0.4 (n = 14) ---- 9.5 ± 0.3 (n = 7) ---- 9.4 ± 0.4 (n = 10) ----
* No difference among groups (P ≥ 0.9)
a,b,c Proportions with different superscripts are different (P < 0.01)
The results of the present study provide support for the hypothesis that the ovulation-inducing effect of seminal plasma is mediated via a systemic rather than a local route. A higher ovulation rate in alpacas treated by intrauterine infusion would have provided evidence to the contrary, but the ovulation rate was significantly lower in the intrauterine infusion group than in the intramuscular group. These results are consistent with those of a previous study [12] in which intramuscular administration of llama seminal plasma was followed by a surge in plasma LH concentration and ovulation. However, results do not unequivocally rule out a potential local contribution of seminal plasma to ovulation induction. In this regard, results of a study of the effects of boar seminal plasma deposited into different segments of the uterine horn in gilts were suggestive of a local unilateral mechanism influencing the interval to ovulation [22]. Ovulation was advanced in the ovary ipsilateral to the side of semen deposition, but interestingly, only when deposited near the utero-tubal junction; no effect was found when seminal plasma was deposited in the middle of the uterine horn between two ligatures. No information has been reported regarding circulating gonadotropin concentrations subsequent to intrauterine or intravaginal deposition of semen.
The results are also consistent with the concept that systemic absorption of OIF from the uterus is facilitated by endometrial curettage. The ovulation rate in the curettage group was intermediate between that of the intramuscular group and the intrauterine group without curettage. Endometrial curettage in the present study was mild and was accomplished by rubbing a smooth, round-tipped plastic infusion pipette against the endometrium for 3 minutes. Perhaps more aggressive curettage would induce sufficient endometrial inflammation to increase absorption of OIF and result in an ovulation rate more typical of natural mating during the period of follicular readiness (i.e., 90%; [23]).
The disparity between the present study and our previous study [12] in the effect of intrauterine treatment in non-curettaged alpacas (ovulation rate of 41% versus 0%, respectively) may be attributed to the dose and site of deposition of seminal plasma. A total of 2 ml of seminal plasma was infused in the uterine horns (1 ml in each horn) in the present study, while only 1 ml of seminal plasma was infused into the uterine body in the previous study [12]. Regarding local versus intramuscular absorption, the addition of a GnRH analogue (Buserelin) to the semen induced ovulation in rabbits after intravaginal artificial insemination [24], but the dose of GnRH required for ovulation induction by intravaginal deposition was ten times higher than that used by intramuscular administration in the control group (8 μg versus 0.8 μg per inseminated female). This is consistent with the results from our previous experiment [12] in which no ovulations were detected in alpacas after intrauterine deposition of 5 mg of LH (Lutropin), a dose that caused ovulation in more that 80% of the females when given intramuscularly [19,20]. Hence, higher systemic concentrations of OIF may have been achieved in the present study by using larger dose and causing greater dispersion of seminal plasma throughout the endometrial surface. No mention was made regarding uterine manipulations in previous studies in llamas and alpacas [1,14] or Bactrian camels [13,5,15], and it is unclear if semen was deposited into the vagina, the cervix, or the uterus.
Results did not support the notion that physical stimulation of the vagina, cervix and uterus is involved in a neuro-endocrine system for ovulation induction, nor was there any evidence that OIF is produced by tissues of the female reproductive tract. Despite purposeful manipulation and irritation of the genitalia in the present study, none of the 45 females treated with saline alone ovulated.
Conclusion
We conclude that 1) OIF in seminal plasma effects ovulation via a systemic rather than a local route, 2) disruption of the endometrial mucosa by curettage facilitated the absorption of OIF and increased the ovulatory effect of seminal plasma, and 3) ovulation in alpacas is not associated with a physical stimulation of the genital tract, and 4) the alpaca represents an excellent biological model to evaluate the bioactivity of OIF.
Acknowledgements
This research was supported by grants from the Canadian Llama and Alpaca Association, the Alpaca Research Foundation, and the Inter-American Institute for Cooperation in Agriculture. Animals and animal maintenance were provided by Quimsachata Research Station under the sponsorship of San Marcos University of Lima, Peru. We gratefully acknowledge Bioniche Animal Health Canada Inc. for providing Lutropin.
==== Refs
Fernandez-Baca S Madden DHL Novoa C Effect of different mating stimuli on induction of ovulation in the alpaca Journal of Reproduction and Fertility 1970 22 261 267 5464117
England BG Foote WC Mathews DH Cardozo AG Riera S Ovulation and corpus luteum function in the llama (Lama glama) Journal of Endocrinology 1969 45 505 513 5366111
Bravo PW Fowler ME Stabenfeldt GH Endocrine response in the llama to copulation Theriogenology 1990 33 891 899 16726785 10.1016/0093-691X(90)90824-D
Bravo PW Stabenfeldt GH Lasley BL Fowler ME The effect of ovarian follicular size on pituitary and ovarian responses to copulation in domesticated South American camelids Biology of Reproduction 1991 45 553 559 1751629
Xu YS Wang HY Zeng GQ Jiang GT Gao HY Hormone concentrations before and after semen-induced ovulation in the Bactrian camel (Camelus bactrianus) Journal of Reproduction and Fertility 1985 74 341 346 4045809
Marie M Anouassi A Mating induced luteinizing hormone surge and ovulation in the female camel (Camelus dromedarius) Biology of Reproduction 1986 35 792 798 3814695
Jones EF Bain JB Odell WD Postcoital luteinizing hormone release in male and female rabbits as determined by radioimmunoassay Fertility and Sterility 1976 27 848 852 950052
Concanon P Hodgson B Lein D Reflex LH release in estrous cats following single and multiple copulations Biology of Reproduction 1980 23 111 117 7191335
Spies HG Pau KYF Yang SP Coital and estrogen signals: A contrast in the preovulatory neuroendocrine networks of rabbits and rhesus monkeys Biology of Reproduction 1997 56 310 319 9116126
Milligan SR Induced ovulation in mammals Oxford Reviews of Reproductive Biology 1982 4 1 46
Bakker J Baum MJ Neuroendocrine regulation of GnRH release in induce ovulators Frontiers in Neuroendocrinology 2000 21 220 262 10882541 10.1006/frne.2000.0198
Adams GP Ratto MH Huanca W Singh J Ovulation-inducing factor in the seminal plasma of alpacas and llamas Biology of Reproduction 2005 Published on May 11, 2005A as DOI:10.1095 /biolreprod.105.040097
Chen BX Yuen ZX Pan GW Semen induced ovulation in the Bactrian camel (Camelus bactrianus) Journal of Reproduction and Fertility 1985 73 335 339 3900379
Sumar J Effects of various ovulation induction stimuli in alpacas and llamas Journal of Arid Environment 1994 26 39 45 10.1006/jare.1994.1006
Pan G Zhao X Chen S Jiang S Huang Y Zu Y Wang H Allen WR, Higgins AJ, Mayhew IG, Snow D, Wade JF The ovulation-inducing effect of seminal plasma in the bactrian camel 1992 Proc First International Camel Conf. R&W Publications, Newmarket 159 161
San-Martin M Copaira M Zuniga J Rodreguez R Bustinza G Acosta L Aspects of reproduction in the alpaca Journal of Reproduction and Fertility 1968 16 395 399 5673739
Bravo PW Moscoso J Ordonez C Alarcon V Transport of spermatozoa and ova in female alpaca Animal Reproduction Science 1996 43 173 179 10.1016/0378-4320(95)01465-9
Bravo PW Flores U Garnica J Ordonez C Collection of semen and artificial insemination of alpacas Theriogenology 1997 47 619 626 16728014 10.1016/S0093-691X(97)00020-4
Ratto MH Singh J Huanca W Adams GP Ovarian follicular wave synchronization and pregnancy rate after fixed-time natural mating in llamas Theriogenology 2003 60 1645 1656 14580647 10.1016/S0093-691X(03)00176-6
Ratto M Huanca W Singh J Adams GP Comparison of the effect of natural mating, LH, and GnRH on interval to ovulation and luteal function in llamas Animal Reproduction Science 2005
Adams GP Griffin PG Ginther OJ In situ morphologic dynamics of ovaries, uterus, and cervix in llamas Biology of Reproduction 1989 41 551 558 2686763
Waberski D Kremer H Borchardt Neto G Jungblut PW Kallweit E Weitze KF Studies on a local effect of boar seminal plasma on ovulation time in gilts Journal of Veterinary Medicine 1999 46 431 438 10.1046/j.1439-0442.1999.00230.x
Adams GP Sumar J Ginther OJ Effects of lactational and reproductive status on ovarian follicular waves in llamas (lama glama) Journal of Reproduction and Fertility 1990 90 535 545 2250251
Quintela L Pena A Vega MD Gullon J Prieto MC Barrio M Becerra JJ Maseda F Herradon PG Ovulation induction in rabbit does submitted to artificial insemination by adding buserelin to the seminal dose Reproduction Nutrition and Development 2004 44 79 88 10.1051/rnd:2004015
|
16018817
|
PMC1190216
|
CC BY
|
2021-01-04 16:37:12
|
no
|
Reprod Biol Endocrinol. 2005 Jul 14; 3:29
|
utf-8
|
Reprod Biol Endocrinol
| 2,005 |
10.1186/1477-7827-3-29
|
oa_comm
|
==== Front
RetrovirologyRetrovirology1742-4690BioMed Central London 1742-4690-2-411598518710.1186/1742-4690-2-41ResearchRole of viral evolutionary rate in HIV-1 disease progression in a linked cohort Mikhail Meriet [email protected] Bin [email protected] Philippe [email protected] Brenda [email protected] Anne-Mieke [email protected] M John [email protected] Nitin K [email protected] Retroviral Genetics Laboratory, Center for Virus Research, Westmead Millennium Institute, Westmead Hospital, The University of Sydney, Westmead NSW 2145. Sydney, Australia2 Department of Clinical and Epidemiological Virology, Rega Institute, Minderbroedersstraat 10, B-3000 Leuven, Belgium3 Department of Medicine, University of Calgary, 3330 Hospital Drive NW Calgary, Albert, T2N 4N1, Canada2005 29 6 2005 2 41 41 19 5 2005 29 6 2005 Copyright © 2005 Mikhail et al; licensee BioMed Central Ltd.2005Mikhail et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The actual relationship between viral variability and HIV disease progression and/or non-progression can only be extrapolated through epidemiologically-linked HIV-infected cohorts. The rarity of such cohorts accents their existence as invaluable human models for a clear understanding of molecular factors that may contribute to the various rates of HIV disease. We present here a cohort of three patients with the source termed donor A – a non-progressor and two recipients called B and C. Both recipients gradually progressed to HIV disease and patient C has died of AIDS recently. By conducting 15 near full-length genome (8.7 kb) analysis from longitudinally derived patient PBMC samples enabled us to investigate the extent of molecular factors, which govern HIV disease progression.
Results
Four time points were successfully amplified for patient A, 4 for patient B and 7 from patient C. Using phylogenetic analysis our data confirms the epidemiological-linkage and transmission of HIV-1 from a non-progressor to two recipients. Following transmission the two recipients gradually progressed to AIDS and one died of AIDS. Viral divergence, selective pressures, recombination, and evolutionary rates of HIV-1 in each member of the cohort were investigated over time. Genetic recombination and selective pressure was evident in the entire cohort. However, there was a striking correlation between evolutionary rate and disease progression.
Conclusion
Non-progressing individuals have the potential to transmit pathogenic variants, which in other host can lead to faster HIV disease progression. This was evident from our study and the accelerated disease progression in the recipient members of he cohort correlated with faster evolutionary rate of HIV-1, which is a unique aspect of this study.
==== Body
Background
The rate of HIV disease progression varies greatly among infected individuals, which is defined invariably by increasing plasma viral loads and concomitant decline in the CD4+ T cell counts. A small but rare subset of chronically-infected individuals comprising <0.8% of total HIV infected population appear to maintain high and stable CD4+ and CD8+ T cell counts, low to undetectable plasma viral loads for >10 years in the absence of antiretroviral therapy [1,2]. In addition, some of these non-progressing individuals harbor <10 copies of proviral DNA/ml blood, show strong immune responses [2,3] and a high secretion of CD8 antiviral factor(s) (CAF) [3,4]. Additionally, in rare cases there is a complete absence of viral evolution over time [5].
HIV disease is a complex interplay of both host and viral factors [6-10], but it has been difficult to derive a consensus on these factor(s) that contribute to disease progression and / or non-progression. In many cases, evidence suggests that viral gene defects contribute to non-progression of HIV disease [6,11-14], yet these molecular changes remain elusive due to the extensive inter-strain variation of HIV-1, which can be investigated using epidemiologically-linked cohorts. The rarity of such cohorts accents their existence as invaluable models for understanding how various host and viral factors govern HIV pathogenesis. For such purposes, we describe detailed molecular analyses of one such cohort comprising of 3 HIV-infected individuals (a non-progressing donor-A and two recipients B and C) whose epidemiological linkage was confirmed through phylogenetic analyses [15]. The donor A likely acquired HIV in 1982, and has remained healthy maintaining non-progressive status with high CD4+ and CD8+ T cell counts and with <7000 HIV-1 copies/ml of plasma. The two recipients were infected in autumn 1983 (recipient B) and in summer of 1983 (recipient C) respectively.
With the help of detailed full-length HIV-1 genome analysis over time from all cohort members, we investigated viral evolution, divergence, recombination and selective forces in contributing to HIV disease development in the two recipients as opposed to the non-progressive donor.
Results
Sequencing of near full-length genomes
Successful amplification of near full-length HIV-1 genomes was achieved from a total of 15 PBMC patient samples collected between 1992 to 2000 from all 3 cohort members A, B and C. Epidemiological-linkage was confirmed by maximum likelihood phylogenetic analysis which was subsequently used for further intra patient evolutionary analysis as discussed previously in Mikhail et al., 2005 [15].
Phylogenetic clustering of cohort members: evidence of HIV transmission via blood transfusion
Within the HIV-1 subtype B phylogenetic tree, the cohort clearly constitutes a single cluster, supported by high bootstrap values as posterior probabilities. Interestingly, the donor A lineage appears to be the out group for the two recipients and it was noted that recipient C revealed one long-branch segregating earlier time points from samples obtained from 1997 till 2000 [15]. As this is in correlation to clinical patient profile, one can deduce that the emergence of host-induced viral variation and hence viral evolution at recent time points occurred in concert with the rapidly progressing status of AIDS patient C. This pattern was also evident through analyses obtained from all the individual genes (data not shown).
Overall, patient-derived virus sequences obtained from corresponding longitudinal samples showed tight clustering within patients, well supported by bootstrap values and posterior probabilities. To analyze within patient evolutionary patterns, a splitstree, allowing the representation of conflicting phylogenetic signal, was reconstructed for all the cohort sequences (Figure 2). In the splitstree the evolutionary patterns within each patient are blurred by discordant relationships indicated by the reticulate pattern of evolution. This pattern of phylogenetic discordance suggests the presence of recombination and/or adaptive evolution, which is acting as a major evolutionary force on the patient's viral variants over time in vivo. Recombination produces networks of sequences rather than strictly bifurcating evolutionary trees. Depicted by the Splitstree program, a tree topology typical of recombination or conflicting phylogenetic signals in the data contains parallel edges between sequences.
Figure 2 Split graph of the cohort reconstructed using the Kimura-2-parameter corrected distances. The splits were refined since this significantly improved the fit. Bootstrap values are indicated on the edges and were performed using the Neighbor-Joining method on 1000 replicates (previously published in Mikhail et al., 2005). Bayesian trees were reconstructed in mrBayes v2.01. Network analysis was performed in Splitstree v 1.0.1, 2.4; Huson 1998).
Recombination analysis
To further delineate the cause of net like pattern seen at the nodes of the splits tree and to determine whether recombination has shaped the evolution of viral sequences, the Informative Sites Tests (IST) together with the Homoplasy test was conducted to test whether the null hypothesis of pure clonal evolution can be significantly rejected [16,17]. In addition, we also attempted to quantify the contribution of recombination to the viral genetic diversity using the Informative Site Index and the Homoplasy Ratio (HR) (Table 1). For the complete genomes, both indices are in the same order of magnitude of 0.3 indicating the presence of recombination. However, for the major genes, the P values still indicate the hallmark of recombination, but the recombination indices become slightly varied and are no longer comparable between the two tests. If this recombination signal is also the cause of reticulate evolution within each patient, then recombination was equally evident in both the donor and recipients (Figure 2). Therefore, even though recombination appears to be an inherent property in this cluster, its exact biological association with progression and non-progression of HIV disease in this cohort is only partially clear, and the possible role of selection pressures on disease progression is needed to be investigated.
Table 1 Results of the Homoplasy Test and the Informative Sites Test
Homoplasy Test Informative Sites Test
P value HR P value ISI
complete genome P < 0.001 0.254 P < 0.001 0.34
gag P < 0.017 0.565 P < 0.098 0.38
pol P < 0.015 0.299 P < 0.007 0.41
env P < 0.043 0.152 P < 0.002 0.42
Selective pressure and evolutionary rate analysis
To investigate the selective pressure exerted on the virus in the cohort members, a non-synonymous/synonymous substitution rate ratio scan was performed on the complete genomes using a maximum likelihood estimation procedure (Figure 3). The average dN/dS ratio shows considerable variation across the genome, with the highest ratios in the env gene, intermediate values in the accessory genes and lower values in the pol gene, with fairly low values for the gag gene. A similar analysis using complete genomes, representative for the HIV-1 diversity group M found from the Los Alamos HIV Database, also resulted in a similar plot, confirming previous reported results [9,17,18]. With the methods at hand, we can quantify the selective pressure across the genome for the complete cohort but it is not possible to document differences in selective pressure between cohort members due to parameter constraints of the mathematical models used. Thus, although over time analyses do demonstrate that differential selective pressure is clearly present in this cohort, its clear relationship with disease progression cannot be unraveled due to the possible contributing role of recombination. And since selection can result in heterogeneous rates along sequences, conflicting phylogenetic signal in this cohort might also have arisen from selection in addition to recombination. This is further confirmed by the correlation of the log likelihood estimates of the overall phylogenetic hypothesis plotted against the dN/dS ratios obtained by the scanning window approach (data not shown).
Figure 3 Non-synonymous : synonymous base rate ratio across the complete genome as estimated under a codon substitution model (MO) in a sliding window fashion with a step size of 81 bp and a window size of 801 bp, indicating the highest ratios within the env gene, followed by the pol, gag and nef genes, respectively.
To investigate differences in evolutionary rate between patients, molecular clock analysis was performed. Figure 4 shows the root-to-tip divergence in function of the sampling time. Linear regression estimates for the evolutionary rates were 2.38 × 10-3 (7.33 × 10-4-3.87 × 10-3), 7.75 × 10-3 (1.86 × l0-3-8.38 × 10-3) and 3.77 × 10-3 (3.07 × 10-3-4.44 × 10-3) nucleotide substitutions/site/year for patient A, B and C, respectively (Figure 4). By incorporating a global molecular clock, constraining all branches with one single evolutionary rate, and local molecular clocks, accommodating for different rates among different branch sets, evolutionary rates were obtained by maximum likelihood under the tip-dated model. Table 2 shows that allowing for different rates among the patients provided a significantly better fit (P < 0.001) than the global clock model, illustrating that the evolutionary rates were significantly different for the three cohort members. It should be noted however that the non-clock model, allowing for a different rate for each branch in the phylogeny, still remained significantly better as determined by the likelihood ratio test. Estimates of the evolutionary rate show a slow evolution for patient A and much higher rates in the two progressors (B and C), with the highest virus evolutionary rate in recipient B in agreement with the linear regression analysis and also consistent with his recent death with AIDS. Thus, from these analyses we have strong evidence showing a considerable influence of viral evolutionary rate on HIV disease progression.
Figure 4 Linear regression plot for root to tip divergence versus sampling date within each patient of the cohort. All regressions had an R2 value above 0.92. This graph indicates the highest slope and thus evolutionary rate for recipient B, followed by recipient C and lowest evolutionary rate for non-progressing donor A.
Table 2 Parameter estimates and log likelihoods under different clock models
Model p Log L Evolutionary rate
Different Rates 34 -24119 n.a.
Global clock 21 -24218 ABC: 2.928 × l0-3 (± 0.72 × l0-3)
Local clock for A and (BC) 22 -24164 A: 1.308 × l0-3 (± 0.19 × 10-3), BC: 5.08810-3 (± 0.41 × 10-3)
Local clock for A, B and C 23 -24156 A: 1.008 × l0-3 (± 0.16 × 10-3), B: 1.2 × l0-2 (± 1.86 × 10-3), C: 4.8 × l0-3 (± 0.38 × 10-3)
p The amount of parameters used in the model.
LogLThe log likelihoods.
Discussion
In this study we have carried-out detailed analyses of molecular factors that might contribute to HIV disease progression in an epidemiologically-linked cohort in which a HIV-infected non-progressor transmitted virus to recipients who gradually progressed to AIDS. With the help of 15 full-length HIV-1 genomes derived from the cohort members, where time and source of infection were known, we are able to show how various genetic changes following transmission of HIV from a non-progressor (donor A) accompanied disease progression in two recipients (B and C). Previously, Sydney Blood Bank Cohort (SBBC) also identified a similar transmission of HIV-1 from a non-progressor to 5 other recipients, but in this case patients did not progress as they were all infected with a nef-deleted HIV-1 strain [19]. We have investigated host-induced viral divergence, selection pressure, recombination and viral evolutionary rates of HIV-1 strains in this cohort.
It is apparent that following transmission of HIV-1 from the donor A, the 2 recipients B and C gradually deteriorated over a 15-year period to low CD4+/CD8+ T cell counts and high viral loads despite the continuation of HAART since 1997. These data suggest a possible role of in vivo viral divergence and host selection pressure over time, in the transition of a virus associated with non-progression in the donor, to a virus associated with gradual progression of HIV in the 2 recipients B and C of the cohort. To investigate this, the contribution of recombination to the genetic diversity and consequently disease progression evident in these cohort members was assessed using IST and the Homoplasy test. As our cohort is epidemiologically-linked, classical techniques such as Simplot, which uses a scanning window approach to detect conflicting topologies, are unreliable. Our methods capture conflicting phylogeny signal at the third codon positions and fourfold degenerate sites, which is unlikely to have resulted from selective pressure, thus indicating recombination. For the complete genomes, similar recombination indices were obtained using both tests. Some differences were observed when individual major genes were considered which could be attributed to different methodology and/or different parameters used by the two different algorithms.
Host-imposed immune selection was investigated by scanning dN/dS ratios across the genome. The variation found across the genome was consistent with that found for HIV-1 group M. Of particular interest was the fairly low ratios obtained for the gag gene which has been extensively implicated in CTL escape [3,20]. Further investigations of our analysis also indicates which genome regions have high dN/dS ratios. Though various reports have documented the evolutionary constraints placed by overlapping reading frames and secondary structures on RNA viruses such as HIV-1 [21,22], it is important to note that the exact number and location of the identified positively selected sites are not under investigation. Rather this study focuses on attributing the discordant phylogenetic patterns detected over time between cohort members by the possible contribution of positive selection. Differential selective pressure was found to have substantially contributed to virus evolution within these three cohort members.
Furthermore, it is noteworthy that while recombination in addition to selection forces may have contributed to the formation of the virus causing the gradual progression of HIV in the 2 recipients, it is possible that the HIV status of these individuals is associated with their HLA types, and not only due to the possible emergence of CTL escape mutations or other host factors as described previously [7,15,23].
In addition, by investigating the divergence of the serially sampled sequences using linear regression [24], we analyzed the rate of viral evolution. Although this analysis is suggestive of higher evolutionary rates in both progressors, the overlapping confidence intervals do not allow us to conclude significant differences. Earlier reports conducted by Ganeshan et al., and Essajee and colleagues based their HIV diversity studies on only partial segments of the env gene [25,26], conducting similar phylogenetic analysis but assessing viral heterogeneity either through heteroduplex assays or nucleotide based distance matrices, respectively. Despite both reports depending only on the env gene, which is naturally variable, both indicate that early quasispecies diversification may be associated with a favorable clinical outcome, with limited heterogeneity correlating to slower HIV disease, and a lack of vertical transmission from mother child pairs, respectively [25,26]. Taken together, literature suggests that an inverse relationship exists between viral diversity and disease progression [25,26], however other studies inclusive of ours also indicate the contrary [15,27]. Moreover, as our analysis relies on predetermined mathematical algorithms the assumption of data independence by linear regression estimates is violated as sequences share a phylogenetic history. Therefore, we estimated the evolutionary rates using a maximum likelihood framework that takes this into account and allows us to test different hypotheses using local clock models imposed onto the genealogy [28,29]. This molecular clock analysis, confirmed a higher rate of evolution in progressors B and C, as opposed to a lower rate in non-progressing donor A. The fact that HIV evolutionary rate could be patient-specific and influenced by immunologic control or even therapy-induced control [30], has major implications for evolutionary and vaccine studies. In our study it is difficult to assess the role of therapy-induced control of HIV-evolution as both patient B and C, who received therapy, had intermittent changes in drug regimen, which usually comprises of a cocktail of drugs and makes it impossible to dissect the role of each drug on the virus. Previous studies have indicated that combinations of RT drugs can act together to further increase HIV-1 mutation frequencies [30]. Thus, although we believe that therapy may have partially influenced viral evolution of HIV-1 strains in cohort patients, it is difficult to assess contribution of individual drugs in affecting viral evolutionary rates. Nonetheless, it is important to reiterate that it does not bias our overall interpretation of HIV disease progression as both recipients prior to initiation of therapy (pre 1997) were showing a gradual decline in T cell counts and rising plasma viremia.
Thus, the most unique aspect of our study the demonstration of patient-specific evolutionary rates as a major contributor to the general lack of a molecular clock in HIV. To date no molecular clock model accommodates for recombination and one can dispute the relevance of the evolutionary rates obtained. However, the genealogy-based estimates are in good agreement with the linear regression estimates, which were based on the viral divergence for each patient separately. Simulations have shown that recombination, even in small amounts, can disturb the molecular clock [31,32], and hence why the more general non-clock model provides a better fit to this data.
Overall, our studies raise the possibility that non-progressors, in some cases may harbor both pathogenic and non-pathogenic variants. Host genetics may act as driving force for positive selection of infecting strains [33]. Although viral recombination and differential selective pressure were found to have significantly affected virus variability in all 3 cohort members, there was striking correlation between faster viral evolutionary rate with accelerated disease progression.
Materials and methods
Cohort patient profiles
By using the well-described approaches of both Lookback and Traceback, clusters of distant HIV transmissions can be identified [34]. One such cluster was identified with the donor A, who likely acquired infection in 1982 and infected 2 recipients B (in 1983 autumn) and C (in 1983 summer) through blood transfusion. These infections were confirmed serologically in late 1990. The donor has remained well for over twenty years without requiring antiretroviral therapy and has maintained CD4+ T cell count above 550 cells/mm3 and CD8+ T cell count over 600 cells/mm3 and a viral load consistently less than 10000 copies /ml. In contrast, both recipients (B and C) have required the use of highly active antiretroviral therapy (HAART) which was initiated in 1995 and 1997 respectively (consisting of ddl/3TC/IMD) with recipient B still alive. On the other hand recipient C experienced a dramatic decline in CD4+ T cell count in 1997 down to CD4+ T cell count of 7 cells / mm3 (Figure 1A, IB and 1C) and has recently died of AIDS-related illness after 14 years post-infection. HLA typing was also conducted revealing patient A to be type A2, A3, B57, B65 and unknown for locus C, patient B showed to be HLA A2, A11, B56, B62 and CW1, while patient C was similariy found to be HLA A2, A24, B7, B13 and unknown for locus C. For a detailed description of patient clinical profiles, patient HLA types and phylogenetic evidence confirming epidemiological linkage refer to Mikhail et al., 2005.
Figure 1 Cohort patient profiles showing CD4+ and CD8+ T cell counts and plasma viral loads for patients A, B and C, respectively.
Full Length genome amplification of HIV-1 strains
Gene-Amp XL PCR kit (Perkin – Elmer Emerville Ca, USA) together with nested internal PCR reactions were used to amplify near full-length HIV genomes (8766 base pairs, the LTR domains were amplified separately) as previously published [5,15]. Population sequencing was conducted on a total of four longitudinal cohort samples obtained from donor A, termed Al, A3, A5, and A6 and corresponded to years 1992, 1997, 1998 and 2000. Similarity 4 time points from patient B were termed B3, B4, B5 and B6 correspond to years: 1992, 1997, 1998 and 2000 for sample collection, with C2, C3, C5, C6, C8, C10 and C11 representing patient C samples obtained from 1993, 1994, 1996, 1993, 1997, 1998 and 2000. To investigate the presence of patient mutations within a known CTL epitope, a database search was conducted within the Los Alamos (NM, USA) immunology database [18]. HIV-1 near full length sequences derived from cohort patients were consequently used to confirm epidemiological linkage and investigate molecular gene by gene comparisons as previously published [15].
Sequencing and phylogenetic analysis of cohort patients
Population nucleotide sequences and peptide sequences were aligned using CLUSTAL W [35] and manually edited in Se-AI according to their reading frame. The best-fitting nucleotide-substitution model was selected using Modeltestv3.06 [36], Phylogenetic trees were reconstructed in PAUP4.0bl0, starting from a Neighbor-Joining tree under a heuristic maximum likelihood search that implemented both nearest-neighbor interchange (NNI) and subtree pruning-regrafting (SPR). Bootstrap analysis was performed using the Neighbor-Joining method on 1000 replicates (previously published in Mikhail et al., 2005). Bayesian trees were reconstructed in mrBayes v2.01. Network analysis was performed in Splitstree 2.4.
Recombination analysis
Since the detection of specific recombination patterns and breakpoints in closely related sequences might be unreliable, evidence for recombination was investigated on a non-overlapping DNA concatemer or in single gene regions using two different tests: (a) the Informative Sites Test (IST) as implemented in PIST on the third codon positions [16], and (b) the Homoplasy Test on the fourfold degenerate sites [16]. The Homoplasy Test determines if there is a statistically significant excess of homoplasies in the phylogenetic tree derived from the data set, compared to an estimate of the number of homoplasies expected by repeated mutation in the absence of recombination [37]. An index of greater than zero indicates linkage equilibrium or recombination, but a value of zero or less indicates pure clonal evolution [34], The IST test detects whether the proportion of two-state parsimony-informative sites to all polymorphic sites is greater than expected from clonally generated data [16].
Selective pressure
Non-synonymous to synonymous substitution rate ratio's (dN/dS) were estimated in a sliding-window fashion under a probabilistic model of codon substitution that restricts all sites to a single dN/dS (M0) index across the complete genome [28]. All calculations were performed using the codeml program from the PAML package.
Evolutionary rate analysis
Root-to-tip divergences were calculated in VirusRates v.0, provided by Andrew Rambaut [24]. Confidence intervals for the linear regression estimates were obtained by bootstrapping the original alignment. Maximum likelihood analysis and local clock modeling was performed in PAML v 3.13 b, provided by Ziheng Yang, which implements a tip-date model estimated as additional parameters under the constraint that the positions of the tips are proportional to the sampling date [28].
Genbank accession numbers
Near full length HIV-1 genomes derived from cohort patient's PBMCs have been allocated Genebank accession numbers AY779550-AY779564.
List of abbreviations used
HIV-l human immunodeficiency virus type 1
AIDS acquired immunodeficiency syndrome
PBMC peripheral blood mononuclear cells
IST Informative site test
HR homoplasy ratio
SBBC Sydney blood bank cohort
CTL cytotoxic T lymphocyte
HLA human leukocyte antigen
NNI nearest neighbor interchange
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
M.M was assisted by B.W in carrying out the molecular genetic studies, generating sequence alignments, and drafting the paper. P.L conducted the evolutionary and recombination studies, B.B together with M.J.G provided the clinical samples, under analysis, while A-M.V participated in the design of the evolutionary study and its analysis. N.K.S conceived of the study, participated in its supervision, design, complete coordination and conclusion. All authors read and approved the final manuscript.
Acknowledgements
Authors would like to thank all members of the cohort for their participation. M.M was supported by the Australian Postgraduate Award (APA) from the University of Sydney and a top-up grant from the Millennium Foundation. P.L. was supported by the Flemish Institute for Scientific-technological Research in Industry (IWT).
==== Refs
Michael ML Chang G d'Arcy LA Tseng CJ Birx DL Sheppard HW Functional characterization of human immunodeficiency virus type 1 nef genes in patients with divergent rates of disease progression J Virol 1995 69 6758 6769 7474087
Trachtenberg E Korber B Sollars C Kepler TB Hraber PL Hayes E Funkhouser R Fugate M Theiler J Hsu YS Kunstman K Wu S Phair J Erlich H Wolinsky S Advantage of rare HLA supertype in HIV disease progression Nat Med 2003 9 928 935 12819779 10.1038/nm893
Wang B Dyer WB Zaunders JJ Mikhail M Sullivan JS Williams L Haddad DN Harris G Holt JA Cooper DA Miranda-Saksena M Boadle R Kelleher AD Saksena NK Comprehensive Analyses of a Unique HIV-1 -Infected Non-progressor Reveal a Complex Association of Immunobiological Mechanisms in the Context of Replication-Incompetent Infection Virology 2000 304 246 264 12504566 10.1006/viro.2002.1706
Harrer T Harrer E Kalams SA Cytotoxic T lymphocytes in asymptomatic longterm non-progressing HIV-1 infection. Breadth and specificity of the response and relation to in vivo viral quasispecies in a person with prolonged infection and low viral load J Immunol 1996 156 2616 2623 8786327
Wang B Mikhail M Dyer WB Zaunders JJ Kelleher AD Saksena NK First demonstration of lack of viral sequence evolution in a non-progressor, defining replication-incompetent HIV-infection Virology 2003 312 315 350
Wilson CC Brown RC Korber BT Wilkes BM Ruhl DJ Sakamoto D Kunstman K Luzuriaga K Hanson 1C Widmayer SM Wiznia A Clapp S Ammann AJ Koup RA Wolinsky SM Walker BD Frequent detection of escape from cytotoxic T-lymphocyte recognition in perinatal human immunodeficiency virus (HIV) type 1 transmission: the ariel project for the prevention of transmission of HIV from mother to infant J Virol 1999 73 3975 3985 10196293
Migueles SA Sabbaghian MS Shupert WL Bettinotti MP Marincola FM Martino L Hallahan CW Selig SM Schwartz D Sullivan J Connors M HLA B*5701 is highly associated with restriction of virus replication in a subgroup of HIV-infected long term nonprogressors Proc Nafl Acad Sci U S A 2000 97 2709 2714 10.1073/pnas.050567397
Kaslow RA Carrington M Apple R Park L Munoz A Saah AJ Goedert JJ Winkler C O'Brien SJ Rinaldo C Detels R Blattner W Phair J Erlich H Mann DL Influence of combinations of human major histocompatibility complex genes on the course of HIV-1 infection Nat Med 1996 2 405 411 8597949 10.1038/nm0496-405
Yusim K Kesmir C Gaschen B Addo MM Altfeld M Brunak S Chigaev A Detours V Korber BT Clustering patterns of cytotoxic T-lymphocyte epitopes in human immunodeficiency virus type 1 (HIV-1) proteins reveal imprints of immune evasion on HIV-1 global variation J Virol 2002 76 8757 8768 12163596 10.1128/JVI.76.17.8757-8768.2002
Rosenberg ES Billingsley JM Caliendo AM Boswell SL Sax PE Kalams SA Walker BD Vigorous HIV-1 -specific CD4+ T cell responses associated with control of viremia Science 1997 278 1447 1450 9367954 10.1126/science.278.5342.1447
Fang G Burger H Chappey C Rowland-Jones S Visosky A Chen CH Moran T Townsend L Murray M Weiser B Analysis of transition from long-term nonprogressive to progressive infection identifies sequences that may attenuate HIV type 1 AIDS Res Hum Retroviruses 2001 17 1395 1404 11679152 10.1089/088922201753197060
Saksena NK Wang B Dwyer WB Biological and Molecular Mechanisms in Progression and non-Progression of HIV Disease AIDS Rev 2001 3 3 10
Saksena NK Ge YC Wang B Xiang SH Ziegler J Palasanthiran P Bolton W Cunningham AL RNA and DMA sequence analysis of the nef gene of HIV type 1 strains from the first HIV type 1 -infected long-term nonprogressing mother-child pair AIDS Res Hum Retroviruses 1997 13 729 732 9168243
Wang B Ge YC Palasanthiran P Xiang SH Ziegler J Dwyer DE Randle C Dowton D Cunningham A Saksena NK Gene defects clustered at the C-terminus of the vpr gene of HIV-1 in long-term nonprogressing mother and child pair: in vivo evolution of vpr quasispecies in blood and plasma Virology 1996 223 224 232 8806556 10.1006/viro.1996.0471
Mikhail M Wang B Lemey P Beckholdt B Vandamme AM Gill JM Saksena NK Full-Length HIV-1 Genome Analysis Showing Evidence For HIV-1 Transmission From A Non-Progressor To Two Recipients Who Progressed To AIDS AIDS Res Hum Retrov 2005
Posada D Crandall KA Evaluation of methods for detecting recombination from DNA sequences: Computer simulations PNAS 2001 98 13757 13762 11717435 10.1073/pnas.241370698
Maynard Smith J Smith NH O'Rouke Spratt BG How Clonal Are Bacteria? Proc Natl Acad Sci 1993 90 4384 4388 8506277
Editors Korber TMB Brander C Haynes BF Koup R Kuiken C Moore JP Walker DB Watkins ID Theoretical Biology and Biophysics, Los Alamos, HIV Molecular Immunology 2. Publisher: Los Alamos National Laboratory 2000 Los Alamos, New Mexico UR03 5816
Birch MR Learmont JC Dyer WB Deacon NJ Zaunders JJ Saksena N Cunningham AL Mills J Sullivan JS An examination of signs of disease progression in survivors of the Sydney Blood Bank Cohort (SBBC) J Clin Virol 2001 22 263 270 11564591 10.1016/S1386-6532(01)00198-6
Yang Z Yoder AD Estimation of the transition/transversion rate bias and species sampling J Mol Evol 1999 48 274 283 10093216
Holmes EC Error thresholds and the constraints to RNA virus evolution Trends Microbiol 2003 11 543 546 14659685 10.1016/j.tim.2003.10.006
Simmonds P Smith DB Structural constraints on RNA virus evolution J Virol 1999 73 5787 5794 10364330
Goulder PJ Brander C Annamalai K Mngqundaniso N Govender U Tang Y He S Hartman KE O'Callaghan CA Ogg GS Altfeld MA Rosenberg ES Cao H Kalams SA Hammond M Bunce M Pelton SI Burchett SA Mclntosh K Coovadia HM Walker BD Differential narrow focusing of immunodominant human immunodeficiency virus gag-specific cytotoxic T-lymphocyte responses in infected African and caucasoid adults and children J Virol 2000 74 5679 5690 10823876 10.1128/JVI.74.12.5679-5690.2000
Rambaut A Estimating the rate of molecular evolution: incorporating non-contemporaneous sequences into maximum likelihood phylogenies Bioinformotics 2000 16 395 399 10.1093/bioinformatics/16.4.395
Ganeshan S Dickover RE Korber BT Bryson YJ Wolinsky SM Human immunodeficiency virus type 1 genetic evolution in children with different rates of development of disease J Virol 1997 71 663 677 8985398
Essajee SM Pollack H Rochford G Oransky I Krasinski K Borkowsky W Early changes in quasispecies repertoire in HIV-infected infants: correlation with disease progression AIDS Res Hum Retroviruses 2000 16 1949 957 11153077 10.1089/088922200750054675
Matala E Crandall KA Baker RC Ahmad N Limited heterogeneity of HIV type 1 in infected mothers correlates with lack of vertical transmission AIDS Res Hum Retroviruses 2000 16 1481 1489 11054261 10.1089/088922200750006001
Yang Z Bielawski JP Statistical methods for detecting molecular adaptation Trends Ecol Evol 2000 15 496 503 11114436 10.1016/S0169-5347(00)01994-7
Drummond A Rodrigo AG Reconstructing genealogies of serial samples under the assumption of a molecular clock using serial-sample UPGMA Mol Biol Evol 2000 17 1807 1815 11110896
Mansky LM HIV mutagenesis and the evolution of antiretroviral drug resistance Drug Resist Updaf 2000 5 219 223 Review 10.1016/S1368-7646(02)00118-8
Schierup MH Hein J Recombination and the molecular clock Mol Biol Evol 2001 17 1578 1579 11018163
Maynard Smith J Smith NH Detecting recombination from gene trees Mol Biol Evol 1998 15 590 599 9580989
Deacon NJ Tsykin A Solomon A Smith K Ludford-Menting M Hooker DJ McPhee DA Greenway AL Ellett A Chatfield C Genomic structure of an attenuated quasi species of HIV-1 from a blood transfusion donor and recipients Science 1995 270 988 991 7481804
Gill MJ Towns D Allaire S Meyers G Transmission of human immunodeficiency virus through blood transfusion: the use of look-back and trace-back approaches to optimize recipient identification in a regional population Transfusion 1997 37 513 516 9149777 10.1046/j.1537-2995.1997.37597293883.x
Thompson JD Higgins DG Gibson TJ CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice Nucleic Acids Res 1994 22 4673 4680 7984417
Posada D Crandall KA MODELTEST: testing the model of DMA substitution Bioinformatics 1998 14 817 818 9918953 10.1093/bioinformatics/14.9.817
Worobey M A novel approach to detecting and measuring recombination: new insights into evolution in viruses, bacteria, and mitochondria Mol Biol Evol 2001 18 1425 1434 11470833
|
15985187
|
PMC1190217
|
CC BY
|
2021-01-04 16:36:39
|
no
|
Retrovirology. 2005 Jun 29; 2:41
|
utf-8
|
Retrovirology
| 2,005 |
10.1186/1742-4690-2-41
|
oa_comm
|
==== Front
RetrovirologyRetrovirology1742-4690BioMed Central London 1742-4690-2-481608078910.1186/1742-4690-2-48ResearchInvolvement of a small GTP binding protein in HIV-1 release Audoly Gilles [email protected] Michel R [email protected] Pablo [email protected] Unité des Rickettsies, CNRS UMR6020, Faculté de Médecine, 27 bd Jean Moulin, 13385 Marseille cedex 05, IFR48, France2 Unité des Bactéries Anaérobies et Toxines, Institut Pasteur, 28 rue du Dr. Roux, 75724 Paris Cedex 15, France2005 4 8 2005 2 48 48 3 3 2005 4 8 2005 Copyright © 2005 Audoly et al; licensee BioMed Central Ltd.2005Audoly et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
There is evidence suggesting that actin binding to HIV-1 encoded proteins, or even actin dynamics themselves, might play a key role in virus budding and/or release from the infected cell. A crucial step in the reorganisation of the actin cytoskeleton is the engagement of various different GTP binding proteins. We have thus studied the involvement of GTP-binding proteins in the final steps of the HIV-1 viral replication cycle.
Results
Our results demonstrate that virus production is abolished when cellular GTP binding proteins involved in actin polymerisation are inhibited with specific toxins.
Conclusion
We propose a new HIV budding working model whereby Gag interactions with pre-existing endosomal cellular tracks as well as with a yet non identified element of the actin polymerisation pathway are required in order to allow HIV-1 to be released from the infected cell.
==== Body
Background
The final step in HIV-1 replication cycle is the release of nascent viral particles from the infected cell. In this way, HIV-1 acquires its lipid bilayer envelope by budding through the plasma membrane of infected T CD4+ cells. The only necessary and sufficient viral element for this event to take place is the expression product of the gag gene; i.e. the Pr55gag precursor [1]. Cells only expressing Pr55gag are able to produce and release vesicles, called viral-like particles (VLP), of size and morphology resembling those of immature viral particles [2,3]. A discrete functional sequence, referred to as the L domain encoded by a PTAP motif in the C-terminal, p6 portion of the Gag precursor, catalyses the pinching off of virus particles from the plasma membrane. Indeed, as demonstrated by EM, virus harbouring a modified L domain have been observed to remain attached to the cell via a thin tether [4]. Further work has shown that the interaction between this viral domain and the cellular cytosolic Tsg101 (the tumor susceptibility gene) molecule, that functions in the biogenesis of the multivesicular body (MVB) endosomal compartment [5], is critical for nascent virus detachment from the plasma membrane of the infected T cell [reviewed in 6].
The biological mechanism involved in the production of either a vesicle or an enclosed membrane surrounded virion through membrane budding, implies plasma membrane curvation prior to phospholipid bilayer fusion. Plasma membrane dynamics are partially governed by actin nucleation, a phenomenon in which several cytosolic molecules, such as small GTP binding proteins among others, are involved [7]. Interestingly, GTP binding protein-dependent actin nucleation, is also a key molecular mechanism in endosomal related vesicular transport [reviewed in 8].
Previous studies reported that HIV-1 release from infected cells could be blocked by disturbing the actin network with specific toxins as Cytochalasin D (Cyto D) or Mycalolyde B [9,10]. The published data shows that, although structural viral proteins are transported and localized to the inner face of the plasma membrane in Cyto D treated cells, HIV-1 virions remain attached to the cell, presenting the same phenotype as observed for L-domain mutated viruses [9].
Since actin dynamics are involved in intracellular vesicular transport, and multiple actin nucleation events at the cell cortex lead to the formation of a dense branched filament network that pushes the membrane forward [11], we postulated that the actin polymerisation pathway itself may play a crucial role in efficient HIV-1 release.
Results
Inhibition of small GTP-binding proteins abolishes HIV budding
We have tested the involvement of plasma membrane related small GTP binding proteins in virus release, using specific bacterial toxins. Toxin B from Clostridium difficile inhibits Cdc42, Rho and Rac molecules by modifying the protein structure through threonine glucosylation [12]. This modification blocks their ability to bind downstream effectors, resulting in actin network disruption. We first asked whether or not Toxin B treatment would interfere with Gag budding and release in a system, where high levels of HIV-1 Pr55gag, as the only viral protein, would be produced. Expression of the HIV-1 Gag precursor, by HeLa-CD4 cells, resulted in VLPs released to the media (Fig. 1A and Material and Methods). The fact that VLP related Pr55gag was neither degraded by Trypsin treatment nor disassembled by Triton X-100 detergent addition, strongly suggested that the viral protein might be surrounded by a lipid bilayer (Fig. 1A). Total degradation of Pr55gag was only obtained after Trypsin treatment of detergent solubilized material (Fig. 1A). Incubation of HIV-1 Gag expressing HeLa-CD4 cells with increasing amounts of Toxin B did not induce cell death, since more than 95% of treated cells excluded the Trypan blue dye. Interestingly, VLP release was inhibited in a dose dependent manner with a maximum effect at a Toxin B concentration of 4 ng/ml (Fig. 1A). Conversely, the overall intracellular Gag production was not significantly modified in these experimental conditions, as shown by p24 quantification and western blot analysis of the soluble fraction of detergent lysed treated cells (Fig. 1B, C). These results show that Pr55gag release was abolished when small GTP binding proteins such as Cdc42, Rac, and/or Rho were inhibited.
Figure 1 Toxin B inhibits VLP release. A. Toxin B dose dependent inhibition of VLP production. Supernatants of MVA infected/HIV-1 Gag transfected HeLa-CD4 cells, treated or not with various concentrations of Toxin B for 16 h, were clarified by low speed centrifugation and treated or not (no) with Trypsin (ty), Triton X-100 (tx) or with Triton X-100 and Trypsin (tt). VLPs were recovered by centrifugation and subjected to Western Blot analysis. B. Lysates of cells from panel A were subjected to Western Blot analysis. (mock: cells transfected with the vector without any insert, T+: is a cellular extract of HIV-1NDK infected H9 cells). C. p24 antigen was quantified in lysates from panel A by ELISA.
In order to define if this is also the case in HIV-1 infected cells, we tested the inhibition of virus production from HIV-1NDK infected Jurkat cells in the presence of Toxin B, exoenzyme C3 from Clostridium botulinum, and Lethal Toxin 82 (LT) from Clostridium sordellii. Exoenzyme C3 ADP-ribosylates specifically Rho proteins, whereas LT glucosylates specific Thr residues from Ras, Rap, Rac and Ral proteins [12,13]. The human Jurkat T cell line was infected with HIV-1NDK and maintained 4 days in culture. After washing 3 times with PBS to ensure elimination of previously produced viral particles, cells were grown for another 20 h in complete medium (RPMI) in the presence or absence of increasing amounts of toxins. The highest toxin concentration used corresponds to the maximal sub-lethal toxin concentration, defined as the maximal toxin amount that did not kill the cell (observed by Trypan blue exclusion) in our experimental system. Under these conditions, toxin activity was confirmed by loss of diffused cortical actin as well as actin aggregate formation, monitored by immunofluorescence microscopy on Phalloidin-FITC treated cells. Cellular morphological changes characterized by cell rounding and loss of numerous filopodial projections was also observed (Fig. 2). Gag and actin co-localized both in treated and untreated cells (Fig. 2b–g). Whereas both proteins were exclusively seen in membrane protrusions in infected untreated cells (Fig 2b), cortical actin disorganisation induced changes in HIV-1 Gag distribution in toxin treated cells (Fig. 2c–g).
Figure 2 Actin polymerisation and intracellular Gag distribution under toxin treatments. HIV-1 or mock infected Jurkat cells treated or not with 0.5 μg/ml toxins for 16 hours, were stained with Phalloidin-FITC (green) and p24 (red), in order to visualize actin organization and Gag distribution, respectively. A field of about 100 cells was studied for each condition, and the percentage of cells presenting disrupted (grey bar) or not (white bar) cortical actin pattern is represented as an histogram. "n": number of counted cells in the field. a) mock infected cells, b-g) HIV-1NDK infected cells. Untreated cells (a-b), cells treated with toxin B (c), LT (d), C3 (e), Cyto D (f), and Iota (g). Bar scale = 10 μm.
We further analysed the viral production capacity of HIV infected T cells treated with the bacterial toxins. Cell culture supernatants of toxin treated or untreated cells were harvested, and intracellular as well as extra cellular p24 antigen was quantified. The intracellular amount of p24 antigen was found to be identical for all cells; i.e. 27.1+/- 1.9 ng p24/105 cells. The release of p24 was unaffected by C3 and LT but was drastically inhibited by Toxin B (Fig. 3A). These data strongly suggest that indeed active small GTP binding proteins are necessary for HIV-1 to be released from the infected target cell.
Figure 3 Engagement of small GTP binding proteins in HIV-1 release. Jurkat HIV-1 infected cells were incubated for 20 h with various concentrations of bacterial toxins and Cyto D. A) Clarified supernatants of the culture medium were harvested for p24 quantification by ELISA. Vertical axis indicates the relative HIV-1 production expressed as a percentage of the p24 antigen in the absence of toxin treatment. B) Titres of infectious virus (TCID50) released/pg of p24 from the highest toxin concentration dose from infected cells shown in panel A. Data presented corresponds to one out of three independent experiments. Each experiment was performed in triplicate.
The increased amount of Toxin B required to inhibit VLP formation in HeLa cells compared to that required to abolish virus release in Jurkat cells (Fig. 1A and 3A) is due to the susceptibility of each cell line to the action of the toxin.
Unexpectedly, when infected Jurkat cells were incubated in the presence of two different actin disrupting agents, Cyto D or Iota Toxin, only Cyto D inhibited HIV production (Fig. 3A), as already reported [9], whereas Iota toxin did not. (fig 3A). Overnight incubation of HIV-1 infected Jurkat cells with various concentrations of these toxins did not induce cell death (as defined by Trypan blue exclusion) and resulted in toxin-dependent actin depolymerisation, as observed by immunofluorescence microscopy on Phalloidin-FITC treated cells (Fig. 2f, g). Since Cyto D reacts with elongating membrane interacting actin [14], whereas Iota sequesters soluble actin monomers [12], our result suggests that active nucleation at the plasma membrane may be necessary for HIV production.
Inhibition of small GTP-binding proteins reduces infectivity of HIV-1 particles
We further investigated whether toxin treatments of HIV-1 producing cells had any effect on the infectivity of the de novo synthesized virions. Infectivity released into the culture media at the highest toxin concentration used in the experiment represented in figure 3A, was quantified by measuring the TCID50/p24 value of supernatants, as described elsewhere [15] (Fig. 3B). Whereas Toxin B lowered the TCID50/p24 value of supernatants with a Toxin treated/Toxin untreated TCID50/p24 ratio of about 0.1, Cyto D only affected virus infectivity by a factor of 1.3 (Fig. 3B). This suggests that the infectivity of the small amount of released virus from Cyto D treated cells remained almost unchanged. Unexpectedly although C3, Iota and LT did not alter p24 release from infected cells (Fig. 3A), they reduced by about two-fold (Toxin treated/Toxin untreated TCID50/p24 ratio ranging from 0.40 to 0.55) the infectivity of cell-free virus (Fig. 3B). This suggests that the status of the actin network in virus-producing cells is relevant for the quality of the virus released into the medium.
It is well documented that Gag assembly in the cytoplasm of infected T cells is required as a key step prior to virus budding [16]. Thus, the inhibition of virus production by the action of toxins (Fig 3) could occur at the Gag assembly level rather than at the level of an interaction between plasma membrane actin polymerisation and the viral protein. In order to rule out this possibility, we studied the assembly status of soluble cytoplasmic Pr55gag in toxin treated cells by sucrose gradient analysis as already reported [17]. HIV-1 infected Jurkat cells treated or not with toxins were lysed in non denaturing conditions and the resulting soluble fraction was loaded on a discontinuous sucrose gradient (see Material and Methods section). In all cases, Pr55gag was recovered in fractions 9–11, at a relative density of about 1.15–1.20 g/ml (Fig. 4), corresponding to assembled non-enveloped Gag structures [18]. Thus, the observed toxin dependent inhibition of virus production was indeed at the level of virus release, and not a result of a modification of intracellular events leading to Pr55gag assembling.
Figure 4 Toxin treatment does not affect intracellular HIV-1 Gag assembly. Non denaturing cytoplasmic lysates of HIV-1NDK Jurkat infected cells treated or not with 0.5 μg/ml of Toxin B, LT, or Cyto D for 16 h, were centrifuged through a discontinuous sucrose gradient. Eleven fractions were collected from top to bottom, concentrated by high speed centrifugation, and analysed by Western Blot. Vertical axis shows the sucrose density fractionation in g/ml. Arrows indicate Pr55gag migration.
Discussion
In infected and transfected cells the HIV Gag precursor is known to be targeted to the inner face of the plasma membrane and to co-localise with actin. In our study we have shown in Jurkat T-cells that this co-localisation takes place in membrane protrusions (Fig. 2b), as previously shown for SupT1 HIV-1 infected cells [19]. Interestingly, incubation of HIV infected Jurkat cells with the toxins that induced cortical actin disorganisation, produced changes in HIV-1 Gag distribution (Fig. 2c–g). This result reinforces the previously reported physical interaction occurring between the Gag precursor and actin [20-23], and argues, as in Sasaki et al. [10], for a potential role for actin dynamics in Pr55gag intracellular localisation.
We have found that disturbing cortical actin dynamics inhibited virus production [Fig. 3A]. This was observed either by modifying the polymerising actin itself, by Cyto D action, or by inhibiting one key GTP binding protein involved in a molecular pathway that leads to actin nucleation, by Toxin B action. Some GTP binding proteins have been shown to govern actin dynamics as well as intracellular vesicular trafficking [8,24]. Since the viral Gag precursor does not travel through the secretory pathway [1] it is reasonable to hypothesize that HIV virus budding and actin polymerisation through activation of a GTP binding protein may be linked. What is thus the molecular mechanism that can explain this observation? We found that Toxin B abolished HIV-1 production whereas C3 and LT did not. Knowing the spectrum of the toxins targets [12,13], it can be inferred that Cdc42 might be a putative cellular partner to virus release. Cdc42 has been found to be specifically down-regulated in cells latently infected with HIV, suggesting an important role for active Cdc42 in virus infection [25]. It can thus be argued that active Cdc42 may induce an actin polymerisation pathway and allow virus budding and release. Analysis of virus production from HIV infected cells harbouring inactive forms of the Cdc42 molecule should help to ultimately define its involvement in this event.
Our study concluded that C3, Iota and LT reduced infectivity of virus produced. However these toxins did not alter total virus production (Figure 3A and 3B). This suggests that the capacity of the budding viral particle to infect a new target cell is modified through disruption of the actin web of the infected cell. How can actin be then correlated to infection in this particular case? The most possible explanation is based on the budding event itself. HIV selectively incorporates cellular membrane proteins, that have been suggested to be involved in virus infectivity [26], while budding from lipid raft domains at the plasma membrane of the infected cell [27] where the Gag precursor is mainly localised [28]. Since disruption of actin filaments modifies the protein content of lipid rafts [29], the action of the studied toxins on the infected cell might modify the cellular protein content of the lipid raft. HIV may then bud as a virus lacking a cellular component, or harbouring an inhibitory cellular molecule.
Virus entry, by a membrane fusion mechanism, requires actin nucleation [30] through activation of Rac-1 but not Cdc42 or Rho proteins [31]. According to our results actin network remodelling would be a key process for HIV replication, since it will play a crucial role in both early (entry) and late (budding) infectious events, by involvement of different sets of cellular GTP binding proteins.
Conclusion
We have shown that inhibiting small GTP binding proteins involved in cortical actin dynamics disrupts virus release. This is not the simple consequence of actin network disorganisation since the action of LT, C3 and Iota did not affect virus production. Our results suggest that the actin polymerisation process, potentially via Cdc42 is involved in the final step of the HIV replication cycle.
Analysis of recently published results shows that the implication of intracellular protein transport pathways to late endosomal compartments (i.e. the multivesicular bodies compartment) acts as pre-existing cellular "tracks" for the viral Gag protein-induced budding [32-34]. The data presented here argues for a more complex working model whereby in addition to using an intracellular "track", HIV requires the specific exploitation of actin dynamics in order to be released from the infected cell. Further experimental studies should be done to define the actin activation pathway used by Gag and the chronology of the molecular events involved.
Materials and methods
Cell culture and transfection
C8166 and Jurkat cells were grown in RPMI 1640 medium, and HeLa-CD4 cells in MEM medium. Both media were completed with 10% FCS, 2 mM glutamine and 100 U/ml of penicillin-streptomycin.
Cells were infected with the Ankara strain/T7 RNA polymerase (MVA) [35,36] at 1 pfu/cell, 30 min before being transfected by fugene-6 (Roche, Basel, Switzerland) with pos7 vector [36] or recombinant pos7-HIV-1Gag [37].
VLP analysis
Supernatants of cells were harvested and clarified by low speed centrifugation 24 h after transfection, and released VLP were concentrated by centrifugation at 100,000 × g at 4°C through a 20% sucrose cushion. The resulted pellet was resuspended in TNE and treated or not with 5 μg/ml trypsine and /or 1% Triton X-100. Treated or mock-treated VLPs were resolved on SDS 10% polyacrylamide gel and transferred onto nitrocellulose membrane. Immunoblotting was carried out with human polyclonal IgG purified from HIV-1 positive individuals (HIVIg), followed by peroxydase-conjugated anti-human antibodies incubation. HIV related proteins were detected using the ECL kit (Amersham Biosciences, Upsala, Sweden).
Cell lysis and density gradient
Cytoplasmic lysates of 5*105 cells were fractionated according to Gorvel et al. [38] with some modifications. Briefly, cells were washed in PBS and resuspended in 0.5 ml of cold homogenisation buffer (HB) (250 mM sucrose, 3 mM imidazole, 0.1% gelatin) completed with the protease inhibitors cocktail (from Roche). Cell lyses was obtained through 2 cycles of freezing and thawing. The lysates were then clarified by centrifugation and the resultant post nuclear supernatants (PNS), were diluted to 1 ml to obtain a final concentration of 32 % sucrose. A discontinuous sucrose gradient was set up, from bottom to top, as follows: 0.3 ml 62% sucrose, 0.3 ml 45% sucrose, 0.3 ml 35 % sucrose, 1 ml of diluted PNS, 0.6 ml 30% sucrose, 0.6 ml 25% sucrose, 0.6 ml 20 % sucrose, and centrifuged for 1 hr at 100,000 × g. Twelve fractions were collected from top to bottom. An aliquot of each fraction was used to determine the density by measuring the refraction index with a refractometer. Each fraction was diluted 1:3 in TNE buffer (10 mM Tris-HCl buffer pH 7, 0.1 M NaCl, 1 mM EDTA) and the assembled Gag protein was recovered as a pellet, after concentration by high speed centrifugation at 70,000 × g for 30 min. The pellets were resuspended in Laemmli loading buffer, and submitted to SDS 10% PAGE prior to Western Blot analysis.
Toxins
All toxins used in this study, but Cyto D, were purified as in [39-41]. Cyto D was purchased from Sigma (France).
Cell infection
Jurkat cells were infected with HIV-1NDK at an MOI of 0.5 and maintained 4 days in culture at 5 × 105 cell/ml. After 3 washes in PBS the cells were grown for another 20 h in complete medium containing serially diluted bacterial toxins. Quantification of viral production by HIV-1 p24 ELISA (Organon Teknika, Boxtel, NL) was done on supernatants, previously clarified by centrifugation at 1500 × g for 5 min. TCID50 was determined on C8166 T-lymphocytes as previously described [20].
Immunofluorescence studies
Cells were incubated on polylysine-covered slides at room temperature for 15 min and immediately fixed in phosphate-buffered saline (PBS) (pH 7.4) containing 3.7% para-formaldehyde and 0.025% glutaraldehyde for 10 min. Fixed cells were treated 10 min in 0.1 M glycine before being permeabilized in PBS containing 0.1% Saponine for 10 min. After two washes in PBS, cells were incubated with 1% bovine serum albumin in PBS (pH 7.4) for 20–30 min. Immunofluorescence staining was performed with phalloidin-FITC (Sigma Aldrich, France) and monoclonal anti p24 (Dako, France) followed by TRITC-labeled anti-mouse antibody (Jakson). The specimens were analysed on a fluorescence microscope. Separate images were taken in the corresponding channels, and merge images were composed. Image acquisition and data processing for all the samples were performed under the same conditions.
List of abbreviations
VLP : viral-like particles, tsg101 : the tumor susceptibility gene, MVB : the multivesicular body endosomal compartment, Cyto D : Cytochalasin D, LT : Lethal Toxin 82.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
GA performed the experiments. PG and GA participated in the experimental design, data interpretation and writing of the manuscript. MP was involved in the interpretation of toxin based experiments
Acknowledgements
This work was partly funded by Ensemble Contre le Sida. The HIVIg reagent was obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH, from NABI. We are deeply indebted to M. Suzan and D. Naniche for fruitful discussions.
==== Refs
Freed EO HIV-1 gag proteins: diverse functions in the virus life cycle Virology 1998 251 1 15 9813197 10.1006/viro.1998.9398
Gheysen D Jacobs E de Foresta F Thiariart C Francotte M Thines D De Wilde M Assembly and release of HIV-1 precursor Pr55gag virus-like particles from recombinant baculovirus-infected insect cells Cell 1989 59 103 112 2676191 10.1016/0092-8674(89)90873-8
Shioda T Shibuta H Production of human immunodeficiency virus (HIV)-like particles from cells infected with recombinant vaccinia viruses carrying the gag gene of HIV Virology 1990 175 139 48 1689917 10.1016/0042-6822(90)90194-V
Gottlinger HG Dorfman T Sodroski JG Haseltine WA Effect of mutations affecting the p6 gag protein on human immunodeficiency virus particle release Proc Natl Acad Sci USA 1991 88 3195 3199 2014240
Katzmann DJ Odorizzi G Emr SD Receptor downregulation and multivesicular-body sorting Nat Rev Mol Cell Biol 2002 3 893 905 12461556 10.1038/nrm973
Amara A Littman D After Hrs with HIV J Cell Biol 2003 162 371 375 12900390 10.1083/jcb.200307062
Yin HL Stull JT Proteins that regulate dynamic actin remodeling in response to membrane signaling minireview series J Biol Chem 1999 274 32529 32530 10551801 10.1074/jbc.274.46.32529
Stamnes M Regulating the actin cytoskeleton during vesicular transport Curr Opin Cell Biol 2002 14 428 433 12383793 10.1016/S0955-0674(02)00349-6
Sasaki H Nakamura M Ohno T Matsuda Y Yuda Y Nonomura Y Myosin-actin interaction plays an important role in human immunodeficiency virus type 1 release from host cells Proc Natl Acad Sci USA 1995 92 2026 2030 7892219
Sasaki H Ozaki H Karaki H Nonomura Y Actin filaments play an essential role for transport of nascent HIV-1 proteins in host cells Biochem Biophys Res Commun 2004 316 588 93 15020258 10.1016/j.bbrc.2004.02.088
Pollard TD Blanchoin L Mullins RD Molecular mechanisms controlling actin filament dynamics in nonmuscle cells Annu Rev Biophys Biomolec Struct 2000 29 545 576 10.1146/annurev.biophys.29.1.545
Barbieri JT Riese MJ Aktories K Bacterial toxins that modify the actin cytoskeleton Annu Rev Cell Dev Biol 2002 18 315 344 12142277 10.1146/annurev.cellbio.18.012502.134748
Hofmann F Rex G Aktories K Just I The ras-related protein Ral is monoglucosylated by Clostridium sordellii lethal toxin Biochem Biophys Res Commun 1996 227 77 81 8858106 10.1006/bbrc.1996.1470
Brown SS Spudich JA Mechanism of action of cytochalasin: evidence that it binds to actin filament ends J Cell Biol 1981 88 487 491 6894300 10.1083/jcb.88.3.487
Salahuddin SZ Markham PD Wong-Staal F Franchini G Kalyanaraman VS Gallo RG Restricted expression of human T-cell leukemia – lymphoma virus (HTLV) in transformed human umbilical cord blood lymphocytes Virology 1983 129 51 64 6412453 10.1016/0042-6822(83)90395-1
Bukrinskaya AG HIV-1 assembly and maturation. A brief review Arch Virol 2004 149 1067 82 15168195 10.1007/s00705-003-0281-8
Lee YM Yu XF Identification and characterization of virus assembly intermediate complexes in HIV-1-infected CD4+ T cells Virology 1998 243 78 93 9527917 10.1006/viro.1998.9064
Lingappa JR Hill RL Wong ML Hegde RS A multistep, ATP-dependent pathway for assembly of human immunodeficiency virus capsids in a cell-free system J Cell Biol 1997 136 567 581 9024688 10.1083/jcb.136.3.567
Sylwester A Wessels D Anderson SA Warren RQ Shutt DC Kennedy RC Soll DR HIV-induced syncytia of a T cell line form single giant pseudopods and are motile J Cell Sci 1993 106 941 953 8308076
Rey O Canon J Krogstad P HIV-1 Gag protein associates with F-actin present in microfilaments Virology 1996 220 530 534 8661406 10.1006/viro.1996.0343
Ott DE Coren LV Kane BP Busch LK Johnson DG Sowder RC Chertova EN Arthur LO Henderson LE Cytoskeletal proteins inside human immunodeficiency virus type 1 virions J Virol 1996 70 7734 7743 8892894
Liu B Dai R Tian CJ Dawson L Gorelick R Yu XF Interaction of the human immunodeficiency virus type 1 nucleocapsid with actin J Virol 1999 73 2901 2908 10074138
Ibarrondo FJ Choi R Geng YZ Canon J Rey O Baldwin GC Krogstad P HIV type 1 Gag and nucleocapsid proteins: cytoskeletal localization and effects on cell motility AIDS Res Hum Retroviruses 2001 17 1489 1500 11709093 10.1089/08892220152644197
Jahn R Südhof TC Membrane fusion and exocytosis Annu Rev Biochem 1999 68 863 911 10872468 10.1146/annurev.biochem.68.1.863
Krishnan V Zeichner SL Host Cell Gene Expression during Human Immunodeficiency Virus Type 1 Latency and Reactivation and Effects of Targeting Genes That Are Differentially Expressed in Viral Latency J Virol 2004 78 9458 9473 15308739 10.1128/JVI.78.17.9458-9473.2004
Gluschankof P Mondor I Gerdelblom HR Sattentau QJ Cell membrane vesicles are a major contaminant of gradient-enriched human immunodeficiency virus type-1 preparations Virology 1997 230 125 33 9126268 10.1006/viro.1997.8453
Nguyen DH Hildreth JE Evidence for budding of human immunodeficiency virus type 1 selectively from glycolipid-enriched membrane lipid rafts J Virol 2000 74 3264 72 10708443 10.1128/JVI.74.7.3264-3272.2000
Lindwasser OW Resh MD Multimerization of human immunodeficiency virus type 1 Gag promotes its localization to barges, raft-like membrane microdomains J Virol 2001 75 7913 24 11483736 10.1128/JVI.75.17.7913-7924.2001
Delaguillaumie A Harriague J Kohanna S Bismuth G Rubinstein E Seigneuret M Conjeaud H Tetraspanin CD82 controls the association of cholesterol-dependent microdomains with the actin cytoskeleton in T lymphocytes: relevance to co-stimulation J Cell Sci 2004 117 5269 82 15454569 10.1242/jcs.01380
Bukrinskaya A Brichacek B Mann A Stevenson M Establishment of a functional human immunodeficiency virus type 1 (HIV-1) reverse transcription complex involves the cytoskeleton J Exp Med 1998 188 2113 2125 9841925 10.1084/jem.188.11.2113
Pontow SE Heyden NV Wei S Ratner L Actin cytoskeletal reorganizations and coreceptor-mediated activation of rac during human immunodeficiency virus-induced cell fusion J Virol 2004 78 7138 47 15194790 10.1128/JVI.78.13.7138-7147.2004
Pornillos O Higginson DS Stray KM Fisher RD Garrus JE Payne M He GP Wang HE Morham SG Sundquist WI HIV Gag mimics the Tsg101-recruiting activity of the human Hrs protein J Cell Biol 2003 162 425 434 12900394 10.1083/jcb.200302138
Raiborg C Rusten TE Stenmark H Protein sorting into multivesicular endosomes Curr Opin Cell Biol 2003 15 446 455 12892785 10.1016/S0955-0674(03)00080-2
Von Schwedler UK Stuchell M Muller B Ward DM Chung HY Morita E Wang HE Davis T He GP Cimbora DM Scott A Krausslich HG Kaplan J Morham SG Sundquist WI The protein network of HIV budding Cell 2003 114 701 713 14505570 10.1016/S0092-8674(03)00714-1
Sutter G Ohlmann M Erfle V Non-replicating vaccinia vector efficiently expresses bacteriophage T7 RNA polymerase FEBS Lett 1995 371 9 12 7664891 10.1016/0014-5793(95)00843-X
Wyatt LS Moss B Rozenblatt S Replication-deficient vaccinia virus encoding bacteriophage T7 RNA polymerase for transient gene expression in mammalian cells Virology 1995 210 202 205 7793072 10.1006/viro.1995.1332
Bouyac M Courcoul M Bertoia G Baudat Y Gabuzda D Blanc D Chazal N Boulanger P Sire J Vigne R Spire B Human immunodeficiency virus type 1 Vif protein binds to the Pr55Gag precursor J Virol 1997 71 9358 9365 9371595
Gorvel JP Chavrier P Zerial M Gruenberg J rab5 controls early endosome fusion in vitro Cell 1991 64 915 25 1900457 10.1016/0092-8674(91)90316-Q
Popoff MR Purification and characterization of Clostridium sordellii lethal toxin and cross-reactivity with Clostridium difficile cytotoxin Infect Immun 1987 55 35 43 3793234
Popoff MR Hauser D Boquet P Eklund MW Gill DM Characterization of the C3 gene of Clostridium botulinum types C and D and its expression in Escherichia coli Infect Immun 1991 59 3673 3679 1910014
Gibert M Petit L Raffestin S Okabe A Popoff MR Clostridium perfringens iota-toxin requires activation of both binding and enzymatic components for cytopathic activity Infect Immun 2000 68 3848 3853 10858193 10.1128/IAI.68.7.3848-3853.2000
|
16080789
|
PMC1190218
|
CC BY
|
2021-01-04 16:36:39
|
no
|
Retrovirology. 2005 Aug 4; 2:48
|
utf-8
|
Retrovirology
| 2,005 |
10.1186/1742-4690-2-48
|
oa_comm
|
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-541601417210.1186/1743-422X-2-54MethodologyUse of a novel cell-based fusion reporter assay to explore the host range of human respiratory syncytial virus F protein Branigan Patrick J [email protected] Changbao [email protected] Nicole D [email protected] Lester L [email protected] Robert T [email protected] Vecchio Alfred M [email protected] Infectious Diseases Research, Centocor, Inc., 145 King of Prussia Road, Radnor, PA, 19087, USA2005 13 7 2005 2 54 54 9 6 2005 13 7 2005 Copyright © 2005 Branigan et al; licensee BioMed Central Ltd.2005Branigan et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Human respiratory syncytial virus (HRSV) is an important respiratory pathogen primarily affecting infants, young children, transplant recipients and the elderly. The F protein is the only virion envelope protein necessary and sufficient for virus replication and fusion of the viral envelope membrane with the target host cell. During natural infection, HRSV replication is limited to respiratory epithelial cells with disseminated infection rarely, if ever, occurring even in immunocompromised patients. However, in vitro infection of multiple human and non-human cell types other than those of pulmonary tract origin has been reported. To better define host cell surface molecules that mediate viral entry and dissect the factors controlling permissivity for HRSV, we explored the host range of HRSV F protein mediated fusion. Using a novel recombinant reporter gene based fusion assay, HRSV F protein was shown to mediate fusion with cells derived from a wide range of vertebrate species including human, feline, equine, canine, bat, rodent, avian, porcine and even amphibian (Xenopus). That finding was extended using a recombinant HRSV engineered to express green fluorescent protein (GFP), to confirm that viral mRNA expression is limited in several cell types. These findings suggest that HRSV F protein interacts with either highly conserved host cell surface molecules or can use multiple mechanisms to enter cells, and that the primary determinants of HRSV host range are at steps post-entry.
==== Body
Background
Human respiratory syncytial virus (HRSV) is the single most common cause of serious lower respiratory tract infections (LRTIs) in infants and young children causing up to 126,000 hospitalizations annually in the U.S. [1] with an estimated 500 deaths per year [2]. HRSV bronchiolitis has been associated with the development and exacerbation of wheezing and other respiratory conditions. Furthermore, HRSV is an increasingly recognized cause of pneumonia, exacerbation of chronic pulmonary and cardiac disease, and death in the elderly [3]. HRSV is also the most common viral respiratory infection of transplant recipients and is responsible for high rates of mortality in this group [4].
HRSV is member of the subfamily Pneumovirinae in the Paramyxoviridae family [5]. Two serologic subgroups (A and B) have been described and co-circulate throughout the world. Three viral transmembrane proteins (G, SH and F) are found on surface of the virus particle in the viral envelope. The G protein is a heavily glycosylated protein that mediates attachment of the virus to host cells, and although not strictly required for virus replication in culture, recombinant viruses lacking the G protein are attenuated in animals [6,7]. While its exact function is unknown, the SH gene is not essential for virus growth in tissue culture, and its deletion only results in slight attenuation in animals [6,8-12]. The F protein is a type 1 membrane protein essential for the packaging and formation of infectious virion particles [6,7,13,14], and is the only viral protein necessary and sufficient for fusion of the viral envelope membrane with the target host cell [15,16]. The HRSV F protein is highly conserved (89% amino acid identity) between subgroups A and B, and shares 81% amino acid identity with the F protein of bovine respiratory syncytial virus (BRSV). The HRSV F protein is synthesized as a 574 amino acid precursor protein designated F0, which is cleaved at two sites [17-20] within the lumen of the endoplasmic reticulum removing a short, glycosylated intervening sequence and generating two subunits designated F1 and F2 [21]. The mature form of the F protein present on the surface of the virus and infected cells is believed to consist of a homotrimer consisting of three non-covalently associated units of F1 disulfide linked to F2[22]. As with many other viral fusion proteins, F-mediated fusion with the host cell membrane is believed to be mediated by insertion of a hydrophobic fusion peptide into the host cytoplasmic membrane after binding of F protein to a target receptor on the host cell. Subsequent conformational changes within F result in the interaction of heptad repeat (HR) 1 with HR2 and formation of a 6-helix bundle structure [22-24]. This process brings the viral and host cell membranes in close proximity resulting in fusion pore formation, lipid mixing, and fusion of the two membranes. The precise number of F trimers and identity of target host surface proteins or molecules required to mediate fusion are currently unknown.
Although initially isolated from a chimpanzee, humans are the primary natural host for HRSV. HRSV will only infect the apical surface of human ciliated lung epithelial cells, and only fully differentiated human bronchial epithelial cells are permissive for HRSV growth [25]. Dissemination of HRSV to other organs is not observed even in immunocompromised individuals. Similarly, disseminated infection with bovine RSV is not observed in infected cattle. In contrast, in vitro infection of multiple human cell types other than those derived from lung [26], cells derived from other animal species, and HRSV infection of several animal species has been reported [27]. This suggests that the F protein interacts with either highly conserved host cell surface molecules or can use multiple mechanisms to mediate fusion. Several previous studies have shown the importance of cell-surface glycosaminoglycans (GAGs) [28-32], in particular iduronic acid, in mediating HRSV infection in vitro [33]; however, GAG independent, F-mediated attachment pathways have been described [13]. In a study comparing the host range of bovine and human respiratory syncytial viruses for cells derived from their respective natural hosts, species specificity mapped to the F2 subunit [10]. These finding allude to the existence of host specific receptor molecules that specifically interact with the F protein to mediate cell fusion.
To better understand the factors governing host range, we developed a HRSV F-based quantitative cell fusion assay and specifically examined the ability of HRSV F protein to mediate fusion with cells derived from a wide range of animal species. As cell permissiveness for virus growth is dependent upon multiple steps, we went on to further characterize the permissiveness of these cells for viral mRNA transcription by using a recombinant HRSV engineered to express GFP [33]. The relevance of these findings to the natural course of HRSV disease is discussed.
Methods
Cells and viruses
All cell lines were obtained from the American Type Culture Collection (ATCC) (Manassas, VA) and were grown at 37°C in a humidified atmosphere of 5% CO2 with the exception of XLK-WG (grown at 32°C). BHK-21, E. Derm, HeLa, HEp-2, LLC-PK1, MDBK, MDCK, Mv1Lu, RK-13, Tb1Lu, Vero and A549 cells (ATCC CCL-10, CCL-57, CCL-2, CCL-23, CL-101, CCL-22, CCL-34, CCL-64, CCL-37, CCL-88, CCL-81, and CCL-185 respectively) were maintained in modified Eagle media (MEM) with 2 mM L-glutamine and Earle's balanced salt solution (BSS) adjusted to contain 1.5 g/L sodium bicarbonate, 0.1 mM non-essential amino acids, 1.0 mM sodium pyruvate and 10% heat-inactivated, gamma-irradiated fetal bovine serum (FBS) (HyClone Laboratories, Salt Lake City, Utah). AK-D cells (CCL-150) were maintained in Ham's F-12K media containing 10% FBS. NCI-H292 (CRL-1848), MT-4 and XLK-WG (CRL-2527) cells were maintained in RPMI 1640 medium with 2 mM L-glutamine adjusted to contain 1.5 g/L sodium bicarbonate, 4.5 g/L glucose, 10 mM HEPES, 1.0 mM sodium pyruvate and 10% FBS. NIH/3T3, QT6 and 293T cells (CRL-1658, CRL-1708, and CRL-1573) were maintained in Dulbecco's modified Eagle media (DMEM) with 4 mM L-glutamine adjusted to contain 1.5 g/L sodium bicarbonate, 4.5 g/L glucose and 10% FBS. Cell lines were maintained at sub-confluence and used for only up to 15 passages after receiving initial stocks from the ATCC. Cell lines were tested and confirmed to be free of mycoplasma contamination. Human RSV (subgroup A, Long strain, ATCC VR-26) was obtained from the ATCC. Virus stocks were prepared by infecting HEp-2 cells with RSV at a multiplicity of infection (MOI) of 0.01 plaque-forming units (PFU) per cell. When cytopathic effect (CPE) was evident (~6 days post infection), the culture supernatant was collected and pooled with the supernatant from the infected cell pellet, which had been subjected to one freeze-thaw cycle. The pooled supernatants were maintained on ice, adjusted to 10% sucrose, flash frozen in liquid nitrogen and stored in liquid nitrogen. RSV titers were determined by plaque assay on HEp-2 cells. Serial dilutions of virus stock were added to monolayers of HEp-2 cells at 80% confluence and allowed to adsorb for 2 hours at 37°C. The virus inoculum was then removed, and cells were overlayed with media containing 0.5% methylcellulose. After plaques became apparent (5–6 days after infection), cell monolayers were fixed and stained with 0.5% crystal violet in 70% methanol, and plaques were counted. HRSV stock titers were typically >106 PFU/ml and remained stable for 6 months without loss of titer. A recombinant RSV rgRSV(224) engineered to express GFP has been previously described [33]. Stocks of rgRSV(224) were prepared as described above. Cell lines were infected with rgRSV(224) at a MOI of 0.1 and infection was visualized by fluorescent microscopy by monitoring fluorescence at 488 nm at 20, 48, and 120 hours post infection.
Plasmids
A DNA fragment encoding HRSV F protein derived from a known infectious cDNA sequence for subgroup A, A2 strain, [34] was synthesized with optimal codon usage for expression in mammalian cells and all potential polyadenylation sites (AATAAA) and splice donor sites (AGGT) removed essentially as described [15]. A similar construct was designed and synthesized for the B subgroup F protein (18537 strain, based upon GenBank accession number D00344). Sequence data is available from the authors upon request. Restriction sites for XbaI and BamHI were added to the 5' and 3' ends of the fragments respectively. The codon optimized HRSV-F DNA fragments (A2 and 18537 strains) were then cloned into the XbaI and BamHI sites of pcDNA 3.1 (Invitrogen, Inc., Carlsbad, CA) to generate pHRSVFOptA2 and pHRSVFOpt18537. The QuikChange® Site-Directed Mutagenesis kit (Stratagene®, La Jolla, CA) was used to change leucine 138 in the fusion peptide region of the F protein to an arginine (pL138R) in pHRSVFOptA2. Plasmids pBD-NFκB encoding the activation domain of NFκB fused to the GAL4 DNA binding domain under the control of the human cytomegalovirus (HCMV) promoter and pFR-Luc containing the luciferase reporter gene under the control of a minimal promoter containing five GAL4 DNA binding sites were obtained from Stratagene®. pGL3-control vector encodes a modified firefly luciferase under the control of the SV40 early promoter (Promega, Inc.). Plasmid pVPack-VSV-G encodes the G protein of vesicular stomatitis virus (Stratagene®, La Jolla, CA).
Transfections
Cells were transiently transfected using FuGENE 6 reagent (Roche Applied Science, IN) according to the manufacturer's recommendations. Briefly, 7.5 × 105 cells were plated in 6-well plates and grown overnight to ~90% confluence. Two micrograms of plasmid DNA was complexed with 6 μl of FuGENE 6 reagent for 30 minutes at room temperature in 100 μl of serum-free medium. The complex was then added to the cells and incubated at 37°C for 20–24 hours.
Metabolic labeling and immunoprecipitation
293T cells were plated the day before transfection in 6-well plates at a density of 0.75 × 106 cells/well in DMEM supplemented with 10% FBS. Cells in 6-well plates were transfected with a total of 2 μgs of plasmid DNA as described above. At 20 hours post-transfection, cells were starved by incubation in DMEM without L-methionine and L-cysteine containing 5% dialyzed FBS for 45 minutes. Cells were then labeled by incubation in DMEM without L-methionine and L-cysteine containing 5% dialyzed FBS and Redivue Pro-mix in vitro cell labeling mix containing (100 μCi/ml, 1.5 mls./well) [35S]-methionine and [35S]-cysteine (Amersham Biosciences, Piscataway, New Jersey) for 4 hrs. Media was removed, and cells were harvested and washed in 1 ml 1X phosphate-buffered saline (PBS) and then lysed with 0.5 mls. of lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% sodium deoxycholate, 1% IGEPAL (Sigma, St. Louis, MO) and Complete ™ protease inhibitor cocktail with EDTA (Roche Biochemicals, Indianapolis, IN). Lysates were spun for 30 minutes at 4°C to remove insoluble material and immunoprecipitated by incubation with a saturating amount (as determined by prior titration) of a cocktail containing 1.5 μgs of anti-F mAbs and protein-A agarose beads (Invitrogen, Inc., Carlsbad CA) overnight at 4?C. Immunoprecipitated complexes were washed three times in lysis buffer and suspended in 20 μl of 4X LDS NuPage loading buffer with reducing agent and resolved by electrophoresis through SDS-containing polyacrylamide gels (SDS-PAGE) under reducing conditions on a NuPage 4–12% Bis-Tris polyacrylamide gel (Invitrogen, Inc., Carlsbad CA). Gels were dried under vacuum for one hour at 80°C followed by autoradiography.
Flow cytometry
To confirm cell surface expression of HRSV F, 293T cells were transfected in 6-well plates as described above for metabolic labeling. Cells were stained with palivizumab (Synagis ®, IgG1κ) at 1 μg/ml in conjunction with an anti-human IgG-Alexa-Fluor-488 conjugated secondary (Molecular Probes, Eugene, OR) at 1 μg/ml in 1X PBS with 2% FBS for analysis with the FACSCalibur (BD Bisociences, CA) by setting a live cell gate in the FSC/SSC plot and determining the mean fluorescence intensity in the FL1 channel. Data analysis was performed with Cell Quest and FloJo Analysis Software.
Cell fusion assays
293T cells were co-transfected with pHRSVFOptA2 and pBD-NFκB (effectors cells), and the panel of cell lines from a variety of different species were transfected using FuGene-6 (Roche Biochemicals, Inc.) with the pFR-Luc reporter plasmid (reporter cells) using conditions described above. Alternatively, 293T cells were co-transfected with pHRSVFOptA2 and pFR-Luc, and the panel of cell lines from a variety of different species were transfected with the pBD-NFκB reporter plasmid. A plasmid encoding the vesicular stomatitis virus (VSV) G protein linked to the HCMV promoter (pVPack-VSV-G) was used as a positive viral fusion protein control. At 24 hours post transfection, unless otherwise specified, 3 × 104 of the transfected 293T (effector cells) were mixed with an equal amount of the various other transfected cells (target cells) in a 96-well plate and incubated an additional 24 hours prior to measurement of luciferase activity using the Steady Glo Luciferase reporter system (Promega, Inc.). The various cell lines were transfected with pGL3-control to determine relative differences in transfection efficiencies and cell-type specific expression of luciferase. To further normalize, a single preparation of effector cells was used per each experiment on the various reporter cell preparations. Cells individually transfected with the F expression plasmids, pBD-NFκB or pFR-Luc individually were used as negative controls.
Results
Expression of RSV F protein
We designed and synthesized a cassette encoding the full-length HRSV-F gene (A2 strain and 18537 strains) in which codon usage was optimized for mammalian expression and all potential splice-donor sites and polyadenylation sites were removed similar to a previous report [15]. Upon transient transfection of 293T cells with this plasmid expressing this optimized HRSV F protein sequence under the control of a human cytomegalovirus immediate early promoter (pHRSVFoptA2), giant multinucleated cells (syncytia) were readily apparent within 24 hours post transfection (Figure 1A). The amount of syncytia qualitatively increased throughout the culture for up to 72 hours, after which extensive cell death and sloughing was observed. This syncytia is phenotypically indistiguishable from that observed following infection of 293T cells with HRSV in tissue culture (data not shown). To confirm appropriate processing of the F protein, 293T cells were transfected with plasmids expressing HRSV F derived from subgroup A (A2 strain) or subgroup B (18537 strain) and metabolically labeled followed by immunoprecipitation of lysates with HRSV F-specific monoclonal antibodies. As a control, 293T cells were infected with HRSV (Long strain). As shown in Figure 1B, immunoprecipitation demonstrates the presence of the unprocessed full length F0 species migrating at approximately 70 kDa, and the processed F1 and F2 fragments of ~50 kDa and 20 kDa, respectively, identical to the F protein immunoprecipitated from HRSV infected 293T cells. The multiple bands observed in the region of 20 kDa likely represent the incompletely processed F2 (F2+), different glycosylated forms of F2, or a combination of both [21]. The band present migrating more rapidly than F1 (~35 kDa) most likely represents a cellular protein as this band was observed in lysates derived from untransfected and control vector transfected cells with varying intensity unrelated to the level of HRSV F expressed (see Figure 2B, lanes 3 and 4). Similar levels of expression were observed for the HRSV F protein from the A and B subgroups (Figure 1B, lane 3 compared to lane 4). Furthermore, the level of F protein expression in the transfected cells was greater at 24 hours post transfection than in HRSV-infected cells at 24 hours post infection (Figure 1B, lane 5). Flow cytometry confirmed abundant cell surface expression of HRSV F protein (Figure 1C).
Figure 1 A) Syncytia formation by RSV-F DNA in transfected cells. 293T cells were mock transfected or transfected with pHRSVFOptA2 and visualized by light microscopy 48 hours post transfection. The arrow indicates giant multinucleated cell formation. B) Processing of RSV F in transfected cells. 293T cells were either mock transfected (lane 1), transfected with pCMV-β-gal (lane 2), transfected with pHRSVFOptA2 (lane 3), pHRSVFOptB18537 (lane 4), or infected with RSV (Long strain, MOI = 1) for 24 hours followed by metabolic labeling for 6 hours with [35S]-cysteine/methionine. Labeled cell lysates were immunoprecipitated with HRSV F specific mAbs, and immunoprecipitates were resolved by SDS-PAGE as described in methods. C) Cell surface expression of RSV F in transfected cells. 293T cells were either mock transfected or transfected with pHRSVFOptA2 for 24 hours followed by flow cytometry using HRSV F specific monoclonal antibodies as described in methods.
Figure 2 A) Dose dependent fusion activity of HRSV F derived. 293T cells were transfected with either pFR-Luc alone (■), pBD-NFκB alone (▲), co-transfected with pFR-Luc and pBD-NFκB (▼), or co-transfected with pHRSVFOptA2 and pBD-NFκB and mixed 24 hours after transfection in various amounts with cells that had been transfected with pFR-Luc alone (◆). Luciferase activity was measured 24 hours post mixing as described in methods and is reported as relative light units. B) Fusion activity of HRSV F derived from subgroups A and B. 293T cells co-transfected with pBD-NFκB and either pHRSVFOptA2, pHRSVFOptB18537, or vector only (NFκB only). Cells were mixed 24 hours later with a separate population of 293T cells transfected with pFR-Luc, and luciferase activity was measured 24 hours post mixing as described in methods. Luciferase activity is reported as relative light units.
HRSV F protein fusion assay
To measure the ability of the HRSV F protein to mediate cell fusion across various cell lines, we developed a quantitative fusion assay. Specifically, 293T cells were co-transfected with plasmids encoding the optimized HRSV F protein and a transcriptional transactivator fusion protein consisting of the GAL4 DNA-binding domain fused to the activation domain of NFκB. These effector cells were mixed with a separate set of reporter cells that were transfected with a reporter plasmid containing the luciferase gene under the control of a GAL4 responsive promoter. HRSV F mediated cell fusion of the two cell populations results in co-localization of the GAL4-NFκB transactivator fusion protein with the GAL4 responsive luciferase reporter plasmid and subsequent transcriptional transactivation of the reporter gene. A dilution series from 50,000 to 100 effector cells were added to a fixed amount of reporter cells (50,000), and luciferase activity was monitored 24 hours later. As a control to determine the maximum signal, cells were co-transfected with the reporter and activator plasmids (pFR-Luc + pBD-NFκB). As shown in figure 2A, luciferase activity was absent in cells transfected with either reporter or activator plasmids alone. Additionally, mixing of cells which had been separately transfected with the reporter or activator plasmids did not produce detectable luciferase activity indicating no spontaneous cell fusion (data not shown). However, mixing an increasing number of cells that had been co-transfected with the GAL4-sensitive reporter plasmid and HRSV F expressing plasmid with those that had been transfected with the GAL-NFκB activator plasmid resulted in a dose-dependent increase in luciferase activity (Figure 2A), indicating fusion of the two cell populations. The maximum signal observed from mixing the effector and reporter populations was similar to the signal obtained when the activator and reporter plasmids were co-transfected into the same cells.
To determine if this property was restricted to the F protein derived from a single strain or subgroup, 293T cells co-transfected with a plasmid encoding the HRSV F protein derived from either subgroup A (A2 strain) or B (18537 strain) together with a plasmid encoding the GAL4-NFκB transactivator fusion protein (effector cells) were then mixed 24 hours later with an equal amount of a separate population of 293T cells (reporter cells) which had been transfected with the GAL4 responsive reporter plasmid. As shown in Figure 2B, the F protein of either HRSV subgroup A and B mediated cell-cell fusion as measured by the increased luciferase activity relative to the negative control (GAL4-NFκB transactivator fusion protein only). The fusion activity of the F protein derived from the A2 strain was approximately 2-fold higher than that observed with the 18537 strain despite similar expression levels. Whether this finding reflects differences in the pathogenicity between these two isolates is unknown, although a recent study suggests similar pathogenicity for both subgroups [35]. To further confirm that the observed fusion activity was inherent to the HRSV F protein, a point mutation (L138R) was generated in the fusion peptide consensus sequence within the fusion peptide region. Mutation of leucine residue 138 to arginine reduced fusion activity to 10% relative to wild-type (Figure 3A). Despite the fact that this mutant appeared to produce somewhat lower levels of fully processed F protein (Figure 3B, lane 2) for unknown reasons, this mutant was expressed at high levels on the cell surface (Figure 3C) indicating that the cell fusion observed was attributable to the HRSV F protein.
Figure 3 A) Comparison of fusion activity of wild-type and a fusion peptide mutant of HRSV F. 293T cells co-transfected with pBD-NFκB and either pHRSVFOptA2 or pL138R were mixed 24 hours later with a separate population of 293T cells that had been transfected with pFR-Luc, and luciferase activity was measured 24 hours post mixing as described in methods. Luciferase activity is reported as relative light units. B) Processing of the wild-type and L138R mutant of HRSV F was determined by metabolic labeling 293T cells transfected with either pHRSVFOptA2 (lane 1), pL138R (lane 2), pCMV-β-gal (lane 3), or mock transfected (lane 4) for 6 hours with [35S]-cysteine/methionine followed by immunoprecipitation of lysates with HRSV F specific mAbs, and analysis of immunoprecipitates by SDS-PAGE as described in methods. C) Cell surface expression of the wild-type and L138R mutant F proteins in 293T cells transfected with either pHRSVFOptA2 or pL138R was compared by flow cytometry as described in methods.
Host range of HRSV-mediated fusion
To determine the host range of HRSV F mediated fusion using the quantitative fusion assay. 293T effector cells were prepared as described above. As we previously demonstrated proper protein processing, abundant cell surface expression of HRSV F protein and cell fusion activity using 293T cells, we selected these as our effector cells. These effector cells were mixed with reporter cells derived from a diverse range of species (Table 1) that were transfected with the GAL4-responsive reporter plasmid. To account for any differences in relative transfection efficiencies and expression of the luciferase reporter among the various target cells lines, the target cell lines were transfected with a plasmid containing the luciferase gene under the control of the SV40 early promoter (pGL3-control), and relative luciferase activity was measured. To account for potential differences in host cell transcription factors that mediate activation of the reporter plasmid, the assay was flipped and 293T cells were co-transfected with the HRSV F expression plasmid and the GAL4 responsive reporter plasmid, and the cells from the various species were transfected with the GAL4-NFκB transactivator fusion protein plasmid. For further comparison, we used the VSV G protein, which is known to mediate entry into cells derived from a wide range of species.
Table 1 Species and tissue origin of cell lines used in this study are listed.
Cell line Species, tissue
XLK-WG Xenopus laevis (S. African clawed frog), kidney
QT6 Coturnix coturnix japonica (Japanese quail), fibrosarcoma
Tb1Lu Tadarida brasiliensis (free-tailed bat), lung
NIH/3T3 Mus musculus (mouse), fibroblast
BHK-21 Mesocricetus auratus (Syrian golden hamster), kidney
RK-13 Oryctolagus cuniculus (rabbit), kidney
LLC-PK1 Sus scrofa (pig), kidney
Mv1Lu Musteal vison (mink), lung
AK-D Felis catus (domestic cat), fetal epithelial
MDCK Canis familiaris (domestic dog), kidney
MDBK Bos taurus (cow), kidney
E. Derm Equus caballus (horse), dermal
Vero Cercopithecus aethiops (African green monkey), kidney
HEp-2 Homo sapiens (human), laryngeal carcinoma
HeLa Homo sapiens (human), cervical carcinoma
MT-4 Homo sapiens (human), T-cell
293T Homo sapiens (human), kidney
NCI-H292 Homo sapiens (human), epidermoid pulmonary carcinoma
A549 Homo sapiens (human), lung
As shown in figure 4, despite a limited host range in nature, HRSV F was able to mediate fusion to various degrees with cells derived from all species examined. This fusion activity was within 5-fold of the fusion activity mediated by the VSV G protein in the cell types tested here. Generally, there was little qualitative difference between results obtained when either the reporter plasmid or the activator plasmid were co-transfected with the F expression plasmid (compare figures 4A and 4B with figures 4C and 4D). As expected, the relative transfection efficiencies of the various cell lines as measured by the luciferase activity from the plasmid pGL3-control varied; however, there was no direct correlation between transfection efficiencies and fusion activity. For example, cell lines such as BHK-21 and LLC-PK1 cells which transfected well, had lower relative levels of fusion. In contrast, cell lines such as MT-4, MDCK and XLK-WG which had low transfection efficiency, had appreciable levels of HRSV F mediated fusion. These findings support the hypothesis that HRSV F protein interacts with evolutionarily conserved host cell surface molecules or can use multiple mechanisms to enter cells.
Figure 4 Fusion activity of HRSV F with cell lines derived from various species. Cell lines derived from various species (target cells) were either transfected with pFR-Luc and mixed 24 hours later with 293T cells that had been co-transfected for 24 hours with pHRSVFOptA2 or pVPack-VSV-G and pBD-NFκB (Figs. 4A and 4B), or the target cells were transfected with pBD-NFκB and mixed 24 hours later with 293T cells that had been co-transfected for 24 hours with pHRSVFOptA2 or pVPack-VSV-G together with pFR-Luc (Figs. 4C and 4D). Cell lines derived from various species were transfected with either pFR-Luc or pBD-NFκB only as negative controls. Luciferase activity was measured 24 hours post mixing of the cell populations as described in methods and is reported as relative light units.
Infections using recombinant HRSV expressing GFP
The results obtained from the fusion assays indicated that HRSV F is able to mediate fusion with cells from multiple diverse species, suggesting that virus entry is not the primary determinant of host range. To examine whether viral mRNA transcription had occurred, the various cell lines were infected with a recombinant HRSV (rgRSV224) expressing GFP [33] and fluorescence scored at 20, 48, and 120 hours post infection. As expected, rgRSV(224) infection of human (HEp-2, HeLa, A549) and animal (Vero, Mv1Lu, MDBK) [36-39] cell lines commonly used to propagate HRSV resulted in a time dependent increase in the number of cells expressing GFP (≥50% by day 5) as seen by fluorescent microscopy indicating spread of infection throughout the culture (Figure 5). Infection of other human cell lines such as NCI-H292 [40], and 293T also resulted in a time dependent increase in the number of cells expressing GFP. Infection of hamster BHK-21 cells also resulted in a time dependent increase in the number of GFP positive cells, although the appearance of a large number of bright GFP positive cells seemed delayed. Interestingly, hamsters are considered to be a semi-permissive host for HRSV [27,41] and produce lung titers similar to those achieved in mice. Whether this reflects a tissue-specific phenomenon (kidney versus lung) remains to be determined. Infection of cell lines (Tb1Lu, AK-D, E. Derm, NIH/3T3, LLC-PK1, and XLG-WG) derived from other species (bat, cat, horse, mouse, and frog respectively) produced few or occasional GFP expressing cells over the course of the five-day infection. The number of positive cells did not increase over time, and in some cases (AK-D cells) appeared to decrease. Aside from mice, infection in vivo of these other species by HRSV has not been described. This finding also supports the finding that high titers of virus (>105 PFU) are typically required to initiate infection in mice after intranasal inoculation, and that relatively few cells become viral antigen positive.
Figure 5 Infection of various cell lines by rg224(RSV). Cell lines derived from various species were infected with rgRSV(224) at an MOI = 0.1 and GFP-expressing cells were visualized at 20, 48, and 120 hours post infection by fluorescent microscopy by monitoring fluorescence at 488 nm.
Discussion
We have developed a quantitative reporter gene based cell-cell fusion assay for HRSV F. Prior assays have been based upon visual counting of plaques or syncytia after staining of infected monolayers, or infection another virus such as vaccinia, to provide HRSV F protein, which could potentially complicate interpretation. The assay described herein is a means of quantifying the fusion activity of the HRSV F protein. This assay has multiple applications. For example, this assay can be used as a means of studying the structure-function of the HRSV F protein, or for evaluating the activity of mutations in the F protein without the need to select for antibody or compound escape mutants or generate point mutations in a reverse genetics system. We propose that this assay also has utility in the identification and characterization of inhibitors of HRSV entry for the development of specific agents to prevent and treat HRSV infections. We have used this assay as a means of exploring the host-range of HRSV and have shown that the HRSV F protein is able to mediate fusion with cells derived from a wide range of vertebrate species.
Cell lines known to be permissive for HRSV growth such as HEp-2, HeLa, A549, Vero, MDBK, and Mv1Lu were highly competent for F protein fusion as expected. Somewhat surprisingly, a wide variety of cells derived from species not known to be normally infected by HRSV were also capable of undergoing HRSV F protein mediated fusion. Most surprising were the results obtained using the XLK-WG cells which are derived from the amphibian Xenopus laevis. Although this finding implies that HRSV virion is able to enter a wide range of cells, the results of the infection studies using the GFP-expressing RSV indicate that viral mRNA transcription seems limited in cell lines derived from certain species. Taken together these results suggest that events post-viral entry are the primary determinants that mediate the host range of HRSV. During natural infection of humans, viral replication is restricted to epithelial cells of the upper and lower respiratory tract. Although limited HRSV replication within human alveolar macrophages and detection of HRSV sequences in peripheral blood monocytes (PBMCs) has also been reported [42,43], dissemination of HRSV to other organs is not observed even in immunocompromised individuals. Similarly, disseminated infection with bovine RSV is not observed in infected cattle [44]. Given the ability of the HRSV F protein to mediate fusion with cells derived from a diverse range of vertebrate species, the implication is that HRSV may not be able to access these sites or undergoes non-productive infection in many cell types other than epithelial cells of the respiratory tract. Although the overall biological significance of such an abortive infection is unclear, biological effects of the individual HRSV proteins have been reported.
HRSV F protein has also been shown to be a ligand for TLR4, and HRSV infection persists longer in TLR4-/- deficient mice [45,46]. HRSV F protein also binds surfactant proteins A and D (SP-A and SP-D) [47,48], although the implications of these findings in human infection are unclear. G protein has been shown to modulate multiple immune related activities. Soluble G suppresses some PBMC and lung CD8+ T-cell effector and peripheral memory responses [49], induces chemotaxis, eosinophilia, and both soluble and membrane forms of G bind the fractalkine receptor, CX3CRI [50]. Additionally, G has a domain with similarity to the TNF-α receptor (p55), although it has not been directly shown to be a TNFR antagonist. Additionally, G has been shown to modify CC and CXC chemokine mRNA expression [50], and suppress lymphoproliferative responses to antigens in PBMCs [51].
It is tempting to speculate that entry of HRSV into cell types other than those permissive for complete virus growth may be a strategy by which the virus is able to modulate immune responses while avoiding the induction of antiviral responses such as the interferon (IFN) pathway by production of double-stranded RNA replication intermediates in these cells. Limited viral mRNA transcription in the absence of virus RNA replication would result in expression of NS1 and NS2 which have been shown to block the IFN response [52] possibly preventing these unproductively infected cells from responding to external cytokines such as IFNs. Such a strategy may help explain why despite little antigenic drift in the F protein, infection by HRSV infection only confers partial protection, with reinfections occurring throughout life [53-55]. As the fusion proteins of other members of the Paramyxoviridae family, such as Hendra virus [56], are also able to mediate fusion with a wide variety of cells derived from multiple species, it is possible that such a strategy is shared by other members of this virus family.
Competing interests
The author(s) declare that they are all employees of Centocor, Inc. which provided supported for this work.
Authors' contributions
PB and CL contributed equally to this work. PB and ND performed the fusion assays, immunoprecipitations, and flow cytometry. CL generated reagents and developed the fusion assay. LG conducted site-directed mutagenesis of the HRSV F protein. AD and RS participated in the design of the experiments, oversight of the conduct of the experiments, and in the interpretation of the results.
Table 2 Infection of various cell lines with GFP-expressing HRSV.
Cell line 20 hrs 48 hrs 120 hrs
Vero +++ ++++ ++++
AK-D + ++ -
MDBK +++ ++++ +++
MDCK + + +
Tb1Lu + + +
XLK-WG - + +
E. Derm ++ ++ +
HeLa ++ +++ ++++
NCI-H292 ++ +++ +++
293T ++ ++++ ++++ *
HEp-2 +++ ++++ ++++
Mv1Lu +++ ++++ +++
NIH/3T3 + + +
LLC-PK1 + - -
RK-13 ++ ++ ++
BHK-21 - ++ +++
QT6 + ++ ++++
A549 + +++ ++
MT-4 + ++ ++
- = few isolated weakly positive cells
+ = <1–5% GFP positive cells in culture
++ = 5–30% GFP positive cells in culture
+++ = 30–60% GFP positive cells in culture
++++ = >60% GFP positive cells in culture
* wide spread cell death
Acknowledgements
Recombinant HRSV expressing green fluorescent protein rgRSV(224) was generously provided by Dr. Peter Collins (NIAID, NIH). We thank Jose Melero, Geraldine Taylor, and Paul Bates for helpful discussion and comments, and Lamine Mbow, Lani San Mateo, and William Glass for critical review of this manuscript.
==== Refs
Shay DKHRCNRDLLLSJWLJA Bronchiolitis-associated hospitalizations among U.S. children, 1980-1996 JAMA 1999 282 1440 1446 10535434 10.1001/jama.282.15.1440
Shay DKHRCRGECMJALJ Bronchiolitis-associated mortality and estimates of respiratory syncytial virus-associated deaths among U.S. children, 1979-1997 J Infect Dis 2001 183 16 22 11076709 10.1086/317655
Han LLJPALJA Respiratory syncytial virus pneumonia among the elderly: an assessment of disease burden J Infect Dis 1999 179 25 30 9841818 10.1086/314567
Ison MGHFG Viral infections in immunocompromised patients:what's new with respiratory viruses? Curr Opin Infect Dis 2002 15 355 367 12130931
Collins PLCRMMBR D.M. Knipe HPM Respiratory syncytial virus Fields Virology 2001 1 Philadelphia, Lippincott, Williams, and Wilkins 1443 1485
Karron RADABAFGSSWJEAMLCMDOHVBRSUBRMMSS Respiratory syncytial virus (RSV) SH and G proteins are not essential for viral replication in vitro: clinical evaluation and molecular characterization of a cold-passaged, attenuated RSV subgroup B mutant Proc Natl Acad Sci USA 1997 94 13961 13966 9391135 10.1073/pnas.94.25.13961
Teng MNSSWPLC Contribution of the respiratory syncytial virus G glycoprotein and its secreted and membane-bound forms to virus replication in vitro and in vivo Virology 2001 289 283 296 11689051 10.1006/viro.2001.1138
Bukreyev AWSSMBRCPL Recombinant respiratory syncytial virus from which the entire SH gene has been deleted grows efficiently in cell culture and exhibits site-specific attenuation in the respiratory tract of the mouse. J Virol 1997 71 8973 8982 9371553
Jin HZHCXTRMMNN Recombinant respiratory syncytial viruses with deletions in the NS1, NS2, SH, and M2-2 genes are attenuated in vitro and in vivo Virology 2000 273 210 218 10891423 10.1006/viro.2000.0393
Schlender JGZGHKKC Respiratory syncytial virus (RSV) fusion protein subunit F2, not attachment protein G, determines the specificity of RSV infection J Virol 2003 77 4609 4616 12663767 10.1128/JVI.77.8.4609-4616.2003
Techaarpornkul SBNPME Functional analysis of recombinant respiratory syncytial virus deletion mutants lacking the small hydrophobic and/or attachment glycoprotein J Virol 2001 75. 6825 6834 11435561 10.1128/JVI.75.15.6825-6834.2001
Whitehead SSABMNTCYFMSCWREPLCBRM Recombinant respiratory syncytial virus bearing a deletion of either the NS2 or SH gene is attenuated in chimpanzees J Virol 1999 73 3438 3442 10074199
Techaarpornkul SPLCMEP Respiratory syncytial virus with the fusion protein as its only viral glycoprotein is less dependent on cellular glycosaminoglycans for attachment than complete virus Virology 2002 294 296 304 12009871 10.1006/viro.2001.1340
Teng MNPLC Identification of the respiratory syncytial virus proteins required for formation and passage of helper-dependent infectious particles J Virol 1998 72 5707 5716 9621029
Morton CJRCLJLBLMLALAJMMKBMWPJRMSJSSPTPRY Structural characterization of respiratory syncytial virus fusion inhibitor escape mutants: homology model of the F protein and a syncytium formation assay Virology 2003 311 275 288 12842618 10.1016/S0042-6822(03)00115-6
Heminway BRYYYTKGPEGJMBMSG Analysis of respiratory syncytial virus F, G, and SH proteins in cell fusion Virology 1994 200 801 805 8178462 10.1006/viro.1994.1245
Bolt G POLBHH Cleavage of the respiratory syncytial virus fusion protein is required for its surface expression: role of furin Virus Res 2000 68 25 33 10930660 10.1016/S0168-1702(00)00149-0
Collins PLMG Post-translational processing and oligomerization of the fusion glycoprotein of human respiratory syncytial virus J Gen Virol 1991 72 3095 3101 1765771
Gonzalez-Reyes LMBRABGBLCJAJPAJJSDCWJAM Cleavage of the human respiratory syncytial virus fusion protein at two distinct sites is required for activation of membrane fusion Proc Natl Acad Sci 2001 98 9859 9864 11493675 10.1073/pnas.151098198
Sugrue RLCBGBJAHWMLR Furin cleavage of the respiratory syncytial virus fusion protein is not a requirement for its transport to the surface of virus-infected cells J Gen Virol 2001 82 1375 1386 11369882
Rixon HWMLCBGBRJS Multiple glycosylated forms of the respiratory syncytial virus fusion protein are expressed in virus-infected cells J Gen Virol 2002 83 61 66 11752701
Matthews JMTFYSPTJPM The core of the repiratory syncytial virus fusion protein is a trimeric coiled coil J Virol 2000 74 5911 5920 10846072 10.1128/JVI.74.13.5911-5920.2000
Zhao XMSVNMPSK Structural characterization of the human respiratory syncytial virus fusion protein core Proc Natl Acad Sci USA 2000 97 14172 14177 11106388 10.1073/pnas.260499197
Lawless-Delmedico MKPSRSNCMJBAJMWRJGKCLTJMDML Heptad-repeat regions of respiratory syncytial virus F1 protein form a six-membered coiled-coil complex Biochemistry 2000 39 11684 11695 10995236 10.1021/bi000471y
Zhang LMEPRCBPLCRJP Respiratory syncytial virus infection of human airway epithelial cells is polarized, specific to ciliated cells, and without obvious cytopathology J Virol 2002 76 5654 5666 11991994 10.1128/JVI.76.11.5654-5666.2002
Sarmiento RERTBG Characteristics of a respiratory syncytial virus persistently infected macrophage-like culture Virus Res 2002 84 45 58 11900838 10.1016/S0168-1702(01)00420-8
Byrd LGPGA Animal models of respiratory syncytial virus infection Clin Infect Dis 1997 25 1363 1368 9431379
Bourgeois CBJBLKGCPP Heparin-like structures on respiratory syncytial virus are involved in its infectivity in vitro J Virol 1998 72 7221 7227 9696816
Hallak LKDSPLCMEP Glycosaminoglycan sulfation requirements for respiratory syncytial virus infection J Virol 2000 74 10508 10513 11044095 10.1128/JVI.74.22.10508-10513.2000
Feldman SASAJAB The fusion glycoprotein of human respiratory syncytial virus facilitates virus attachment and infectivity via an interaction with cellular heparan sulfate J Virol 2000 74 6442 6447 10864656 10.1128/JVI.74.14.6442-6447.2000
Krusat THJS Heparin-dependent attachment of respiratory syncytial virus (RSV) to host cells Arch Virol 1997 12 1247 1254 9229012 10.1007/s007050050156
Martinez IJAM Binding of human respiratory syncytial virus to cells:implication of sulfated cell surface proteoglycans J Gen Virol 2000 81 2715 2722 11038384
Hallak LKPLCWKMEP Iduronic acid-containing glycosaminoglycans on target cells are required for efficient respiratory syncytial virus infection Virology 2000 271 264 275 10860881 10.1006/viro.2000.0293
Collins PLHMGCEGHCRMMBR Production of infectious human respiratory syncytial virus from cloned cDNA confirms an essential role for the transcription elongation factor from the 5' proximal open reading frame of the M2 mRNA in gene expression and provides a capability for vaccine development Proc Natl Acad Sci USA 1995 92 11563 11567 8524804
Devincenzo JP Natural infection of infants with respiratory syncytial virus subgroups A and B: a study of frequency, disease severity, and viral load Pediatr Res 2004 56 914 917 15470202 10.1203/01.PDR.0000145255.86117.6A
Barenfanger JCDTMTTJOBKG R-Mix cells are faster, at least as sensitive and marginally more costly than conventional cell lines for detection of respiratory viruses. J Clin Virol 2001 22 101 110 11418357 10.1016/S1386-6532(01)00171-8
Bossert BKKC Respiratory syncytial virus (RSV) nonstructural (NS) proteins as host range determinants: a chimeric bovine RSV with NS genes from human RSV is attentuated in interferon-competent bovine cells. J Virol 2002 76 4287 4293 11932394 10.1128/JVI.76.9.4287-4293.2002
Huang YTBMT Mink lung cells and mixed mink lung and A549 cells for rapid detection of influenza virus and other respiratory viruses. J Clin Microbiol 2000 38 422 423 10618131
Parry JEPVSCRP Pneumoviruses: the cell surface of lytically and persistently infected cells. J Gen Virol 1979 44 479 491 118236
Hierholzer JCCEBGGBJAMECT Sensitivity of NCI-H292 human lung mucoepidermoid cells for respiratory and other human viruses. J Clin Microbiol 1993 31 1504 1510 8314992
Wright PFWWGRMC Temperature-sensitive mutants of respiratory syncytial virus: in vivo studies in hamsters. J Infect Dis 1970 122 501 512 5489074
Cirino NMPJRVATHRNAMRTPGIA Restricted replication of respiratory syncytial virus in human alveolar macrophages J Gen Virol 1993 74 1527 1537 8345347
Yui IAHYSTTTN Detection of human respiratory syncytial virus sequences in peripheral blood mononuclear cells J Med Virol 2003 70 481 489 12767015 10.1002/jmv.10421
Viuff BAUCTSA Sites of replication of bovine respiratory syncytial virus in naturally infected calves as determined by in situ hybridization Veterinary Pathology 1996 33 383 390 8817835
Haynes LMMDDKJEAFRWALJTRA Involvement of toll-like receptor 4 in innate immunity to respiratory syncytial virus J Virol 2001 75 10730 10737 11602714 10.1128/JVI.75.22.10730-10737.2001
Kurt-Jones EAPLKLHLMJLPTRAWEEFMWGDTALJFRW Pattern recognition receptors TLR4 and CD14 mediate response to respiratory syncytial virus Nat Immunol 2000 1 398 401 11062499 10.1038/80833
Hickling TPHBKWDGSLMRBSRM A recombinant trimeric surfactant protein D carbohydrate recognition domain inhibits respiratory syncytial virus infection in vitro and in vivo Eur J Immunol 1999 29 3478 3784 10556802 10.1002/(SICI)1521-4141(199911)29:11<3478::AID-IMMU3478>3.3.CO;2-N
Ghildyal RCHAVJMDRVEMAJM Surfactant protein A binds to the fusion glycoprotein of respiratory syncytial virus and neutralizes virion infectivity J Infect Dis 1999 180 2009 2013 10558959 10.1086/315134
Chung WBTJ Respiratory syncytial virus infection suppresses lung CD8+ T-cell effector activity and peripheral CD8+ T-cell memory in the respiratory tract. Nat Med 2002 8 54 60 11786907 10.1038/nm0102-54
Tripp RAJLALJ Respiratory syncytial virus G and/or SH glycoproteins modify CC and CXC chemokine mRNA expression in the BALB/c mouse J Virol 2000 74 6227 6229 10846112 10.1128/JVI.74.13.6227-6229.2000
Ray RHDFMKBRLLMBRB Immunoregulatory role of secreted glycoprotein G from respiratory syncytial virus Virus Res 2001 75 147 154 11325469 10.1016/S0168-1702(01)00237-4
Spann KMKCTBCRLRPLC Suppression of the induction of alpha, beta, and lambda interferons by the NS1 and NS2 proteins of human respiratory syncytial virus in human epithelial cells and macrophages J Virol 2004 78 4363 4369 15047850 10.1128/JVI.78.8.4363-4369.2004
Dowell SFALJGHEJEDDPJFFTMJMBJBRF Respiratory syncytial virus is an important cause of community-acquired lower respiratory infection among hospitalized adults. J Infect Dis 1996 174 456 462 8769600
Falsey AREEW Respiratory syncytial virus infection in adults Clin Microbiol Rev 2000 13 371 10885982 10.1128/CMR.13.3.371-384.2000
Glezen WPTLHFALKJA Risk of primary infection and reinfection with respiratory syncytial virus Am J Dis Child 1986 140 543 546 3706232
Bossart KNWLFFMNCKBLSKEBTBCC Membrane Fusion tropism and heterotypic functional activities of the Nipah virus and Hendra virus envelope glycoproteins J Virol 2002 76 11186 11198 12388678 10.1128/JVI.76.22.11186-11198.2002
|
16014172
|
PMC1190219
|
CC BY
|
2021-01-04 16:39:00
|
no
|
Virol J. 2005 Jul 13; 2:54
|
utf-8
|
Virol J
| 2,005 |
10.1186/1743-422X-2-54
|
oa_comm
|
==== Front
Virol JVirology Journal1743-422XBioMed Central London 1743-422X-2-541601417210.1186/1743-422X-2-54MethodologyUse of a novel cell-based fusion reporter assay to explore the host range of human respiratory syncytial virus F protein Branigan Patrick J [email protected] Changbao [email protected] Nicole D [email protected] Lester L [email protected] Robert T [email protected] Vecchio Alfred M [email protected] Infectious Diseases Research, Centocor, Inc., 145 King of Prussia Road, Radnor, PA, 19087, USA2005 13 7 2005 2 54 54 9 6 2005 13 7 2005 Copyright © 2005 Branigan et al; licensee BioMed Central Ltd.2005Branigan et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Human respiratory syncytial virus (HRSV) is an important respiratory pathogen primarily affecting infants, young children, transplant recipients and the elderly. The F protein is the only virion envelope protein necessary and sufficient for virus replication and fusion of the viral envelope membrane with the target host cell. During natural infection, HRSV replication is limited to respiratory epithelial cells with disseminated infection rarely, if ever, occurring even in immunocompromised patients. However, in vitro infection of multiple human and non-human cell types other than those of pulmonary tract origin has been reported. To better define host cell surface molecules that mediate viral entry and dissect the factors controlling permissivity for HRSV, we explored the host range of HRSV F protein mediated fusion. Using a novel recombinant reporter gene based fusion assay, HRSV F protein was shown to mediate fusion with cells derived from a wide range of vertebrate species including human, feline, equine, canine, bat, rodent, avian, porcine and even amphibian (Xenopus). That finding was extended using a recombinant HRSV engineered to express green fluorescent protein (GFP), to confirm that viral mRNA expression is limited in several cell types. These findings suggest that HRSV F protein interacts with either highly conserved host cell surface molecules or can use multiple mechanisms to enter cells, and that the primary determinants of HRSV host range are at steps post-entry.
==== Body
Background
Human respiratory syncytial virus (HRSV) is the single most common cause of serious lower respiratory tract infections (LRTIs) in infants and young children causing up to 126,000 hospitalizations annually in the U.S. [1] with an estimated 500 deaths per year [2]. HRSV bronchiolitis has been associated with the development and exacerbation of wheezing and other respiratory conditions. Furthermore, HRSV is an increasingly recognized cause of pneumonia, exacerbation of chronic pulmonary and cardiac disease, and death in the elderly [3]. HRSV is also the most common viral respiratory infection of transplant recipients and is responsible for high rates of mortality in this group [4].
HRSV is member of the subfamily Pneumovirinae in the Paramyxoviridae family [5]. Two serologic subgroups (A and B) have been described and co-circulate throughout the world. Three viral transmembrane proteins (G, SH and F) are found on surface of the virus particle in the viral envelope. The G protein is a heavily glycosylated protein that mediates attachment of the virus to host cells, and although not strictly required for virus replication in culture, recombinant viruses lacking the G protein are attenuated in animals [6,7]. While its exact function is unknown, the SH gene is not essential for virus growth in tissue culture, and its deletion only results in slight attenuation in animals [6,8-12]. The F protein is a type 1 membrane protein essential for the packaging and formation of infectious virion particles [6,7,13,14], and is the only viral protein necessary and sufficient for fusion of the viral envelope membrane with the target host cell [15,16]. The HRSV F protein is highly conserved (89% amino acid identity) between subgroups A and B, and shares 81% amino acid identity with the F protein of bovine respiratory syncytial virus (BRSV). The HRSV F protein is synthesized as a 574 amino acid precursor protein designated F0, which is cleaved at two sites [17-20] within the lumen of the endoplasmic reticulum removing a short, glycosylated intervening sequence and generating two subunits designated F1 and F2 [21]. The mature form of the F protein present on the surface of the virus and infected cells is believed to consist of a homotrimer consisting of three non-covalently associated units of F1 disulfide linked to F2[22]. As with many other viral fusion proteins, F-mediated fusion with the host cell membrane is believed to be mediated by insertion of a hydrophobic fusion peptide into the host cytoplasmic membrane after binding of F protein to a target receptor on the host cell. Subsequent conformational changes within F result in the interaction of heptad repeat (HR) 1 with HR2 and formation of a 6-helix bundle structure [22-24]. This process brings the viral and host cell membranes in close proximity resulting in fusion pore formation, lipid mixing, and fusion of the two membranes. The precise number of F trimers and identity of target host surface proteins or molecules required to mediate fusion are currently unknown.
Although initially isolated from a chimpanzee, humans are the primary natural host for HRSV. HRSV will only infect the apical surface of human ciliated lung epithelial cells, and only fully differentiated human bronchial epithelial cells are permissive for HRSV growth [25]. Dissemination of HRSV to other organs is not observed even in immunocompromised individuals. Similarly, disseminated infection with bovine RSV is not observed in infected cattle. In contrast, in vitro infection of multiple human cell types other than those derived from lung [26], cells derived from other animal species, and HRSV infection of several animal species has been reported [27]. This suggests that the F protein interacts with either highly conserved host cell surface molecules or can use multiple mechanisms to mediate fusion. Several previous studies have shown the importance of cell-surface glycosaminoglycans (GAGs) [28-32], in particular iduronic acid, in mediating HRSV infection in vitro [33]; however, GAG independent, F-mediated attachment pathways have been described [13]. In a study comparing the host range of bovine and human respiratory syncytial viruses for cells derived from their respective natural hosts, species specificity mapped to the F2 subunit [10]. These finding allude to the existence of host specific receptor molecules that specifically interact with the F protein to mediate cell fusion.
To better understand the factors governing host range, we developed a HRSV F-based quantitative cell fusion assay and specifically examined the ability of HRSV F protein to mediate fusion with cells derived from a wide range of animal species. As cell permissiveness for virus growth is dependent upon multiple steps, we went on to further characterize the permissiveness of these cells for viral mRNA transcription by using a recombinant HRSV engineered to express GFP [33]. The relevance of these findings to the natural course of HRSV disease is discussed.
Methods
Cells and viruses
All cell lines were obtained from the American Type Culture Collection (ATCC) (Manassas, VA) and were grown at 37°C in a humidified atmosphere of 5% CO2 with the exception of XLK-WG (grown at 32°C). BHK-21, E. Derm, HeLa, HEp-2, LLC-PK1, MDBK, MDCK, Mv1Lu, RK-13, Tb1Lu, Vero and A549 cells (ATCC CCL-10, CCL-57, CCL-2, CCL-23, CL-101, CCL-22, CCL-34, CCL-64, CCL-37, CCL-88, CCL-81, and CCL-185 respectively) were maintained in modified Eagle media (MEM) with 2 mM L-glutamine and Earle's balanced salt solution (BSS) adjusted to contain 1.5 g/L sodium bicarbonate, 0.1 mM non-essential amino acids, 1.0 mM sodium pyruvate and 10% heat-inactivated, gamma-irradiated fetal bovine serum (FBS) (HyClone Laboratories, Salt Lake City, Utah). AK-D cells (CCL-150) were maintained in Ham's F-12K media containing 10% FBS. NCI-H292 (CRL-1848), MT-4 and XLK-WG (CRL-2527) cells were maintained in RPMI 1640 medium with 2 mM L-glutamine adjusted to contain 1.5 g/L sodium bicarbonate, 4.5 g/L glucose, 10 mM HEPES, 1.0 mM sodium pyruvate and 10% FBS. NIH/3T3, QT6 and 293T cells (CRL-1658, CRL-1708, and CRL-1573) were maintained in Dulbecco's modified Eagle media (DMEM) with 4 mM L-glutamine adjusted to contain 1.5 g/L sodium bicarbonate, 4.5 g/L glucose and 10% FBS. Cell lines were maintained at sub-confluence and used for only up to 15 passages after receiving initial stocks from the ATCC. Cell lines were tested and confirmed to be free of mycoplasma contamination. Human RSV (subgroup A, Long strain, ATCC VR-26) was obtained from the ATCC. Virus stocks were prepared by infecting HEp-2 cells with RSV at a multiplicity of infection (MOI) of 0.01 plaque-forming units (PFU) per cell. When cytopathic effect (CPE) was evident (~6 days post infection), the culture supernatant was collected and pooled with the supernatant from the infected cell pellet, which had been subjected to one freeze-thaw cycle. The pooled supernatants were maintained on ice, adjusted to 10% sucrose, flash frozen in liquid nitrogen and stored in liquid nitrogen. RSV titers were determined by plaque assay on HEp-2 cells. Serial dilutions of virus stock were added to monolayers of HEp-2 cells at 80% confluence and allowed to adsorb for 2 hours at 37°C. The virus inoculum was then removed, and cells were overlayed with media containing 0.5% methylcellulose. After plaques became apparent (5–6 days after infection), cell monolayers were fixed and stained with 0.5% crystal violet in 70% methanol, and plaques were counted. HRSV stock titers were typically >106 PFU/ml and remained stable for 6 months without loss of titer. A recombinant RSV rgRSV(224) engineered to express GFP has been previously described [33]. Stocks of rgRSV(224) were prepared as described above. Cell lines were infected with rgRSV(224) at a MOI of 0.1 and infection was visualized by fluorescent microscopy by monitoring fluorescence at 488 nm at 20, 48, and 120 hours post infection.
Plasmids
A DNA fragment encoding HRSV F protein derived from a known infectious cDNA sequence for subgroup A, A2 strain, [34] was synthesized with optimal codon usage for expression in mammalian cells and all potential polyadenylation sites (AATAAA) and splice donor sites (AGGT) removed essentially as described [15]. A similar construct was designed and synthesized for the B subgroup F protein (18537 strain, based upon GenBank accession number D00344). Sequence data is available from the authors upon request. Restriction sites for XbaI and BamHI were added to the 5' and 3' ends of the fragments respectively. The codon optimized HRSV-F DNA fragments (A2 and 18537 strains) were then cloned into the XbaI and BamHI sites of pcDNA 3.1 (Invitrogen, Inc., Carlsbad, CA) to generate pHRSVFOptA2 and pHRSVFOpt18537. The QuikChange® Site-Directed Mutagenesis kit (Stratagene®, La Jolla, CA) was used to change leucine 138 in the fusion peptide region of the F protein to an arginine (pL138R) in pHRSVFOptA2. Plasmids pBD-NFκB encoding the activation domain of NFκB fused to the GAL4 DNA binding domain under the control of the human cytomegalovirus (HCMV) promoter and pFR-Luc containing the luciferase reporter gene under the control of a minimal promoter containing five GAL4 DNA binding sites were obtained from Stratagene®. pGL3-control vector encodes a modified firefly luciferase under the control of the SV40 early promoter (Promega, Inc.). Plasmid pVPack-VSV-G encodes the G protein of vesicular stomatitis virus (Stratagene®, La Jolla, CA).
Transfections
Cells were transiently transfected using FuGENE 6 reagent (Roche Applied Science, IN) according to the manufacturer's recommendations. Briefly, 7.5 × 105 cells were plated in 6-well plates and grown overnight to ~90% confluence. Two micrograms of plasmid DNA was complexed with 6 μl of FuGENE 6 reagent for 30 minutes at room temperature in 100 μl of serum-free medium. The complex was then added to the cells and incubated at 37°C for 20–24 hours.
Metabolic labeling and immunoprecipitation
293T cells were plated the day before transfection in 6-well plates at a density of 0.75 × 106 cells/well in DMEM supplemented with 10% FBS. Cells in 6-well plates were transfected with a total of 2 μgs of plasmid DNA as described above. At 20 hours post-transfection, cells were starved by incubation in DMEM without L-methionine and L-cysteine containing 5% dialyzed FBS for 45 minutes. Cells were then labeled by incubation in DMEM without L-methionine and L-cysteine containing 5% dialyzed FBS and Redivue Pro-mix in vitro cell labeling mix containing (100 μCi/ml, 1.5 mls./well) [35S]-methionine and [35S]-cysteine (Amersham Biosciences, Piscataway, New Jersey) for 4 hrs. Media was removed, and cells were harvested and washed in 1 ml 1X phosphate-buffered saline (PBS) and then lysed with 0.5 mls. of lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% sodium deoxycholate, 1% IGEPAL (Sigma, St. Louis, MO) and Complete ™ protease inhibitor cocktail with EDTA (Roche Biochemicals, Indianapolis, IN). Lysates were spun for 30 minutes at 4°C to remove insoluble material and immunoprecipitated by incubation with a saturating amount (as determined by prior titration) of a cocktail containing 1.5 μgs of anti-F mAbs and protein-A agarose beads (Invitrogen, Inc., Carlsbad CA) overnight at 4?C. Immunoprecipitated complexes were washed three times in lysis buffer and suspended in 20 μl of 4X LDS NuPage loading buffer with reducing agent and resolved by electrophoresis through SDS-containing polyacrylamide gels (SDS-PAGE) under reducing conditions on a NuPage 4–12% Bis-Tris polyacrylamide gel (Invitrogen, Inc., Carlsbad CA). Gels were dried under vacuum for one hour at 80°C followed by autoradiography.
Flow cytometry
To confirm cell surface expression of HRSV F, 293T cells were transfected in 6-well plates as described above for metabolic labeling. Cells were stained with palivizumab (Synagis ®, IgG1κ) at 1 μg/ml in conjunction with an anti-human IgG-Alexa-Fluor-488 conjugated secondary (Molecular Probes, Eugene, OR) at 1 μg/ml in 1X PBS with 2% FBS for analysis with the FACSCalibur (BD Bisociences, CA) by setting a live cell gate in the FSC/SSC plot and determining the mean fluorescence intensity in the FL1 channel. Data analysis was performed with Cell Quest and FloJo Analysis Software.
Cell fusion assays
293T cells were co-transfected with pHRSVFOptA2 and pBD-NFκB (effectors cells), and the panel of cell lines from a variety of different species were transfected using FuGene-6 (Roche Biochemicals, Inc.) with the pFR-Luc reporter plasmid (reporter cells) using conditions described above. Alternatively, 293T cells were co-transfected with pHRSVFOptA2 and pFR-Luc, and the panel of cell lines from a variety of different species were transfected with the pBD-NFκB reporter plasmid. A plasmid encoding the vesicular stomatitis virus (VSV) G protein linked to the HCMV promoter (pVPack-VSV-G) was used as a positive viral fusion protein control. At 24 hours post transfection, unless otherwise specified, 3 × 104 of the transfected 293T (effector cells) were mixed with an equal amount of the various other transfected cells (target cells) in a 96-well plate and incubated an additional 24 hours prior to measurement of luciferase activity using the Steady Glo Luciferase reporter system (Promega, Inc.). The various cell lines were transfected with pGL3-control to determine relative differences in transfection efficiencies and cell-type specific expression of luciferase. To further normalize, a single preparation of effector cells was used per each experiment on the various reporter cell preparations. Cells individually transfected with the F expression plasmids, pBD-NFκB or pFR-Luc individually were used as negative controls.
Results
Expression of RSV F protein
We designed and synthesized a cassette encoding the full-length HRSV-F gene (A2 strain and 18537 strains) in which codon usage was optimized for mammalian expression and all potential splice-donor sites and polyadenylation sites were removed similar to a previous report [15]. Upon transient transfection of 293T cells with this plasmid expressing this optimized HRSV F protein sequence under the control of a human cytomegalovirus immediate early promoter (pHRSVFoptA2), giant multinucleated cells (syncytia) were readily apparent within 24 hours post transfection (Figure 1A). The amount of syncytia qualitatively increased throughout the culture for up to 72 hours, after which extensive cell death and sloughing was observed. This syncytia is phenotypically indistiguishable from that observed following infection of 293T cells with HRSV in tissue culture (data not shown). To confirm appropriate processing of the F protein, 293T cells were transfected with plasmids expressing HRSV F derived from subgroup A (A2 strain) or subgroup B (18537 strain) and metabolically labeled followed by immunoprecipitation of lysates with HRSV F-specific monoclonal antibodies. As a control, 293T cells were infected with HRSV (Long strain). As shown in Figure 1B, immunoprecipitation demonstrates the presence of the unprocessed full length F0 species migrating at approximately 70 kDa, and the processed F1 and F2 fragments of ~50 kDa and 20 kDa, respectively, identical to the F protein immunoprecipitated from HRSV infected 293T cells. The multiple bands observed in the region of 20 kDa likely represent the incompletely processed F2 (F2+), different glycosylated forms of F2, or a combination of both [21]. The band present migrating more rapidly than F1 (~35 kDa) most likely represents a cellular protein as this band was observed in lysates derived from untransfected and control vector transfected cells with varying intensity unrelated to the level of HRSV F expressed (see Figure 2B, lanes 3 and 4). Similar levels of expression were observed for the HRSV F protein from the A and B subgroups (Figure 1B, lane 3 compared to lane 4). Furthermore, the level of F protein expression in the transfected cells was greater at 24 hours post transfection than in HRSV-infected cells at 24 hours post infection (Figure 1B, lane 5). Flow cytometry confirmed abundant cell surface expression of HRSV F protein (Figure 1C).
Figure 1 A) Syncytia formation by RSV-F DNA in transfected cells. 293T cells were mock transfected or transfected with pHRSVFOptA2 and visualized by light microscopy 48 hours post transfection. The arrow indicates giant multinucleated cell formation. B) Processing of RSV F in transfected cells. 293T cells were either mock transfected (lane 1), transfected with pCMV-β-gal (lane 2), transfected with pHRSVFOptA2 (lane 3), pHRSVFOptB18537 (lane 4), or infected with RSV (Long strain, MOI = 1) for 24 hours followed by metabolic labeling for 6 hours with [35S]-cysteine/methionine. Labeled cell lysates were immunoprecipitated with HRSV F specific mAbs, and immunoprecipitates were resolved by SDS-PAGE as described in methods. C) Cell surface expression of RSV F in transfected cells. 293T cells were either mock transfected or transfected with pHRSVFOptA2 for 24 hours followed by flow cytometry using HRSV F specific monoclonal antibodies as described in methods.
Figure 2 A) Dose dependent fusion activity of HRSV F derived. 293T cells were transfected with either pFR-Luc alone (■), pBD-NFκB alone (▲), co-transfected with pFR-Luc and pBD-NFκB (▼), or co-transfected with pHRSVFOptA2 and pBD-NFκB and mixed 24 hours after transfection in various amounts with cells that had been transfected with pFR-Luc alone (◆). Luciferase activity was measured 24 hours post mixing as described in methods and is reported as relative light units. B) Fusion activity of HRSV F derived from subgroups A and B. 293T cells co-transfected with pBD-NFκB and either pHRSVFOptA2, pHRSVFOptB18537, or vector only (NFκB only). Cells were mixed 24 hours later with a separate population of 293T cells transfected with pFR-Luc, and luciferase activity was measured 24 hours post mixing as described in methods. Luciferase activity is reported as relative light units.
HRSV F protein fusion assay
To measure the ability of the HRSV F protein to mediate cell fusion across various cell lines, we developed a quantitative fusion assay. Specifically, 293T cells were co-transfected with plasmids encoding the optimized HRSV F protein and a transcriptional transactivator fusion protein consisting of the GAL4 DNA-binding domain fused to the activation domain of NFκB. These effector cells were mixed with a separate set of reporter cells that were transfected with a reporter plasmid containing the luciferase gene under the control of a GAL4 responsive promoter. HRSV F mediated cell fusion of the two cell populations results in co-localization of the GAL4-NFκB transactivator fusion protein with the GAL4 responsive luciferase reporter plasmid and subsequent transcriptional transactivation of the reporter gene. A dilution series from 50,000 to 100 effector cells were added to a fixed amount of reporter cells (50,000), and luciferase activity was monitored 24 hours later. As a control to determine the maximum signal, cells were co-transfected with the reporter and activator plasmids (pFR-Luc + pBD-NFκB). As shown in figure 2A, luciferase activity was absent in cells transfected with either reporter or activator plasmids alone. Additionally, mixing of cells which had been separately transfected with the reporter or activator plasmids did not produce detectable luciferase activity indicating no spontaneous cell fusion (data not shown). However, mixing an increasing number of cells that had been co-transfected with the GAL4-sensitive reporter plasmid and HRSV F expressing plasmid with those that had been transfected with the GAL-NFκB activator plasmid resulted in a dose-dependent increase in luciferase activity (Figure 2A), indicating fusion of the two cell populations. The maximum signal observed from mixing the effector and reporter populations was similar to the signal obtained when the activator and reporter plasmids were co-transfected into the same cells.
To determine if this property was restricted to the F protein derived from a single strain or subgroup, 293T cells co-transfected with a plasmid encoding the HRSV F protein derived from either subgroup A (A2 strain) or B (18537 strain) together with a plasmid encoding the GAL4-NFκB transactivator fusion protein (effector cells) were then mixed 24 hours later with an equal amount of a separate population of 293T cells (reporter cells) which had been transfected with the GAL4 responsive reporter plasmid. As shown in Figure 2B, the F protein of either HRSV subgroup A and B mediated cell-cell fusion as measured by the increased luciferase activity relative to the negative control (GAL4-NFκB transactivator fusion protein only). The fusion activity of the F protein derived from the A2 strain was approximately 2-fold higher than that observed with the 18537 strain despite similar expression levels. Whether this finding reflects differences in the pathogenicity between these two isolates is unknown, although a recent study suggests similar pathogenicity for both subgroups [35]. To further confirm that the observed fusion activity was inherent to the HRSV F protein, a point mutation (L138R) was generated in the fusion peptide consensus sequence within the fusion peptide region. Mutation of leucine residue 138 to arginine reduced fusion activity to 10% relative to wild-type (Figure 3A). Despite the fact that this mutant appeared to produce somewhat lower levels of fully processed F protein (Figure 3B, lane 2) for unknown reasons, this mutant was expressed at high levels on the cell surface (Figure 3C) indicating that the cell fusion observed was attributable to the HRSV F protein.
Figure 3 A) Comparison of fusion activity of wild-type and a fusion peptide mutant of HRSV F. 293T cells co-transfected with pBD-NFκB and either pHRSVFOptA2 or pL138R were mixed 24 hours later with a separate population of 293T cells that had been transfected with pFR-Luc, and luciferase activity was measured 24 hours post mixing as described in methods. Luciferase activity is reported as relative light units. B) Processing of the wild-type and L138R mutant of HRSV F was determined by metabolic labeling 293T cells transfected with either pHRSVFOptA2 (lane 1), pL138R (lane 2), pCMV-β-gal (lane 3), or mock transfected (lane 4) for 6 hours with [35S]-cysteine/methionine followed by immunoprecipitation of lysates with HRSV F specific mAbs, and analysis of immunoprecipitates by SDS-PAGE as described in methods. C) Cell surface expression of the wild-type and L138R mutant F proteins in 293T cells transfected with either pHRSVFOptA2 or pL138R was compared by flow cytometry as described in methods.
Host range of HRSV-mediated fusion
To determine the host range of HRSV F mediated fusion using the quantitative fusion assay. 293T effector cells were prepared as described above. As we previously demonstrated proper protein processing, abundant cell surface expression of HRSV F protein and cell fusion activity using 293T cells, we selected these as our effector cells. These effector cells were mixed with reporter cells derived from a diverse range of species (Table 1) that were transfected with the GAL4-responsive reporter plasmid. To account for any differences in relative transfection efficiencies and expression of the luciferase reporter among the various target cells lines, the target cell lines were transfected with a plasmid containing the luciferase gene under the control of the SV40 early promoter (pGL3-control), and relative luciferase activity was measured. To account for potential differences in host cell transcription factors that mediate activation of the reporter plasmid, the assay was flipped and 293T cells were co-transfected with the HRSV F expression plasmid and the GAL4 responsive reporter plasmid, and the cells from the various species were transfected with the GAL4-NFκB transactivator fusion protein plasmid. For further comparison, we used the VSV G protein, which is known to mediate entry into cells derived from a wide range of species.
Table 1 Species and tissue origin of cell lines used in this study are listed.
Cell line Species, tissue
XLK-WG Xenopus laevis (S. African clawed frog), kidney
QT6 Coturnix coturnix japonica (Japanese quail), fibrosarcoma
Tb1Lu Tadarida brasiliensis (free-tailed bat), lung
NIH/3T3 Mus musculus (mouse), fibroblast
BHK-21 Mesocricetus auratus (Syrian golden hamster), kidney
RK-13 Oryctolagus cuniculus (rabbit), kidney
LLC-PK1 Sus scrofa (pig), kidney
Mv1Lu Musteal vison (mink), lung
AK-D Felis catus (domestic cat), fetal epithelial
MDCK Canis familiaris (domestic dog), kidney
MDBK Bos taurus (cow), kidney
E. Derm Equus caballus (horse), dermal
Vero Cercopithecus aethiops (African green monkey), kidney
HEp-2 Homo sapiens (human), laryngeal carcinoma
HeLa Homo sapiens (human), cervical carcinoma
MT-4 Homo sapiens (human), T-cell
293T Homo sapiens (human), kidney
NCI-H292 Homo sapiens (human), epidermoid pulmonary carcinoma
A549 Homo sapiens (human), lung
As shown in figure 4, despite a limited host range in nature, HRSV F was able to mediate fusion to various degrees with cells derived from all species examined. This fusion activity was within 5-fold of the fusion activity mediated by the VSV G protein in the cell types tested here. Generally, there was little qualitative difference between results obtained when either the reporter plasmid or the activator plasmid were co-transfected with the F expression plasmid (compare figures 4A and 4B with figures 4C and 4D). As expected, the relative transfection efficiencies of the various cell lines as measured by the luciferase activity from the plasmid pGL3-control varied; however, there was no direct correlation between transfection efficiencies and fusion activity. For example, cell lines such as BHK-21 and LLC-PK1 cells which transfected well, had lower relative levels of fusion. In contrast, cell lines such as MT-4, MDCK and XLK-WG which had low transfection efficiency, had appreciable levels of HRSV F mediated fusion. These findings support the hypothesis that HRSV F protein interacts with evolutionarily conserved host cell surface molecules or can use multiple mechanisms to enter cells.
Figure 4 Fusion activity of HRSV F with cell lines derived from various species. Cell lines derived from various species (target cells) were either transfected with pFR-Luc and mixed 24 hours later with 293T cells that had been co-transfected for 24 hours with pHRSVFOptA2 or pVPack-VSV-G and pBD-NFκB (Figs. 4A and 4B), or the target cells were transfected with pBD-NFκB and mixed 24 hours later with 293T cells that had been co-transfected for 24 hours with pHRSVFOptA2 or pVPack-VSV-G together with pFR-Luc (Figs. 4C and 4D). Cell lines derived from various species were transfected with either pFR-Luc or pBD-NFκB only as negative controls. Luciferase activity was measured 24 hours post mixing of the cell populations as described in methods and is reported as relative light units.
Infections using recombinant HRSV expressing GFP
The results obtained from the fusion assays indicated that HRSV F is able to mediate fusion with cells from multiple diverse species, suggesting that virus entry is not the primary determinant of host range. To examine whether viral mRNA transcription had occurred, the various cell lines were infected with a recombinant HRSV (rgRSV224) expressing GFP [33] and fluorescence scored at 20, 48, and 120 hours post infection. As expected, rgRSV(224) infection of human (HEp-2, HeLa, A549) and animal (Vero, Mv1Lu, MDBK) [36-39] cell lines commonly used to propagate HRSV resulted in a time dependent increase in the number of cells expressing GFP (≥50% by day 5) as seen by fluorescent microscopy indicating spread of infection throughout the culture (Figure 5). Infection of other human cell lines such as NCI-H292 [40], and 293T also resulted in a time dependent increase in the number of cells expressing GFP. Infection of hamster BHK-21 cells also resulted in a time dependent increase in the number of GFP positive cells, although the appearance of a large number of bright GFP positive cells seemed delayed. Interestingly, hamsters are considered to be a semi-permissive host for HRSV [27,41] and produce lung titers similar to those achieved in mice. Whether this reflects a tissue-specific phenomenon (kidney versus lung) remains to be determined. Infection of cell lines (Tb1Lu, AK-D, E. Derm, NIH/3T3, LLC-PK1, and XLG-WG) derived from other species (bat, cat, horse, mouse, and frog respectively) produced few or occasional GFP expressing cells over the course of the five-day infection. The number of positive cells did not increase over time, and in some cases (AK-D cells) appeared to decrease. Aside from mice, infection in vivo of these other species by HRSV has not been described. This finding also supports the finding that high titers of virus (>105 PFU) are typically required to initiate infection in mice after intranasal inoculation, and that relatively few cells become viral antigen positive.
Figure 5 Infection of various cell lines by rg224(RSV). Cell lines derived from various species were infected with rgRSV(224) at an MOI = 0.1 and GFP-expressing cells were visualized at 20, 48, and 120 hours post infection by fluorescent microscopy by monitoring fluorescence at 488 nm.
Discussion
We have developed a quantitative reporter gene based cell-cell fusion assay for HRSV F. Prior assays have been based upon visual counting of plaques or syncytia after staining of infected monolayers, or infection another virus such as vaccinia, to provide HRSV F protein, which could potentially complicate interpretation. The assay described herein is a means of quantifying the fusion activity of the HRSV F protein. This assay has multiple applications. For example, this assay can be used as a means of studying the structure-function of the HRSV F protein, or for evaluating the activity of mutations in the F protein without the need to select for antibody or compound escape mutants or generate point mutations in a reverse genetics system. We propose that this assay also has utility in the identification and characterization of inhibitors of HRSV entry for the development of specific agents to prevent and treat HRSV infections. We have used this assay as a means of exploring the host-range of HRSV and have shown that the HRSV F protein is able to mediate fusion with cells derived from a wide range of vertebrate species.
Cell lines known to be permissive for HRSV growth such as HEp-2, HeLa, A549, Vero, MDBK, and Mv1Lu were highly competent for F protein fusion as expected. Somewhat surprisingly, a wide variety of cells derived from species not known to be normally infected by HRSV were also capable of undergoing HRSV F protein mediated fusion. Most surprising were the results obtained using the XLK-WG cells which are derived from the amphibian Xenopus laevis. Although this finding implies that HRSV virion is able to enter a wide range of cells, the results of the infection studies using the GFP-expressing RSV indicate that viral mRNA transcription seems limited in cell lines derived from certain species. Taken together these results suggest that events post-viral entry are the primary determinants that mediate the host range of HRSV. During natural infection of humans, viral replication is restricted to epithelial cells of the upper and lower respiratory tract. Although limited HRSV replication within human alveolar macrophages and detection of HRSV sequences in peripheral blood monocytes (PBMCs) has also been reported [42,43], dissemination of HRSV to other organs is not observed even in immunocompromised individuals. Similarly, disseminated infection with bovine RSV is not observed in infected cattle [44]. Given the ability of the HRSV F protein to mediate fusion with cells derived from a diverse range of vertebrate species, the implication is that HRSV may not be able to access these sites or undergoes non-productive infection in many cell types other than epithelial cells of the respiratory tract. Although the overall biological significance of such an abortive infection is unclear, biological effects of the individual HRSV proteins have been reported.
HRSV F protein has also been shown to be a ligand for TLR4, and HRSV infection persists longer in TLR4-/- deficient mice [45,46]. HRSV F protein also binds surfactant proteins A and D (SP-A and SP-D) [47,48], although the implications of these findings in human infection are unclear. G protein has been shown to modulate multiple immune related activities. Soluble G suppresses some PBMC and lung CD8+ T-cell effector and peripheral memory responses [49], induces chemotaxis, eosinophilia, and both soluble and membrane forms of G bind the fractalkine receptor, CX3CRI [50]. Additionally, G has a domain with similarity to the TNF-α receptor (p55), although it has not been directly shown to be a TNFR antagonist. Additionally, G has been shown to modify CC and CXC chemokine mRNA expression [50], and suppress lymphoproliferative responses to antigens in PBMCs [51].
It is tempting to speculate that entry of HRSV into cell types other than those permissive for complete virus growth may be a strategy by which the virus is able to modulate immune responses while avoiding the induction of antiviral responses such as the interferon (IFN) pathway by production of double-stranded RNA replication intermediates in these cells. Limited viral mRNA transcription in the absence of virus RNA replication would result in expression of NS1 and NS2 which have been shown to block the IFN response [52] possibly preventing these unproductively infected cells from responding to external cytokines such as IFNs. Such a strategy may help explain why despite little antigenic drift in the F protein, infection by HRSV infection only confers partial protection, with reinfections occurring throughout life [53-55]. As the fusion proteins of other members of the Paramyxoviridae family, such as Hendra virus [56], are also able to mediate fusion with a wide variety of cells derived from multiple species, it is possible that such a strategy is shared by other members of this virus family.
Competing interests
The author(s) declare that they are all employees of Centocor, Inc. which provided supported for this work.
Authors' contributions
PB and CL contributed equally to this work. PB and ND performed the fusion assays, immunoprecipitations, and flow cytometry. CL generated reagents and developed the fusion assay. LG conducted site-directed mutagenesis of the HRSV F protein. AD and RS participated in the design of the experiments, oversight of the conduct of the experiments, and in the interpretation of the results.
Table 2 Infection of various cell lines with GFP-expressing HRSV.
Cell line 20 hrs 48 hrs 120 hrs
Vero +++ ++++ ++++
AK-D + ++ -
MDBK +++ ++++ +++
MDCK + + +
Tb1Lu + + +
XLK-WG - + +
E. Derm ++ ++ +
HeLa ++ +++ ++++
NCI-H292 ++ +++ +++
293T ++ ++++ ++++ *
HEp-2 +++ ++++ ++++
Mv1Lu +++ ++++ +++
NIH/3T3 + + +
LLC-PK1 + - -
RK-13 ++ ++ ++
BHK-21 - ++ +++
QT6 + ++ ++++
A549 + +++ ++
MT-4 + ++ ++
- = few isolated weakly positive cells
+ = <1–5% GFP positive cells in culture
++ = 5–30% GFP positive cells in culture
+++ = 30–60% GFP positive cells in culture
++++ = >60% GFP positive cells in culture
* wide spread cell death
Acknowledgements
Recombinant HRSV expressing green fluorescent protein rgRSV(224) was generously provided by Dr. Peter Collins (NIAID, NIH). We thank Jose Melero, Geraldine Taylor, and Paul Bates for helpful discussion and comments, and Lamine Mbow, Lani San Mateo, and William Glass for critical review of this manuscript.
==== Refs
Shay DKHRCNRDLLLSJWLJA Bronchiolitis-associated hospitalizations among U.S. children, 1980-1996 JAMA 1999 282 1440 1446 10535434 10.1001/jama.282.15.1440
Shay DKHRCRGECMJALJ Bronchiolitis-associated mortality and estimates of respiratory syncytial virus-associated deaths among U.S. children, 1979-1997 J Infect Dis 2001 183 16 22 11076709 10.1086/317655
Han LLJPALJA Respiratory syncytial virus pneumonia among the elderly: an assessment of disease burden J Infect Dis 1999 179 25 30 9841818 10.1086/314567
Ison MGHFG Viral infections in immunocompromised patients:what's new with respiratory viruses? Curr Opin Infect Dis 2002 15 355 367 12130931
Collins PLCRMMBR D.M. Knipe HPM Respiratory syncytial virus Fields Virology 2001 1 Philadelphia, Lippincott, Williams, and Wilkins 1443 1485
Karron RADABAFGSSWJEAMLCMDOHVBRSUBRMMSS Respiratory syncytial virus (RSV) SH and G proteins are not essential for viral replication in vitro: clinical evaluation and molecular characterization of a cold-passaged, attenuated RSV subgroup B mutant Proc Natl Acad Sci USA 1997 94 13961 13966 9391135 10.1073/pnas.94.25.13961
Teng MNSSWPLC Contribution of the respiratory syncytial virus G glycoprotein and its secreted and membane-bound forms to virus replication in vitro and in vivo Virology 2001 289 283 296 11689051 10.1006/viro.2001.1138
Bukreyev AWSSMBRCPL Recombinant respiratory syncytial virus from which the entire SH gene has been deleted grows efficiently in cell culture and exhibits site-specific attenuation in the respiratory tract of the mouse. J Virol 1997 71 8973 8982 9371553
Jin HZHCXTRMMNN Recombinant respiratory syncytial viruses with deletions in the NS1, NS2, SH, and M2-2 genes are attenuated in vitro and in vivo Virology 2000 273 210 218 10891423 10.1006/viro.2000.0393
Schlender JGZGHKKC Respiratory syncytial virus (RSV) fusion protein subunit F2, not attachment protein G, determines the specificity of RSV infection J Virol 2003 77 4609 4616 12663767 10.1128/JVI.77.8.4609-4616.2003
Techaarpornkul SBNPME Functional analysis of recombinant respiratory syncytial virus deletion mutants lacking the small hydrophobic and/or attachment glycoprotein J Virol 2001 75. 6825 6834 11435561 10.1128/JVI.75.15.6825-6834.2001
Whitehead SSABMNTCYFMSCWREPLCBRM Recombinant respiratory syncytial virus bearing a deletion of either the NS2 or SH gene is attenuated in chimpanzees J Virol 1999 73 3438 3442 10074199
Techaarpornkul SPLCMEP Respiratory syncytial virus with the fusion protein as its only viral glycoprotein is less dependent on cellular glycosaminoglycans for attachment than complete virus Virology 2002 294 296 304 12009871 10.1006/viro.2001.1340
Teng MNPLC Identification of the respiratory syncytial virus proteins required for formation and passage of helper-dependent infectious particles J Virol 1998 72 5707 5716 9621029
Morton CJRCLJLBLMLALAJMMKBMWPJRMSJSSPTPRY Structural characterization of respiratory syncytial virus fusion inhibitor escape mutants: homology model of the F protein and a syncytium formation assay Virology 2003 311 275 288 12842618 10.1016/S0042-6822(03)00115-6
Heminway BRYYYTKGPEGJMBMSG Analysis of respiratory syncytial virus F, G, and SH proteins in cell fusion Virology 1994 200 801 805 8178462 10.1006/viro.1994.1245
Bolt G POLBHH Cleavage of the respiratory syncytial virus fusion protein is required for its surface expression: role of furin Virus Res 2000 68 25 33 10930660 10.1016/S0168-1702(00)00149-0
Collins PLMG Post-translational processing and oligomerization of the fusion glycoprotein of human respiratory syncytial virus J Gen Virol 1991 72 3095 3101 1765771
Gonzalez-Reyes LMBRABGBLCJAJPAJJSDCWJAM Cleavage of the human respiratory syncytial virus fusion protein at two distinct sites is required for activation of membrane fusion Proc Natl Acad Sci 2001 98 9859 9864 11493675 10.1073/pnas.151098198
Sugrue RLCBGBJAHWMLR Furin cleavage of the respiratory syncytial virus fusion protein is not a requirement for its transport to the surface of virus-infected cells J Gen Virol 2001 82 1375 1386 11369882
Rixon HWMLCBGBRJS Multiple glycosylated forms of the respiratory syncytial virus fusion protein are expressed in virus-infected cells J Gen Virol 2002 83 61 66 11752701
Matthews JMTFYSPTJPM The core of the repiratory syncytial virus fusion protein is a trimeric coiled coil J Virol 2000 74 5911 5920 10846072 10.1128/JVI.74.13.5911-5920.2000
Zhao XMSVNMPSK Structural characterization of the human respiratory syncytial virus fusion protein core Proc Natl Acad Sci USA 2000 97 14172 14177 11106388 10.1073/pnas.260499197
Lawless-Delmedico MKPSRSNCMJBAJMWRJGKCLTJMDML Heptad-repeat regions of respiratory syncytial virus F1 protein form a six-membered coiled-coil complex Biochemistry 2000 39 11684 11695 10995236 10.1021/bi000471y
Zhang LMEPRCBPLCRJP Respiratory syncytial virus infection of human airway epithelial cells is polarized, specific to ciliated cells, and without obvious cytopathology J Virol 2002 76 5654 5666 11991994 10.1128/JVI.76.11.5654-5666.2002
Sarmiento RERTBG Characteristics of a respiratory syncytial virus persistently infected macrophage-like culture Virus Res 2002 84 45 58 11900838 10.1016/S0168-1702(01)00420-8
Byrd LGPGA Animal models of respiratory syncytial virus infection Clin Infect Dis 1997 25 1363 1368 9431379
Bourgeois CBJBLKGCPP Heparin-like structures on respiratory syncytial virus are involved in its infectivity in vitro J Virol 1998 72 7221 7227 9696816
Hallak LKDSPLCMEP Glycosaminoglycan sulfation requirements for respiratory syncytial virus infection J Virol 2000 74 10508 10513 11044095 10.1128/JVI.74.22.10508-10513.2000
Feldman SASAJAB The fusion glycoprotein of human respiratory syncytial virus facilitates virus attachment and infectivity via an interaction with cellular heparan sulfate J Virol 2000 74 6442 6447 10864656 10.1128/JVI.74.14.6442-6447.2000
Krusat THJS Heparin-dependent attachment of respiratory syncytial virus (RSV) to host cells Arch Virol 1997 12 1247 1254 9229012 10.1007/s007050050156
Martinez IJAM Binding of human respiratory syncytial virus to cells:implication of sulfated cell surface proteoglycans J Gen Virol 2000 81 2715 2722 11038384
Hallak LKPLCWKMEP Iduronic acid-containing glycosaminoglycans on target cells are required for efficient respiratory syncytial virus infection Virology 2000 271 264 275 10860881 10.1006/viro.2000.0293
Collins PLHMGCEGHCRMMBR Production of infectious human respiratory syncytial virus from cloned cDNA confirms an essential role for the transcription elongation factor from the 5' proximal open reading frame of the M2 mRNA in gene expression and provides a capability for vaccine development Proc Natl Acad Sci USA 1995 92 11563 11567 8524804
Devincenzo JP Natural infection of infants with respiratory syncytial virus subgroups A and B: a study of frequency, disease severity, and viral load Pediatr Res 2004 56 914 917 15470202 10.1203/01.PDR.0000145255.86117.6A
Barenfanger JCDTMTTJOBKG R-Mix cells are faster, at least as sensitive and marginally more costly than conventional cell lines for detection of respiratory viruses. J Clin Virol 2001 22 101 110 11418357 10.1016/S1386-6532(01)00171-8
Bossert BKKC Respiratory syncytial virus (RSV) nonstructural (NS) proteins as host range determinants: a chimeric bovine RSV with NS genes from human RSV is attentuated in interferon-competent bovine cells. J Virol 2002 76 4287 4293 11932394 10.1128/JVI.76.9.4287-4293.2002
Huang YTBMT Mink lung cells and mixed mink lung and A549 cells for rapid detection of influenza virus and other respiratory viruses. J Clin Microbiol 2000 38 422 423 10618131
Parry JEPVSCRP Pneumoviruses: the cell surface of lytically and persistently infected cells. J Gen Virol 1979 44 479 491 118236
Hierholzer JCCEBGGBJAMECT Sensitivity of NCI-H292 human lung mucoepidermoid cells for respiratory and other human viruses. J Clin Microbiol 1993 31 1504 1510 8314992
Wright PFWWGRMC Temperature-sensitive mutants of respiratory syncytial virus: in vivo studies in hamsters. J Infect Dis 1970 122 501 512 5489074
Cirino NMPJRVATHRNAMRTPGIA Restricted replication of respiratory syncytial virus in human alveolar macrophages J Gen Virol 1993 74 1527 1537 8345347
Yui IAHYSTTTN Detection of human respiratory syncytial virus sequences in peripheral blood mononuclear cells J Med Virol 2003 70 481 489 12767015 10.1002/jmv.10421
Viuff BAUCTSA Sites of replication of bovine respiratory syncytial virus in naturally infected calves as determined by in situ hybridization Veterinary Pathology 1996 33 383 390 8817835
Haynes LMMDDKJEAFRWALJTRA Involvement of toll-like receptor 4 in innate immunity to respiratory syncytial virus J Virol 2001 75 10730 10737 11602714 10.1128/JVI.75.22.10730-10737.2001
Kurt-Jones EAPLKLHLMJLPTRAWEEFMWGDTALJFRW Pattern recognition receptors TLR4 and CD14 mediate response to respiratory syncytial virus Nat Immunol 2000 1 398 401 11062499 10.1038/80833
Hickling TPHBKWDGSLMRBSRM A recombinant trimeric surfactant protein D carbohydrate recognition domain inhibits respiratory syncytial virus infection in vitro and in vivo Eur J Immunol 1999 29 3478 3784 10556802 10.1002/(SICI)1521-4141(199911)29:11<3478::AID-IMMU3478>3.3.CO;2-N
Ghildyal RCHAVJMDRVEMAJM Surfactant protein A binds to the fusion glycoprotein of respiratory syncytial virus and neutralizes virion infectivity J Infect Dis 1999 180 2009 2013 10558959 10.1086/315134
Chung WBTJ Respiratory syncytial virus infection suppresses lung CD8+ T-cell effector activity and peripheral CD8+ T-cell memory in the respiratory tract. Nat Med 2002 8 54 60 11786907 10.1038/nm0102-54
Tripp RAJLALJ Respiratory syncytial virus G and/or SH glycoproteins modify CC and CXC chemokine mRNA expression in the BALB/c mouse J Virol 2000 74 6227 6229 10846112 10.1128/JVI.74.13.6227-6229.2000
Ray RHDFMKBRLLMBRB Immunoregulatory role of secreted glycoprotein G from respiratory syncytial virus Virus Res 2001 75 147 154 11325469 10.1016/S0168-1702(01)00237-4
Spann KMKCTBCRLRPLC Suppression of the induction of alpha, beta, and lambda interferons by the NS1 and NS2 proteins of human respiratory syncytial virus in human epithelial cells and macrophages J Virol 2004 78 4363 4369 15047850 10.1128/JVI.78.8.4363-4369.2004
Dowell SFALJGHEJEDDPJFFTMJMBJBRF Respiratory syncytial virus is an important cause of community-acquired lower respiratory infection among hospitalized adults. J Infect Dis 1996 174 456 462 8769600
Falsey AREEW Respiratory syncytial virus infection in adults Clin Microbiol Rev 2000 13 371 10885982 10.1128/CMR.13.3.371-384.2000
Glezen WPTLHFALKJA Risk of primary infection and reinfection with respiratory syncytial virus Am J Dis Child 1986 140 543 546 3706232
Bossart KNWLFFMNCKBLSKEBTBCC Membrane Fusion tropism and heterotypic functional activities of the Nipah virus and Hendra virus envelope glycoproteins J Virol 2002 76 11186 11198 12388678 10.1128/JVI.76.22.11186-11198.2002
|
16107218
|
PMC1190220
|
CC BY
|
2021-01-04 16:39:00
|
no
|
Virol J. 2005 Aug 17; 2:66
|
latin-1
|
Virol J
| 2,005 |
10.1186/1743-422X-2-66
|
oa_comm
|
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-421598751210.1186/1477-7819-3-42Case ReportCase report: late perianal mucinous adenocarcinoma after Crohn's disease proctectomy: an oncological rarity Keese Michael [email protected] Walter [email protected] Dietmar [email protected] Rainer [email protected] Andreas [email protected] Pablo [email protected] Department of Surgery, University Hospital of Mannheim, 68135 Mannheim, Germany2 Department of Pathology, University Hospital of Mannheim, 68135 Mannheim, Germany3 Department of Radiology, University Hospital of Mannheim, 68135 Mannheim, Germany4 Department of Medicine, University Hospital of Mannheim, 68135 Mannheim, Germany2005 29 6 2005 3 42 42 23 2 2005 29 6 2005 Copyright © 2005 Keese et al; licensee BioMed Central Ltd.2005Keese et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
As in ulcerative colitis, there is an increased incidence of colorectal carcinoma in Crohn's disease. While carcinoma formation originating from ano-rectal fistulas is generally considered as a rare event there are different publications reporting on mucinous adenocarcinoma formation in association with a neovagina and rectovaginal fistulas. To the best of our knowledge this is the first description of a perianal mucinous adenocarcinoma arising in a patient after Crohn's disease proctocolectomy.
Case presentation
We report the case of a 50-year old female with a mucinous adenocarcinoma forming in the perineum eleven years after proctocolectomy for Crohn's disease. The patient was readmitted with perineal pain, leucocytosis and a perineal mass highly suspicious of abscess formation in the MRI-Scan. Histological examination revealed a mucinous adenocarcinoma. Exenteration including vagina, uterus and ovaries together with the coccygeal-bone was performed.
Conclusion
Mucinous adenocarcinoma formation is a rare complication of Crohn's disease and so far unreported after proctocolectomy.
==== Body
Background
As in ulcerative colitis, there is an increased incidence of colorectal carcinoma in Crohn's disease [1]. An increased risk of cancer in patients with Crohn's disease has been shown to be related to an early onset and a prolonged duration of the inflammatory bowel disease [2]. Furthermore, several cases have been reported in which carcinoma formation originated from ano-rectal fistulas which are commonly associated with Crohn's disease [3]. While carcinoma formation originating from anorectal fistulas is generally considered as a rare event, these tumors are most commonly either mucinous carcinomas or squamous cell carcinomas [4]. The causative relationship between anorectal fistulas and cancer is not known. Cancer development in these cases remains a diagnostic challenge especially if carcinomas arise in the midst of abscess formation [5]. We report late formation of a mucinous adenocarcinoma in a patient with Crohn's disease who presented severe perineal fistulous lesions after proctocolectomy.
Case presentation
In October 2004 a 50-year old woman was admitted presenting with gluteal and perineal fistula formation after proctocolectomy eleven years earlier. The clinical examination showed a single fistula opening on the perineum with purulent secretion. A second opening was found on the posterior wall of the vagina.
The MRI-scanning revealed a large formation in the lower pelvis reaching to the sacrum. This mass showed a thickened, contrast-enhancing wall and necrotic, liquid inner parts suggestive of abscess-formation. No involvement of the small bowel or bladder was detected (Figures 1, 2, 3).
Figure 1 A T2-weighted axial MR image of the pelvis showing an irregular presacral fluid collection (arrows) extending from the cervix to the coccyx and to the sciatic foramen on the left side.
Figure 2 T1- weighted fat saturated MR image after i.v. application of contrast medium coronal. The arrow is pointing to a fistulating formation spreading in the gluteal muscles on the left side with typical rim enhancement and central liquid components; the soft tissue mass seen in midline has a centrally necrotizing appearance.
Figure 3 T1- weighted fat saturated MR image after i.v. application of contrast medium sagittal view. Presacrally, the mass characterized by an inhomogeneous contrast medium uptake in the thickened, nodular wall spreads in the subperiosteal space, as indicated.
The patient's history revealed Crohn's disease first diagnosed in 1981. Ten years later (1991), the patient was readmitted because of severe perineal fistula disease. The clinical and radiological examination showed multiple transphincteric fistulas and a single rectovaginal fistula as well as one enterocutaneous fistula originating from the terminal ileum. After medical management (parenteral nutrition and corticoids) she was referred to the department of surgery where an ileocoecal resection with a diversion ileostomy and drainage of the perianal fistulas were performed. Postoperatively the patient developed all signs of an Addison crisis, which could be satisfactorily treated.
The patient was readmitted in 1993 with a new inflammatory episode including perianal fistulating disease. Because of the disease recurrence and due to further impairment of the sphincter mechanism a proctocolectomy with exstirpation of the sphincter ani muscles and a terminal ileostomy was recommended and performed. The histopathological examination showed signs of Crohn's disease with some epitheloid cell granulomas and giant cells as well as microgranulomas in both the large bowel and the rectum. No signs of malignancy and no dysplastic changes were found. The patient developed a paralytic ileus that could be treated conservatively and a perianal wound infection which required secondary wound closure.
Three years later, in 1996, the patient was readmitted because of a persistent perineal secretion. No fistulas and no signs of Crohn's disease could be found in the small bowel. The patient was treated with local wound therapy without any further surgical procedure until readmission in 2004.
Surgical therapy
After informed consent a perineal drainage with entire resection of the clinically inflamed tissue was performed. Intraoperatively the tissue showed a colloidal consistency and fistulation into the coccygeal bone and the vagina. Histological analysis of the tissue detected a mucinous adenocarcinoma occupying the resected tissue and the resection margins. The tumour consisted of moderately atypical glandular cell elements lying in pools of PAS-positive mucin. A distinct fibroblastic stromal reaction could be found in the surrounding mesenchyma.
Upon diagnosis, a second intervention was performed including an en-bloc exenteration of the uterus, vagina and ovaries via laparotomy and resection of the sacrum through a posterior approach. The pelvis was closed using a Vypro® mesh (polypropylene-polygalactine composite) and the perineum was left open for secondary wound healing.
Postoperatively, bladder function was impaired, otherwise the patient recovered well and could be dismissed into out patient oncological care.
Histopathology
Examination of the specimen showed typical cylindrical epithelium of colorectal type lining some residual lumina of fistulas and cysts (Figure 4). The resection margins did not show any residual fistulas. No rests of original bowel mucosa or remaining bowel wall structures could be found in the "en-bloc" specimen. But there were some minor residual carcinomatous infiltrates in the soft tissues between coccygeal bone and dorsal wall of the vagina (Figure 5). Bone tissue and the vaginal wall structures themselves proofed to be free of tumour histologically. No florid, ulcerating or granulomatous inflammation was found in the remaining mucosal tissues.
Figure 4 Histopathological overview of the fistula opening on the perineum consisting of mucus containing cystic spaces, partially lined by cylindrical epithelium of the mucinous adenocarcinoma. H&E, × 25.
Figure 5 Dermal infiltration of the mucinous adenocarcinoma undermining the perineal skin. PAS-stain, × 50.
Discussion
Despite the original Crohn's disease description in 1932, it was not until 1938 that Penner and Crohn described the presence of a perianal fistula in a patient with Crohn's disease [6]. Isolated perianal involvement has been reported in fewer than 5% of cases [7]. Patients with colonic involvement are most likely to develop perianal Crohn's disease, seen in 46% to 68% of this patient group, and 5% to 27% of patients with small bowel disease develop perianal lesions [8].
Surgical options for perianal Crohn's disease range from abscess drainage as ''first-aid management'' to major interventions [9], such as proctocolectomy and permanent stoma formation as in the presented case. A proctocolectomy was performed in this patient because of severe and recurrent perineal disease that did not respond to local surgery combined with medical treatment. However, among colorectal surgeons poor wound healing and perineal sinus formation are well recognized complications after proctectomy.
As for ulcerative colitis, patients with longstanding Crohn's disease are at increased risk (3.7 per cent) of developing adenocarcinoma. This has been attributed to an early onset and prolonged duration of disease. The incidence of carcinoma is 0.7 per cent in patients with perineal Crohn's disease; both adenocarcinomas and squamous cell carcinomas occur [4,10].
For mucinous adenocarcinoma the overall incidence among all colorectal carcinomas is ranging from 7.8 per cent to 18 per cent. As a rare tumor entity among sporadic colon carcinomas, these tumors are most frequently found in the right-sided colon followed by the rectum and tend to be associated with an inflammatory process, such as colitis, ulcerative colitis and Crohn's disease [12,13]. As assessed by our histopathological studies no residual rectal mucosa was found in the perineum. In the en-bloc specimen signs of persistant perineal inflammatory Crohn's disease were found. There are several publications reporting this kind of adenocarcinoma in association with a neovagina and rectovaginal fistulas [14-16]. Pathogenetically, perianal mucinous adenocarcinoma are thought to arise from the anal ducts if triggered by chronic inflammation [13]. One single case of mucinous carcinoma development in a patient with persistant perineal crohn's disease has been described after ileorectal anastomosis. Here the tumor had formed within the fistula [17]. In our case histomorphologically, it cannot be decided if the tumour developed from residual perianal gland mucosa or has evolved in residual perianal fistulas.
To the best of our knowledge, this is the first case report of a late perianal mucinous adenocarcinoma arising in a patient after proctocolectomy for Crohn's disease. The present case should make surgeons and gastroenterologist aware of the risk of poor healing and the associated morbidity after rectal excision performed for perianal Crohn's disease in order to consider alternative therapeutical options. As a possible alternative, the literature shows that a low Hartmann's procedure may result in a 60 per cent healing rate in patients with perineal disease [18]. However, up to 50 per cent of these patients (perineal Crohn's disease treated by Hartmann's procedure) required a completion proctectomy, because they showed residual disease in the rectal stump [19]. Diversion ileostomy before proctocolectomy, which may convert active disease to a quiescent state, could be another effective alternative in order to avoid sinus development. There are also favorable reports about the use of gracilis transposition flap to treat perineal wounds after proctectomy [20].
The presented case with cancer formation shows, that even a radical surgical approach to treat perineal fistulous disease does not withhold late complications of chronic perineal inflammatory Crohn's disease. Moreover, this report should alert surgeons to avoid a perineal sinus or poor wound healing after proctectomy. Alternative surgical procedures combined with optimized medical therapy should be preferred over proctectomy in Crohn's disease.
==== Refs
Mpofu C Watson AJ Rhodes JM Strategies for detecting colon cancer and/or dysplasia in patients with inflammatory bowel disease Cochrane Database Syst Rev 2004 CD000279 15106148
Shanahan F Colitis-associated cancer – time for new strategies Aliment Pharmacol Ther 2003 18 6 9 12950414 10.1046/j.1365-2036.18.s2.5.x
Winkler R Wittmer A Heusermann U Cancer and Crohn's disease Z Gastroenterol 2002 40 569 76 12297980 10.1055/s-2002-33417
Ky A Sohn N Weinstein MA Korelitz BI Carcinoma arising in anorectal fistulas of Crohn's disease Dis Col Rectum 1998 41 992 996
Cuenod CA de Parades V Siauve N Marteau P Grataloup C Hernigou A Berger A Cugnenc PH Frija G MR imaging of ano-perineal suppurations Radiol 2003 84 516 28
Penner A Crohn BB Perianal fistulae as a complication of regional ileitis Ann Surg 1938 108 867 873 17857277
Lockhart-Mummery H Crohn's disease: anal lesions Dis Colon Rectum 1975 18 200 202 1140047
Alabaz O Beck D, Wexner SD Anorectal Crohn's disease Fundamentals of Anorectal Surgery 1999 498 509
Makowiec F Makowiec F Jehle EC Becker HD Starlinger M Perianal abscess in Crohn's disease Dis Colon Rectum 1997 40 443 450 9106694
Sjodahl RI Myrelid P Soderholm JD Anal and rectal cancer in Crohn's disease Colorectal Dis 2003 5 490 495 12925087 10.1046/j.1463-1318.2003.00510.x
Okuno M Ikehara T Nagayama M Mucinous colorectal carcinoma: Clinical pathology and prognosis Am Surg 1988 54 681 685 2847606
Symonds DA Vichery AL Mucinous adenocarcinoma of the colon and rectum Cancer 1976 37 1891 1900 177180
Prioleau PG Allen MS Roberts T Perianal mucinous adenocarcinoma Cancer 1977 39 1295 1299 199346
Munkarah A Malone JM Budev HD Evans TN Mucinous adenocarcinoma arising in a neovagina Gynecologic Oncology 1994 52 272 275 8314151 10.1006/gyno.1994.1045
Hiroi H Yasugi T Matsumoto K Fujii T Watanabe T Yoshikawa H Taketani Y Mucinous adenocarcinoma arising in a neovagina using the sigmoid colon thirty years after operation: a case report J Surg Oncol 2001 77 61 64 11344485 10.1002/jso.1067
Moore-Maxwell CA Robboy SJ Mucinous adenocarcinoma arising in rectovaginal fistulas associated with Crohn's disease Gynecologic Oncology 2004 93 266 268 15047250 10.1016/j.ygyno.2003.11.056
Buchmann P Allan RN Thompson H Alexander-Williams J Carcinoma in a rectovaginal fistula in a patient with crohn's disease Am J Surg 1980 140 642 4 7435823 10.1016/0002-9610(80)90048-3
Sher ME Bauer JJ Gorphine S Gelernt I Low Hartmann's procedure for severe anorectal Crohn's disease Dis Col Rectum 1992 35 975 980
Guillem JG Roberts PL Murray JJ Coller JA Veidenheimer MC Schoetz DJ Jr Factors predictive of persistent or recurrent Crohn's disease in excluded rectal segments Dis Col Rectum 1992 35 768 772
Rius J Nessim A Nogueras JJ Wexner SD Gracilis transposition in complicated perianal fistula and unhealed perineal wounds in Crohn's disease Eur J Surg 2000 166 218 22 10755336 10.1080/110241500750009311
|
15987512
|
PMC1190221
|
CC BY
|
2021-01-04 16:39:04
|
no
|
World J Surg Oncol. 2005 Jun 29; 3:42
|
utf-8
|
World J Surg Oncol
| 2,005 |
10.1186/1477-7819-3-42
|
oa_comm
|
==== Front
World J Surg OncolWorld Journal of Surgical Oncology1477-7819BioMed Central London 1477-7819-3-531608349810.1186/1477-7819-3-53ResearchClinico-pathological features of bladder carcinoma in women in Pakistan and smokeless tobacco as a possible risk factor Rafique Muhammad [email protected] Department of Urology, Nishtar Medical College, Multan, Pakistan2005 5 8 2005 3 53 53 12 4 2005 5 8 2005 Copyright © 2005 Rafique; licensee BioMed Central Ltd.2005Rafique; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Bladder carcinoma is one of the common urological malignancies occurring worldwide in both sexes. Use of smokeless tobacco by women is common in rural areas of Pakistan. The clinico-pathological features of bladder carcinoma in women and association of smokeless tobacco as a possible risk factor for bladder carcinoma has not been well described in the literature. The objective of the study was to determine the clinico-pathological features of histologically confirmed bladder carcinoma in women and to investigate the role of smokeless tobacco use as a possible risk factor for its development.
Patients and methods
Of the 204 patients (160 male and 44 female M:F ratio 3.6:1) of newly diagnosed bladder carcinoma treated at Nishtar Medical College Hospital Multan from January 1998 to December 2004, the 44 female patients were evaluated with respect to age, clinical presentation, cystoscopic findings, histopathological reports and possible etiological factors. Data were collected and prospectively updated at the time of discharge from hospital and during follow-up in urology out-patient clinic.
Results
Transitional cell carcinoma accounted for all of the bladder carcinoma in women. Median age of the patients was 55 years and 68% patients were under 60 years of age. Majority of patients (88%) presented with hematuria. Eleven (25%) patients had superficial (pTa/pT1) while 33 (75%) patients had muscle invasive (T2–T4) bladder carcinoma. Most (81%) superficial tumors were papillary while muscle invasive tumors had solid configuration at cystoscopy. Of these, 21 (47%) patients had long history of smokeless tobacco use (chewable or moist snuff).
Conclusion
Transitional cell carcinoma is the most common bladder malignancy in women in Pakistan. Many women with bladder carcinoma had long history of use of smokeless tobacco. Majority of patients presented with hematuria and were under 60 years of age. At the time of diagnosis 75% women had muscle invasive bladder carcinoma. In women using smokeless tobacco, the correlation between stage of bladder carcinoma and duration of smokeless tobacco use was significant (p = 0.03). Further studies are needed to clarify the role of smokeless tobacco in the development of bladder carcinoma.
==== Body
Background
Bladder carcinoma is one of the most common malignancies occurring worldwide. It is seen mainly in men. The incidence in women is approximately 3 to 4 times lower than in men but it seems to be rising [1]. Bladder cancer has been associated pathogenetically with many etiological factors which include occupational exposure to certain chemicals e.g. aniline dyes, cigarette smoking, viral agents, bacterial and parasitic infections, cystolithiasis, cyclophosphamide therapy and pelvic irradiation [2].
The initial clinical evaluation consists of history and physical examination, upper tract studies (IVU +/- Ultrasonography) and urine cytology followed by cystoscopy and transurethral resection of bladder tumor.
Most cases of the bladder carcinoma are superficial at the time of diagnosis (stage Ta-T1). The recurrence of the superficial tumors can be as high as 70%, with 10–15% progressing to muscle invasive disease [3].
Despite the fact that bladder carcinoma is among one of the common malignancies in women worldwide, the etiological and clinico-pathological aspects of bladder carcinoma are not well described in the literature. In contrast to many Western countries use of smokeless tobacco in women is quite common in rural areas of Pakistan. The primary objective of the present study was to determine the clinico-pathological features of histologically confirmed bladder carcinoma in women and the secondary objective was to investigate smokeless tobacco use as a possible risk factor for it.
Patients and methods
Two hundreds and four patients of newly diagnosed bladder carcinoma were treated in the department of urology, Nishtar Hospital Multan, Pakistan from Jan 1998 to December 2004. Age, clinical presentation, cystoscopic findings and histopathological reports, and possible etiological risk factors of bladder carcinoma in women were studied prospectively.
After initial clinical evaluation and routine hematological, biochemical and radiological investigations all patients underwent cystoscopy and transurethral resection of bladder tumor (TURBT). All women were treated as inpatients and none of the patients had undergone TURBT previously. In all cases complete removal of papillary tumor was performed. In cases of solid muscle invasive tumors either complete resection or generous debulking of exophytic tumor was carried out. Cystoscopic tumor configuration was compared with the histopathological reports. Data were collected and prospectively updated at the time of discharge from hospital and during follow-up in urology out-patient clinic.
Detailed information about the smoking habits, use of smokeless tobacco (chewable or snuff), use of hair coloring dyes, occupational exposure to chemicals was obtained from all patient. Many of the women had long history of smokeless tobacco use, its duration and frequency was inquired from such patients. Possible effect of smokeless tobacco on the depth (T category) of bladder carcinoma in such patients was studied and compared with patients not using any form of tobacco.
Results
Two hundred and four patients (160 male and 44 female with male female ratio 3.6:1) were treated. The age of the female patients ranged from 26–80 years (median age 55 years). Hematuria was the predominant symptom in 39 (88.6%) patients at the time of presentation. A total of 21 patients had painless and 18 patients had painful hematuria. Four patients presented with various urinary complaints but had no hematuria. In one patient the bladder tumor was incidentally detected on ultrasonography performed for some other complaints. The mean duration of symptoms was 4 months (range 2 weeks to 16 years)
At presentation mean hemoglobin concentration was 9.5 grams/dL (range 3.9 grams to 13.3 grams). Six (13.6%) patients had renal insufficiency (serum creatinine > 1.5 mg %) secondary to ureteric obstruction from bladder carcinoma. All patients had transitional cell carcinoma. Eleven (25%) patients had non-invasive superficial (i.e. pTa or pT1) transitional cell carcinoma while 33 (75%) patients had muscle invasive (T2–T4) transitional cell carcinoma. The median duration of symptoms for noninvasive transitional cell carcinoma was 1.5 years (range 2 weeks to 16 years) and it was 4 months (range 1 month to 2 years) for muscle invasive carcinoma.
Most superficial tumors had papillary and muscle invasive tumors had solid configuration at cystoscopy. Average size of the superficial and invasive tumor was 4 cm (range 1–8 cm) and 3.8 cm (range 2–8 cm) respectively. Of the superficial tumors 2 (18%) were pTa and 9 (82%) were pT1 tumors. Three pT1 tumors were grade I and five patients had grade II tumors. One patient had high-grade pT1 grade III carcinoma. There was no carcinoma in situ although no random biopsies were taken.
Of the muscle invasive tumors T2, T3 and T4 tumors were present in 14, 15 and 4 patients respectively. Thirteen patients with muscle invasive disease had histological grade III carcinoma while twenty patients had grade II carcinoma. None of the patients had GI tumor. Two patients had marked iliac and para-aortic lymphadenopathy while one patient had iliac and para-aortic lymphadenopathy and liver metastases at the time of diagnosis.
All females were non smokers but 21 (47%) patients had long history of smokeless tobacco use (moist snuff (niswar) 12 patients, chewable tobacco (beera) in 5 patients and chewed tobacco with betel nuts (pan) in 4 patients most of these were currently using these substances at the time of presentation. All such patients came from rural areas of Punjab and were uneducated. Patients were asked whether they were using smokeless tobacco 5 or less than 5 times/day or greater than 5 times /day.
By employing cross tabulation (table 1) of duration of smokeless tobacco use and depth (T category) of bladder carcinoma, it appears that majority of such patients had muscle invasive carcinoma at presentation and about 60% of patients have been using smokeless tobacco between 20–30 years. The correlation between the depth of bladder carcinoma and duration of smokeless tobacco use was 0.473 which is statistically significant (p = 0.03). However correlation between bladder carcinoma and intensity of exposure to smokeless tobacco was 0.24 which showed a weak relationship. This might be due to the fact that the patients were using different quantities of smokeless tobacco from different sources.
Table 1 Depth (T category) of bladder carcinoma and duration of smokeless tobacco use
Duration of smokeless tobacco use
1–10 years 11–20 years 21–30 years 31–40 years Total
T1 (pTa/pT1) 1 2 1 4
T2 3 3 6
T3 2 6 1 9
T4 2 2
Total 1 7 12 1 21
We applied two independent sample t test for the comparison of depth of invasion. (by using T category of TNM stage) of bladder carcinoma in users and nonusers of smokeless tobacco. Patients using smokeless tobacco were assigned to group I and nonusers to group II (i.e. control group). The mean depth of bladder carcinoma in group I and II was 2.43 (standard deviation 0.93) and 2.13 (standard deviation 0.97) respectively. The datum shows that the carcinoma are of higher depth (T category) in group I, however the t value for the difference between the two groups is 1.04 which indicated that this difference was not statistically significant. Larger studies should be able to clarify the role of smokeless tobacco as an etiological risk-factor for bladder carcinoma.
Discussion
Bladder carcinoma is the fourth most common cancer in men in the USA and eight most common cancers in women [4]. In Pakistan bladder carcinoma is one of the top ten malignancies in men and most common urological malignancy in both sexes [5]. Bladder cancer predominantly affects male, with a sex ratio of 3:1, suggesting sex-linked etiological factors [6]. In the present study the male female sex ratio was 3.6: 1. In women bladder cancer usually occurs above the age of 60 years [7], in the present study however the median age was 55 years and 68% women were less than 60 years of age.
A neoplastic change in the urothelium is a multi-step phenomenon [8]. The exact genetic events leading to this multi-step transformation are unknown, but they are likely to be multiple and may involve the activation of oncogenes and inactivation or loss of tumor suppression genes [9]
Cigarette smoking is the single most important cause of bladder carcinoma. Smokers have up to four fold higher incidence of bladder cancer than do people who never smoked [10]. The risk correlates with the number of cigarettes smoked, the duration of smoking and the degree of inhalation of smoke. Causative agents in cigarette smoke are thought to be alpha and beta naphthylamine, which are secreted in to urine of smokers [11]. When compared by number of cigarettes are smoked, the risk of bladder carcinoma may be higher in women than men [12] Cigarette smoking accounts for 50% and 31% of bladder cancers in men and women respectively [13]. Other forms of tobacco use are associated with only a slightly higher risk of bladder cancer [10].
In Pakistan 36% of men and 9% women are smokers [14]. Tobacco is also used in other forms such as hookah (hubble bubble), moist snuff used as an oral dip (niswar), chewed with betel nuts (pan) and smoking of rolled dry leaves containing tobacco (beedi). The most common form of tobacco use in women in rural Pakistan is chewing tobacco and snuff but because of cultural prohibitions women may under report use of tobacco [15]. In the present study none of the women were smokers but 47% women had history of intake of moist snuff (niswar) or chewable tobacco (beera and pan) and all came from the rural areas. The median duration of use of such tobacco products was 27 years (range 10–40 years).
Chewing tobacco and snuff contains many carcinogens. The most harmful carcinogens in smokeless tobacco are the tobacco specific nitrosamines (TSNA). They are formed during the growing, curing, fermenting and aging of tobacco [16]. Long term use of snuff can lead to a number of adverse health affects including oral cancer, cardiovascular diseases and gingival diseases [17]. However the etiological relationship between smokeless tobacco and bladder carcinoma has not been well elucidated in the literature and there is still no agreement among the researchers whether smokeless tobacco use enhances the risk of bladder cancer. Some studies reported increased risk of bladder cancer in smokeless tobacco users [18,19] while others [12] could find no such risk in smokeless tobacco users. All studies included smaller number of bladder cancer patients using smokeless tobacco.
In the present study 47% women had long history of smokeless tobacco use and the correlation between the stage of bladder carcinoma and duration of smokeless tobacco was significant (p = 0.03). However, there was weak correlation between bladder carcinoma and intensity of exposure to smokeless tobacco. This might be due to the fact that women were using different quantities and forms of the locally available smokeless tobacco. As there is long latent period between exposure to carcinogens and the development of bladder carcinoma, it is possible that prolong use of smokeless tobacco among women in the present study was either the etiological factor or had modifying effect on its development. However further studies are required to clarify the role of smokeless tobacco in the development of bladder carcinoma.
In the USA percentage of bladder cancers attributed to occupational exposure is 21% for men and 11% for women [1]. None of the women in present study were exposed to occupational chemicals.
Majority of the patients with bladder carcinoma present with either hematuria or irritative voiding symptoms [20]. In the present study majority of women (88%) presented with hematuria while some (9 %) had variable urinary symptoms. In one patient the tumor was incidentally detected on ultrasonography.
In the developed world transitional cell carcinoma is reported for most bladder carcinoma. About 25% of newly diagnosed cancers are muscle invasive (T2–T4); the rest are superficial (70%), classified as limited to the mucosa (pTa), lamina propria (pT1) or being in situ (Tis 5%) [21].
In the present study all patients had transitional cell carcinoma. Some authors have reported that bladder cancer is of a higher stage at initial diagnosis in women [1]. In the present study 25% patients had superficial (pTa/pT1) and 75% patients had muscle invasive balder carcinoma. In a report from another Pakistani centre 97% of bladder carcinomas were muscle invasive [22] but clinico-pathological differences in women with bladder cancer were not separately reported. This high percentage of muscle invasive bladder carcinoma in Pakistan is in contrast to all other studies from USA and Europe.
Overall survival in patients with superficial disease is excellent. However 60%-70% of superficial tumors recur and 5% of pTa and 25% of pT1 tumors progress to invasive disease [21]. For recurrence risk, multiplicity of the tumor is the most important followed by recurrence rate, volume of the tumor, grade and T category. For progression the most important factor is the histological grade and the T category T1 GIII tumors carry poor prognosis and up to 50% progress to invasive disease [23].
Patients with tumors just invading detrusor muscles have a 50% five years survival whereas those whose tumors have invaded beyond the detrusor muscle have a 10% five years. At the time of diagnosis 50% of muscle invasive tumors have occult metastases which will manifest themselves clinically within 12 months; few such patients survive beyond two years [21]. In the present study two patients had marked iliac and para-aortic lymphadenopathy while one patient had iliac and para-aortic lymphadenopathy and liver metastases at the time of diagnosis.
Conclusion
Transitional cell carcinoma is the most common bladder malignancy in women in Pakistan. Most women with bladder carcinoma have long history of use of smokeless tobacco. At the time of diagnosis 75% women have muscle invasive bladder carcinoma. In women using smokeless tobacco, there is significant correlation between stage of bladder carcinoma and duration of smokeless tobacco use. Further studies are required to clarify the role of smokeless tobacco in the development of bladder carcinoma.
Acknowledgements
The author is grateful to Professor Abrar Ahmad Javaid, Head of clinical oncology, Nishtar Medical College, Multan and Prof Hayat Awan, Dean Institute of management sciences, Bahauddin Zakariya University, Multan for their help during the preparation of this manuscript.
==== Refs
Van der Poel HG Mungan NA Witjes JA Bladder cancer in women Int Urogynecol J and pelvic floor dysfunction 1999 10 207 212 10.1007/s001920050046
Lee R Droller MJ Natural history of bladder cancer: implications for therapy Urol Clin North Am 2000 27 1 14 10696240 10.1016/S0094-0143(05)70229-9
Konety BR Williams RD Superficial transitional (Ta/T1/CIS) cell carcinoma of the bladder BJU Int 2004 94 18 21 15217424 10.1111/j.1464-410X.2003.04894.x
Jemal A Murray T Samuels A Ghafoor A Ward E Thum M Cancer statistics 2003 CA Cancer J Clin 2003 53 5 26 12568441
Rafique M Javed AA Role of itravenous urography and transabdominal ultrasonography in the diagnosis of bladder carcinoma Int Braz J Urol 2004 30 185 190 discussion 191 15689243 10.1590/S1677-55382004000300002
Rabbani F Cordon-Cardo C Mutation of cell cycle regulators and their impact on superficial bladder cancer Urol Clin North Am 2000 27 83 102 10696248 10.1016/S0094-0143(05)70237-8
Baniel J Bladder cancer in women Int Urogennecol and pelvic floor dysfunction 1999 10 399 404 10.1007/s001920050068
Shirai T Etiology of bladdercancer Semin Urol 1993 11 113 126 8210833
Olumi AF Skinner EC Tsai YC Jones PA Molecular analysis of human bladder cancer Semin Urol 1990 8 270 277 1980959
Burch RD Rohan TE Howe CR Risch HA Hill GB Steele R Miller AB Risk of bladder cancer by source and type of tobacco exposure Int J Caner 1989 44 622 628
Carroll PR "Tanagho EA, McAninch JW" Urolthelial carcinoma: cancers of bladder, ureter and renal pelvis eds Smith General Urology 2000 15 McGraw Hill publishers. USA 355 377
Castelao JE Yuan Jian-Min Skipper PL Tannenbaum SR Gago-Dominguez M Crowder JS Ross RK Yu MC Gender and smoking related bladder cancer risk J Nat Can Inst 2001 93 538 545 10.1093/jnci/93.7.538
Wynder EL Goldsmith R The epidemiology of bladder cancer: a second look Cancer 1971 40 1246 1268 332323
Alam SE Prevalence and pattern of smoking in Pakistan J Pak Med Assoc 1998 48 64 66 9783029
PMRC National Health survey of Pakistan Network publication services 1998
Hoffmann D Djordjevic MC Chemical composition and carcinogenicity of smokeless tobacco Adv Dent Res 1997 11 322 329 9524432
Fant RV Henningfield JE Nelson RA Pickworth WB Pharmacokinetics and pharmacodynamics of moist snuff in humans Tob Control 1999 8 387 392 10629244
Slattery ML Schumacher MC West DW Robinson LM Smoking and bladder cancer. The modifying effect of cigarettes on other factors Cancer 1998 61 402 408 3334975
Kabat GC Dieck GS Wynder EL Bladder cancer in non smokers Cancer 1986 57 362 367 3942969
Carrion Rafael Seigne J Surgical management of bladder carcinoma Cancer Control 2002 9 284 292 12228754
Leung HY Griffiths TRI Neal DE Bladder cancer Post Grad Med J 1996 72 719 724
Roohullah Nusrat J Hamadani SR Burdy GM Khurshid A Carcinoma urinary bladder: 5 years experience at Cenar, Quetta J Ayyub Med Coll Abottabad 2001 13 14 16
Oosterlinck W Guidelines on diagnosis and treatment of superficial bladder cancer Minerva Urol Nefrol 2004 56 65 72 15195031
|
16083498
|
PMC1190222
|
CC BY
|
2021-01-04 16:39:04
|
no
|
World J Surg Oncol. 2005 Aug 5; 3:53
|
utf-8
|
World J Surg Oncol
| 2,005 |
10.1186/1477-7819-3-53
|
oa_comm
|
==== Front
BMC Complement Altern MedBMC Complementary and Alternative Medicine1472-6882BioMed Central London 1472-6882-5-171610721910.1186/1472-6882-5-17Research ArticleProtective action of a hexane crude extract of Pterodon emarginatus fruits against oxidative and nitrosative stress induced by acute exercise in rats Paula Fernanda BA [email protected]êa Cibele MCP [email protected] Patrícia P [email protected] Ione [email protected] Departamento de Análises Clínicas e Toxicológicas, Escola de Farmácia e Odontologia de Alfenas (EFOA), Alfenas, MG, 37130-000, Brazil2 Departamento de Ciências Biológicas, Escola de Farmácia e Odontologia de Alfenas (EFOA), Alfenas, MG, 37130-000, Brazil3 Faculdade de Fisioterapia, Universidade de Alfenas, MG, 37130-000, Brazil4 Departamento de Bioquímica, Instituto de Biologia, Universidade Estadual de Campinas (UNICAMP), Campinas, SP, 13083-970, Brazil2005 17 8 2005 5 17 17 12 4 2005 17 8 2005 Copyright © 2005 Paula et al; licensee BioMed Central Ltd.2005Paula et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The aim of the present work was to evaluate the effect of a hexane crude extract (HCE) of Pterodon emarginatus on the oxidative and nitrosative stress induced in skeletal muscle, liver and brain of acutely exercised rats.
Methods
Adult male rats were subjected to acute exercise by standardized contractions of the tibialis anterior (TA) muscle (100 Hz, 15 min) and treated orally with the HCE (once or three times with a fixed dose of 498 mg/kg), before and after acute exercise. Serum creatine kinase activity was determined by a kinetic method and macrophage infiltration by histological analyses of TA muscle. Lipid peroxidation was measured as malondialdehyde (MDA) levels. Nitric oxide production was evaluated by measuring nitrite formation, using Griess reagent, and nitrotyrosine was assessed by western blotting.
Results
Serum creatine kinase activities in the controls (111 U/L) increased 1 h after acute exercise (443 U/L). Acute exercise also increased the infiltration of macrophages into TA muscle; lipid peroxidation levels in TA muscle (967%), liver (55.5%) and brain (108.9%), as well as the nitrite levels by 90.5%, 30.7% and 60%, respectively. The pattern of nitrotyrosine formation was also affected by acute exercise. Treatment with HCE decreased macrophage infiltration, lipid peroxidation, nitrite production and nitrotyrosine levels to control values.
Conclusion
Acute exercise induced by functional electrical stimulation in rats resulted in increase in lipid peroxidation, nitrite and nitrotyrosine levels in brain, liver and skeletal muscle. The exercise protocol, that involved eccentric muscle contraction, also caused some muscle trauma, associated with over-exertion, leading to inflammation. The extract of P. emarginatus abolished most of these oxidative processes, thus confirming the high antioxidant activity of this oil which infusions are used in folk medicine against inflammatory processes.
==== Body
Background
Although exercise has salutary effects, strenuous or unaccustomed physical exercise can cause morphological insult to the myocardial and skeletal muscle involved in the activity, as well as to others organs [1]. Tissue damage resulting from acute or chronic exercise ranges from considerable fiber disruption to subcellular damage [2,3]. Such damage may arise from oxidative stress caused by reactive oxygen species (ROS), which elicit different responses, depending on the organ or tissue and its levels of endogenous antioxidant [1]. In response to eccentric contraction-induced muscle damage, neutrophils and macrophages migrate to the site, infiltrate the muscle tissue, activate cytokines, and produce additional ROS [4,5]. These activities may overwhelm the natural antioxidant defenses of the cell and lead to lipid peroxidation and delayed-onset muscle damage [4,5]. Increased levels of lipid peroxidation have been detected after sprint exercise in blood samples from sprint-trained athletes [6] and in the skeletal muscle of rats after sprint and acute exercise [3,5].
Acute exercise also increases nitric oxide (NO) formation and NO synthase (NOS) activity [7,8]. Reid [9] proposed that, in addition to ROS, NO may play an important role in skeletal muscle, and that muscles involved in repetitive contraction may produce sufficient NO to influence the oxidant-antioxidant balance. NO has been implicated in the metabolic control of muscle by its effects on blood delivery, glucose uptake, oxidative phosphorylation, contractility, and excitation-contraction coupling [10]. However, exposure to reactive oxygen and nitrogen species (RONS) may cause lipid peroxidation in cell membranes, which in turn may generate species that damage cell proteins and promote their degradation [11]. These actions may impair the performance, integrity and metabolism of muscle cells [4]. In addition to the increase in lipid peroxidation following acute exercise, the effects of nitration need to be considered, especially since tyrosine nitration is frequently linked to altered protein function during inflammatory conditions [12]. Protein nitration has been suggested to be a final product of the highly reactive nitrogen oxide intermediates(e.g. peroxynitrite) formed in reactions between NO and oxygen-derived species such as superoxide. However, Gunther et al. [13] described a mechanism for NO-dependent tyrosine nitration that does not require the formation of highly reactive nitrogen oxide intermediates such as peroxynitrite or nitrogen dioxide. Peroxidases can also oxidize nitrite to nitrogen dioxide radical, which can cause nitration of tyrosine and tyrosine residues in proteins [14].
Pterodon emarginatus Vog. (Leguminosae, Papilonaceae), popularly known as sucupira branca, is a native tree species widely distributed throughout central Brazil, in the states of Goiás, Minas Gerais and São Paulo. Sucupira infusions are used in folk medicine for the treatment of rheumatism, sore throats and back-bone problems [15]. Pterodon seed oil has cercaricidal activities [16], an acute anti-inflammatory action in rats [17] and an analgesic effect in mice [18]. The seed oil contains diterpenes and furanditerpenes [19]. The most ubiquitous furanditerpene of P. emarginatus fruits is 6α,7β-di-hydroxyvouacapan-17β-oic acid [19], the anti-inflammatory activity of which has been evaluated in rodent experimental models of paw edema induced by carrageenin and nystatin [17,20].
Although there have been a number of reports on the anti-inflammatory activity of Pterodon oil, there has been no investigation of its antioxidant activity. The aim of the present work was to investigate the antioxidant activity and the effect on nitrite production of hexane crude extract of P. emarginatus fruits in acutely exercised rats.
Methods
Animals
The present study is in compliance with the principals and guidelines of the Brazilian Institute for Animal Experimentation (COBEA) and it had institutional ethical approval by the Ethics Committee for Animal Research, Escola de Farmácia e Odontologia de Alfenas/Centro Universitário Federal (Efoa/Ceufe; Permission: 15/04/2004).
Male Wistar rats (270 ± 20 g) were obtained from the Efoa/Ceufe. The rats were housed in a temperature-controlled room on a 12 h light/dark schedule with food and water available ad libitum. The rats were allocated to six experimental groups: (1) rested control group; (2) acutely exercised group; (3) rested group treated with P. emarginatus hexane crude extract (HCE); (4) acutely exercised group treated with the P. emarginatus HCE 0.5 h before exercise, (5) acutely exercised group treated with the P. emarginatus HCE 0.5 h after exercise and (6) acutely exercised group treated with three administrations of HCE given at 24 h, 12 h and 0.5 h before exercise. The rats were sacrificed 1 h, 6 h and 48 h after HCE administration (group 3) or after acute exercise (groups 2, 4–6). For the determination of CK levels the rats were sacrificed immediately or 1 h, 6 h and 48 h after exercise. For each treatment, 8–10 rats were used, as specified in the figure legends.
Acute exercise
The rats were acutely exercised using functional electrical stimulation to produce standardized repetitive activation of the fast tibialis anterior (TA) muscle, according to Matsuura et al. [21] with modifications. Briefly, the animals were boxed and two square electrodes (1 cm2) were then fixed on the skin of their right legs at proximal and distal ends of the TA muscle. TA muscle was stimulated electrically (Neurodyn II-Ibramed) through an electrode with 10 s square wave biphasic pulses and interpulse intervals of 10 s of 5 V for 10 min. Stimulation frequency was 100 Hz, with an on-off ratio of 1:1.
Plant extract preparation and administration to rats
Pterodon emarginatus Vog. fruits were harvested in the state of Minas Gerais, Brazil. A crude hexane extract (HCE) was prepared according to Carvalho et al. [20]. Briefly, the fruits were ground in the presence of hexane using an industrial blender and then placed in a percolator at room temperature and successively washed with hexane. The extraction liquid was collected, filtered and concentrated by rotatory evaporation under reduced pressure. The resulting HCE had an oily characteristic and a density of 0.98 g/mL and was administered orally (by gavage) to the rats, in one or three doses at 498 mg/kg each.
Serum creatine kinase (CK) activity and lactate level determinations
Blood samples were obtained by cardiac punction from anesthetized rats either before or after the acute exercise. The serum CK activity was determined by a colorimetric assay (CK-NAC kit, Bayer), at 340 nm, and was expressed in U/L, where 1 U resulted in the phoshorylation of 1 mmol of creatine per min at 25°C. Serum lactate concentrations were determined by a colorimetric assay at 550 nm, using lactate oxidase [22].
Histological analysis
The tibialis anterior (TA) muscle was removed from all rats and fixed in 4% (v/v) formalin, embedded in paraffin, cut longitudinally into 5 μm sections and stained with hematoxylin-eosin. The density of macrophages and the extent muscle fiber damage were estimated by point counting morphometry, using 15 randomly selected fields of 0.625 mm2 per section.
Measurement of TBARS in brain, liver and TA muscle
As an index of oxidative damage, lipid peroxidation was evaluated by using the level of thiobarbituric acid-reactive substances (TBARS) test [23]. Brain, liver and TA muscle were homogenized in four volumes of 0.1 M phosphate buffered saline (PBS) and centrifuged at 3,000 g, 4°C, for 10 min. Aliquots (0.5 ml) of each homogenate were mixed with 0.5 mL of a solution containing 2% thiobarbituric acid, 0.5 mL of 25% HCl and 45 μL of 2% BHT (butylated hydroxytoluene). The mixtures were heated at 95°C for 10 min and then centrifuged. The supernatant was transferred to a quartz cuvette and the absorption was measured at 532 nm. The absorption values were transformed into concentrations of MDA using 1,1,3,3-tetraethoxypropane (TEP) as standard.
Determination of nitrite in brain, liver and TA muscle
Nitrite is a stable end product of NO, and its concentration was determined by the Griess method [see [24]]. Brain, liver and TA muscle were homogenized in four volumes of 30 mM Tris-HCl buffer, pH 6.8, containing 5 mM EDTA, 250 mM sucrose, 30 mM KCl, 2% β-mercaptoethanol, PMSF (100 μg/mL), benzamidine (5 μg/ml), aprotinin (2 μg/mL) and leupeptin (2 μg/mL), and then centrifuged (12,000 × g, 4°C, 15 min). Protein concentrations were determined by the Coomassie blue dye-binding method [25] using bovine serum albumin as standard. Aliquots (1.0 mL) of the homogenates were removed and diluted with 1.0 mL of Griess reagent (1% sulphanilamide, 2% phosphoric acid and 0.1% naphthyl ethylene diamine dihydrochloride). The absorbance of the chromophore formed during diazotization of the nitrite with sulphanilamide and subsequent coupling with naphthylethelene diamine was measured at 545 nm. Appropriate blanks and controls were run in parallel.
Western blotting analysis
Aliquots (30 μg) of protein from brain, liver and TA muscle were run on 9% gels using sodium dodecyl sulfate-polyacrylamide gel electrophoresis and were electrophoretically transferred overnight, at 4°C, to nitrocellulose membranes in 20 mM Tris, 192 mM glycine and 20% methanol. The membranes were blocked for 1 h at 4°C in Tris-buffered saline containing 30% Tween 20 and 50% Triton X-100 (TBSTT) with 1% BSA. The blots were then incubated for 2 h with anti-nitrotyrosine antibody (1:1000) (Sigma) in TBSTT with 1% BSA at room temperature. The membranes were washed five times in TBSTT with 0.1% BSA, and incubated for 1 h with peroxidase-conjugated anti-goat IgG (1:2500) (Amersham) in TBSTT with 1% BSA at room temperature. The blots were again washed five times in TBSTT with 0.1% BSA and then visualized using an Enhanced Chemiluminescence (ECL) detection system (Amersham) [26].
Statistical analysis
The results were expressed as the means ± SE of the number of measurements (independent experiments) shown. Statistical comparisons were done using ANOVA and the Tukey-Kramer test, with P<0.05 indicating significance.
Results
Acute exercise increases serum creatine kinase activity
Serum creatine kinase (CK) activity increased in exercised rats. Maximum CK activity was seen 1 h after acute exercise and remained elevated for up 48 h after exercise (Fig. 1). The serum lactate concentration also increased in the exercised rats (75.36 ± 1.48 mg/dL, n = 8) when compared to the controls (54.67 ± 3.06 mg/dL, n = 8).
Figure 1 Effect of acute exercise on serum creatine kinase (CK) activity. The CK levels were determined in rested control and after acute exercise, at the indicated times. The results are the mean ± SE of 8–10 rats per treatment. Significance was determined with one-way ANOVA followed by the Tukey-Kramer test. Columns with different letters differed significantly (p < 0.05).
HCE reduces macrophage infiltration in TA muscle induced by acute exercise
As shown in Fig. 2, intense exercise increased the infiltration of macrophages into the tibialis anterior (TA) muscle. Higher density was already observed 6 h after exercise and after 48 h it was still elevated. For our study, we have chosen an exercise protocol that involved a polymetric (eccentric) muscle component. Untrained exercised rats therefore experience both oxidative and polymetric stress, with the latter causing some muscle trauma, associated with over-exertion, leading to inflammation. Oral administration of the HCE alone did not affect the density of macrophages in TA muscle. However, the increased density of macrophages, seen in TA muscle of acutely exercised rats was significantly reduced, when the HCE was administered 0.5 h before or 0.5 h after acute exercise. When the HCE was administered 0.5 h after acute exercise, cell density of macrophages reduced to almost control values 48 h after exercise indicating its effectiveness against the delayed-onset of muscle trauma.
Figure 2 Effect of HCE (498 mg/Kg), administrated orally to acute exercised rats, on the density of macrophages in the tibialis anterior muscle. The number of cells per 0.625 mm2 area was determined in muscles of rested control rats (1); acutely exercised rats (2); rested rats treated with HCE (3); acutely exercised rats treated with HCE 0.5 h before exercise (4) and acutely exercised rats treated with HCE 0.5 h after exercise (5). The number of cells was determined at 6 h (A) and 48 h (B) after acute exercise. Each column shows the mean ± SE of 15 randomly selected fields of 8–10 rats. Significance was determined with one-way ANOVA followed by the Tukey-Kramer test. Columns with different letters differed significantly (p < 0.05).
HCE reduces lipid peroxidation in TA muscle, brain and liver induced by acute exercise
Lipid peroxidation, evaluated by TBARS formation (MDA), was assessed in brain, liver and tibialis anterior muscle. Lipid peroxidation increased after acute exercise in all of the tissues analyzed (Fig. 3A). The highest TBARS formation was seen in TA muscle, with an increase of 967%, followed by brain (109%) and liver (55.5%). The time at which maximum lipid peroxidation was seen after exercise differed among the tissues – 48 h in brain and 6 h in liver and TA muscle. In TA muscle, the MDA levels were still elevated after 48 h (Fig. 3A).
Figure 3 Effect of the HCE from P. emarginatus on brain, liver and TA muscle lipid peroxidation after acute exercise. (A) TBARS were determined in tissue homogenates of rested control rats and 1 h (Exer 1 h), 6 h (Exer 6 h) and 48 h (Exer 48 h) after acute exercise. TBARS were determined in brain and liver (B) and in TA muscle (C) homogenates of rested control rats (1); acutely exercised rats (2); rested rats treated with HCE (3); acutely exercised rats treated with HCE 0.5 h before exercise (4), acutely exercised rats treated with HCE 0.5 h after exercise (5) and acutely exercised rats treated with HCE 24, 12 and 0.5 h before exercise (6). TBARS levels were estimated at 48 h (brain), 6 h (liver) and 6 h and 48 h (TA muscle) after exercise, as indicated. Each column shows the mean ± SE of 8–10 rats. Significance was determined with one-way ANOVA followed by the Tukey-Kramer test. Columns with different letters differed significantly (p < 0.05).
Based on the results shown in figure 3A, we studied the effects of the HCE on lipid peroxidation in TA muscle 6 h and 48 h after acute exercise. In brain and liver, the effects of HCE were analyzed at 48 h and 6 h after acute exercise, respectively. The effect of HCE on lipid peroxidation was assessed by administering the extract 0.5 h before and 0.5 h after acute exercise. These treatments completely decreased lipid peroxidation in brain and liver tissues (Fig. 3B). However, in TA muscle, the HCE was effective in reducing lipid peroxidation only 48 h after acute exercise; the levels of MDA were still elevated 6 h after acute exercise (Fig. 3C). Then we examined the effect of three administrations of a fixed dose (498 mg/Kg) of HCE given 24 h, 12 h and 0.5 h before acute exercise. This protocol significantly reduced the TBARS in TA muscle 6 h after exercise (Fig. 3C), indicating that this treatment was more effective in reducing the lipid peroxidation induced by acute exercise in TA muscle, when compared with a single treatment protocol.
HCE reduces nitric oxide levels in TA muscle, brain and liver increased by acute exercise
As shown in Fig. 4A, the basal nitrite concentration in TA muscle was seven and six times greater than in brain and liver, respectively. Despite this difference, acute exercise increased the nitrite production by 60.0%, 30.7% and 90.5% in brain, liver and TA muscle, respectively. The maximum nitrite levels in brain, liver and TA muscle were observed 48 h, 6 h and 1 h after exercise, respectively (Fig. 4A). When HCE was administered 0.5 h before or 0.5 h after acute exercise, the nitrite production decreased to basal levels in all of the tissues (Fig. 4B). The effects of the HCE on nitrite production were assessed at 1 h (TA muscle), 6 h (liver) and 48 h (brain) after exercise because at those times we found the maximal nitrite production in the respective tissues (Fig. 4A). The effect of HCE was similar, regardless of whether it was administered three times or only once (data not shown).
Figure 4 Effect of the HCE from P. emarginatus on nitrite levels in brain, liver and TA muscle after acute exercise. (A) Nitrite levels were estimated by the Griess reaction in tissue homogenates of rested control rats and at 1 h (Exer 1 h), 6 h (Exer 6 h) and 48 h (Exer 48 h) after exercise. (B) Nitrite was determined in tissue homogenates of rested control rats (1); acutely exercised rats (2); rested rats treated with HCE (3); acutely exercised rats treated with HCE 0.5 h before exercise (4) acutely exercised rats treated with HCE 0.5 h after exercise (5). Nitrite levels were estimated at 1 h (TA muscle), 6 h (liver) and 48 h (brain) after exercise. Each column shows the mean ± SE of 8–10 rats. Significance was determined with one-way ANOVA followed by the Tukey-Kramer test. Columns with different letters differed significantly (p < 0.05).
HCE prevents nitration of brain and muscle proteins increased by acute exercise
Western blot analysis using an antibody specific for nitrotyrosine revealed different staining patterns for brain (48 h after exercise), liver (6 h after exercise) and TA muscle (1 h after exercise) that were consistent with the nitration of one or more tyrosine residues on proteins (Fig. 5). In the brain, two more bands (66 and 51 kDa) were detected in the exercised rats. The administration of three doses of HCE (498 mg/kg each), given 24 h, 12 h and 0.5 h before acute exercise, prevented the formation of these bands, indicating that the extract prevented exercise-induced nitrotyrosine formation in the brain. In liver, no nitrotyrosine was detected in the control or HCE administered rats, but in exercised rats three bands (66, 55 and 45 kDa) were evident and the treatment with HCE did not prevent nitrotyrosine formation in this tissue. In TA muscle, a triplet band pattern and a band of 66 kDa were seen in the control. These bands were less intense in the HCE administered rats. The exercised rats showed an intense band at 97 kDa, in addition to the 66 kDa band and the triplet. In TA muscle from HCE-exercised rats, labeling of the 97 kDa was abolished and the intensity of the triplet decreased, indicating that the extract prevented nitrotyrosine formation. The 97 kDa band can be related to phagocyte activity in the TA muscle.
Figure 5 Western blot analysis for nitrotyrosine formation in brain, liver and TA muscle. M, molecular mass markers. Lane 1, rested control rats. Lane 2, rested control rats treated with HCE. Lane 3, acutely exercised rats. Lane 4, acutely exercised rats treated with HCE 24, 12 and 0.5 h before acute exercise. Nitrotyrosine formation was analyzed at 1 h (TA muscle), 6 h (liver) and 48 h (brain) after exercise. The results are representative of three independent experiments.
Discussion
In this work we examined the effects of an HCE from Pterodon emarginatus on the oxidative and nitrosative stress induced by acute exercise. Whereas a large number of studies have tested the effects of vitamins and minerals in preventing oxidation in exercise, there is little information on the effect of plant extracts on lipid peroxidation and nitric oxide formation during acute exercise.
An increase in serum CK and in neutrophil and macrophage infiltration after acute exercise has been reported in humans and animals [27-29]. However, there is controversy about the relationships between CK kinetics and the time at which the most prominent muscle cell damage occurs. The degree of muscle damage is considered to be proportional to muscle loading [4]. Our results showed that CK activity increased after acute exercise, with maximum activity after 1 h and remained elevated for at least 48 h. The number of inflammatory cells increased after 6 h and also persisted for more than 48 h. The HCE prevented the muscle fiber damage induced by electrical stimulation, regardless of whether it was given before or after the acute exercise, and this was reflected in the decrease in macrophage density in the HCE administered rats.
Acute exercise induced the lipid peroxidation in all of the tissues studied, although there were differences in the kinetics of lipid peroxidation among the tissues. These differences have previously been described and it appears to be dependent on factors such as oxygen consumption, oxidant susceptibility and the activation of antioxidant enzymes and other repair systems [1]. Increased levels of lipid peroxidation have also been detected in the skeletal muscle of rats after acute [3] and long-term submaximal [4] exercise. However, supramaximal exercise may produce more ROS and be more damaging. In fact there is a larger increase in TBARS after high-intensity exercise in which lactate production is substantial when compared with moderate-intensity exercise [3,30]. Moreover, a significant relationship between the plasma lactate concentration and lipid peroxidation during progressive incremental exercise has been reported [31]. In agreement with this, lactate increases hydroxyl radical generation by the Fenton reaction resulting in lipid peroxidation [32]. Thus, the increase in lactate concentration seen here after acute exercise could account for the increase in TBARS levels detected in skeletal muscle.
The TBARS also increased in liver after 6 h and in brain after 48 h of acute exercise. In contrast, Kayatekin et al. [33] found that TBARS levels in the liver were unchanged after acute sprint exercise. Part of the reason for these contradictory findings could be attributed to the use of different types and intensities of exercise.
Our results also showed that lipid peroxidation induced by acute exercise were markedly higher in brain than in liver. In fact brain tissue has high content of oxidizable substrates, such as polyunsaturated fatty acids, poor catalase activity and low iron-binding capacity making it particularly prone to oxidative stress damage [1].
Although the HCE from P. emarginatus reduced TBARS formation in all of the tissues, it was most effective in TA muscle when administered at three different times before exercise. Thus, the decrease in MDA levels seen after administration of the HCE indicates that the extract has antioxidant activity. The higher level of HCE needed to prevent lipid peroxidation in TA muscle agreed with the greater levels of lipid peroxidation seen in this organ after electrical stimulation.
Nitrite is a stable metabolite of NO and can be used as an indicator of the overall formation of NO in vivo. Increased nitrite levels as a result of increased NOS activity have been observed in skeletal muscle after contractile activity [34]. In our study, acute exercise increased the nitrite levels in all of the tissues. In TA muscle, an increase on nitrite production and lipid peroxidation occurred within 1 h after acute exercise. However, in brain and liver, the increase in the levels of nitrite occurred at 48 h and 6 h after acute exercise, respectively and the lipid peroxidation increased after 6 h in the brain and liver. These findings indicate that despite the differences in time response of the organs to acute exercise, all of them suffered oxidative and nitrosative stress. Additionally, trauma caused by acute exercise in muscle was shown to be transferable to other organs. However, the highest levels of these compounds were detected in TA muscle, probably because it was the direct site of injury.
Our results also show that the basal nitrite level in TA muscle was five times greater than in brain and liver. This could be partly explained by the fact that skeletal muscle has a relatively high NOS activity compared with other organs [35]. Acute exercise increases NOS activity [8] and thus NO formation [7] in muscle. The increased nitrite formation seen after exercise could be mediated by increased intracellular calcium content of muscle fibers during contraction, which would activate NOS. Consistent with this is the observation that both of the constitutively expressed isoforms of NOS (eNOS and nNOS) require calcium as a cofactor for activation [36], and there is evidence that extracellular calcium may also enhance NOS activation during increased loading [37]. Thus, as calcium enters the sarcoplasm, not only does it induce muscle contraction, but it also activates calcium-dependent, constitutively expressed isoforms of NOS.
The HCE reduced nitrite production in TA muscle when administered before or after acute exercise. The extract was also effective in liver and brain tissues, where it decreased the nitrite formation induced by acute exercise. Our results for lipid peroxidation and nitrite formation agree with previous reports, indicating that the rate of NO production in rat liver is similar to the rate of superoxide anion production [38].
Determination of the protein nitrotyrosine content is frequently used to detect oxidative damage to tissues. Protein nitration has been suggested to be a final target of highly reactive nitrogen oxide intermediates (e.g. peroxynitrite) formed in reactions between NO and oxygen-derived species such as superoxide. Since NO is made by a variety of cell types, the opportunity for the formation of a tyrosine iminoxyl free radical exists in any protein in which tyrosyl radical formation occurs [13]. Additionally, myeloperoxidase and peroxidase can also oxidize nitrite to nitrogen dioxide radicals, which can participate in the nitration of tyrosine residues in protein [14]. Western blotting showed more than one band of nitrotyrosine in proteins in all of the tissues analyzed. Acute exercise increased nitrotyrosine formation in the tissues and HCE prevented the nitration of brain and skeletal muscle proteins. The reduction in nitrite levels and nitrotyrosine formation by HCE indicated that the extract prevented the production of reactive nitrogen species generated from the activity of NOS after acute physical exercise. In liver the HCE had no effect on protein nitration suggesting that, at least, part of the RNS formation and protein nitration in muscle are linked to phagocyte activity. As the HCE presented anti-inflammatory activity, it seemed also to prevent the secondary oxidative damage, caused by activated phagocytes.
The antioxidant action of the HCE could be partly explained by its terpene and phenol content [19]. Studies of the inhibition of lipid peroxidation have demonstrated a role of terpens and phenols [39,40]. Phenolic compounds have a high reactivity towards lipid peroxyl radicals and are thus able to interact with ROS to interrupt the propagation of lipid peroxidation [40]. Di Mascio et al. [41] showed that furan diterpenes isolated from an alcoholic extract of Pterodon sp had a high capacity to quench oxygen singlets but not to scavenge oxyradicals since the extract did not inhibit microssomal lipid peroxidation. A diterpene isolated of the plant Salvia miltiorrhiza was shown to inhibit lipid peroxidation in rat kidney and brain homogenates, and its activity was attributed to the presence of the furan ring in its structure [39]. In mouse brain, plant extracts containing polyphenols inhibited NMDA-induced lipid peroxidation and restored the glutathione levels [42].
Recently, nNOS was found to accommodate phenolic substituted compounds in its active site, indicating that such substances could act as inhibitors of nNOS and also as antioxidants [43]. Since the HCE contains phenolic constituents, this could explain the ability of the extract to inhibit nitrite formation in brain and muscle homogenates.
Conclusion
In conclusion, acute exercise induced by functional electrical stimulation in rats resulted in increase in TBARS, nitrite and nitrotyrosine levels in brain, liver and skeletal muscle and macrophage infiltration in muscle fibers. The HCE of P. emarginatus abolished most of these oxidative processes, thus providing for a high antioxidant and anti-inflammatory action against acute exercise. Despite the well-documented protective role of plant antioxidant substances and their ability to protect against the deleterious effects of oxidation, most studies have examined the antioxidant effect of only one class of compound or of an isolated compound, mainly in vitro. Our results showed that a crude extract of P. emarginatus had anti-inflammatory, antioxidant and anti-nitrosative activities in vivo in various organs following acute exercise. Crude plant homogenates are generally less expensive to obtain and more accessible to the population than an isolated compound. Thus, our results support the beneficial effects of the Pterodon extract used popularly, and suggest its potential use in humans as a therapeutic agent against oxidative damage.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
FBAP carried out all of the experiments and participated in the drafting of the manuscript, PPA participated in the experiments of histological analysis, CMCPG supervised the work of FBAP and PPA, participated in the design drifting of the manuscript and IS coordinated the study and drafted the manuscript. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
This work was partly supported by the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) and by research fellowships from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES).
==== Refs
Liu J Yeo HC Övervik-Douki E Hagen T Doniger SJ Chyu DW Brooks GA Ames BN Chronically and acutely exercised rats: biomarkers of oxidative stress and endogenous antioxidants J App Physiol 2000 89 21 28
Davies KJ Quintanilha AT Brooks GA Packer L Free radicals and tissue damage produced by exercise Biochem Biophys Res Commun 1982 107 1198 1205 6291524 10.1016/S0006-291X(82)80124-1
Alessio HM Goldfarb AH Lipid peroxidation and scavenger enzymes during exercise: adaptive response to training J Appl Physiol 1988 64 1333 1336 3378967
Sacheck J Blumberg J The role of vitamin E and oxidative stress in exercise Nutrition 2001 17 809 814 11684385 10.1016/S0899-9007(01)00639-6
Proske U Morgan DL Muscle damage from eccentric exercise: mechanism, mechanical signs, adaptation and clinical applications J Physiol 2001 537 333 345 11731568 10.1111/j.1469-7793.2001.00333.x
Marzatico F Pansara O Bertorelli L Somenzini L Della Valle G Blood free radical antioxidant enzymes and lipid peroxides following long-distance and lactacidemic performances in highly trained aerobic and sprint athletes J Sports Med Phys Fitness 1997 37 235 239 9509820
Jungersten L Ambring A Wall B Wennmalm A Both physical fitness and acute exercise regulate nitric oxide formation in healthy humans J Appl Physiol 1997 82 760 764 9074960 10.1063/1.365855
Roberts CK Barnard RJ Jasman A Balon TW Acute exercise increases nitric oxide synthase activity in skeletal muscle Am J Physiol 1999 277 E390 E394 10444436
Reid MB Role of nitric oxide in skeletal muscle: synthesis, distribution and functional importance Acta Physiol Scand 1998 162 401 409 9578386 10.1046/j.1365-201X.1998.0303f.x
Stamler JS Meissner G Physiology of Nitric Oxide in Skeletal Muscle Physiol Rev 2001 81 209 237 11152758
Davies KJ Goldberg AL Proteins damaged by oxygen radicals are rapidly degraded in extracts of red blood cells J Biol Chem 1987 262 8227 8234 3597373
Eiserich JP Patel RP O'Donnell VB Pathophysiology of nitric oxide and related species: free radical reactions and modification of biomolecules Mol Aspects Med 1998 19 221 357 10231805 10.1016/S0098-2997(99)00002-3
Gunther MR His LC Curtis JF Gierse JK Marnett LJ Eling TE Mason RP Nitric oxide trapping of the tyrosyl radical of prostaglandin H synthase-2 leads to tyrosine iminoxyl radical and nitrotyrosine formation J Biol Chem 1997 272 17086 17090 9202025 10.1074/jbc.272.27.17086
Hirota S Takahama U Ly TN Yamauchi R Quercetin-dependent inhibition of nitration induced by peroxidase/H2O2/nitrite systems in human saliva and characterization of an oxidation product of quercetin formed during the inhibition J Agric Food Chem 2005 53 3265 3272 15853358 10.1021/jf0404389
Correa MP Dicionário de plantas úteis do Brasil e das exóticas cultivadas 1984 Ministério da Agricultura. Rio de Janeiro: Instituto Brasileiro de Desenvolvimento Florestal
Mors WB dos Santos Filho MF Monteiro HJ Gilbert B Chomoprophylactic agent in Schistosomiasis: 14,15-epoxygeranylgeraniol Science 1967 157 950 951 5006418
Nunan EA Carvalho MG Piló-Veloso D Turchetti-Maia RMM Ferreira-Alves DL Furane diterpenes with anti- and pro-inflammatory activity Braz J Med Biol Res 1982 15 450
Duarte ID Ferreira-Alves DL Veloso DP Nakamura-Craig M Evidence of the involvement of biogenic amines in the antinociceptive effect of vouacapan extracted from Pterodon polygalaeflorus Benth J Ethnopharmcol 1996 55 13 18 10.1016/S0378-8741(96)01465-1
Mahjan JR Monteiro MB New diterpenoids from Pterodon emarginatus Vog Anais Acad Bras Ciênc 1970 42 103 107
Carvalho JC Sertié JA Barbosa MV Patrício KC Caputo LR Sarti SJ Ferreira LP Bastos JK Anti-inflammatory activity of the crude extract from the fruits of Pterodon emarginatus Vog J Ethnopharmacol 1999 64 127 133 10197747 10.1016/S0378-8741(98)00116-0
Matsuura T Ikata T Takata S Kashiwaguchi S Niwa M Sogabe T Koga K Effect of weight bearing on recovery from nerve injury in skeletal muscle J Appl Physiol 2001 91 2334 2341 11641378
Moss DW Burtis CA, Ashwood ER Clinical Enzymology Tietz Textbook of Clinical Chemistry 1999 3 WB: Saunders Company 617 721
Winterbourn CC Gutteridge JM Halliwell B Doxorubicin-dependent lipid peroxidation at low partial pressures of O2 J Free Rad Biol Med 1985 1 43 49 10.1016/0748-5514(85)90028-5
Scuro LS Simioni PU Gabriel DL Saviani EE Modolo LV Tamashiro WMSC Salgado I Supression of nitric oxide production in mouse macrophage by soybean flavonoids accumulated in response to nitroprusside and fungal eliciation BMC Biochem 2004 5 5 12 15102332 10.1186/1471-2091-5-5
Bradford MM A rapid and sensitive method for the quantification of microgram quantities of protein utilizing the principle of protein-dye binding Anal Bioch 1976 72 248 254
Kanski J Alterman MA Schöneich C Proteomic identification of age-dependent protein nitration in rat skeletal muscle Free Radic Biol Med 2003 35 1229 1239 14607522 10.1016/S0891-5849(03)00500-8
Pizza FX Koh TJ Mcgregor SJ Brooks SV Muscle inflammatory cells after passive stretches, isometric contractions, and lengthening contractions J Appl Physiol 2002 92 1873 1878 11960936 10.1063/1.1489068
Clarkson PM Tremblay I Exercise-induced muscle damage, repair, and adaptation in humans J Appl Physiol 1988 65 1 6 3403453
Sacheck JM Millbury PE Cannon JG Roubenoff R Blumberg JB Effect of vitamin e and eccentric exercise on selected biomarkers of oxidative stress in young and elderly men Free Rad Biol Med 2003 34 1575 1588 12788477 10.1016/S0891-5849(03)00187-4
Salminen A Vihko V Lipid peroxidation in exercise myopathy Exp Mol Pathol 1983 38 380 388 6852210 10.1016/0014-4800(83)90077-1
Lovlin R Cottle W Pyke I Kavanagh M Belcastrom AN Are indices of free radical damage related to exercise intensity? Eur J Appl Physiol Occup Physiol 1987 56 313 316 3569239 10.1007/BF00690898
Halliwell B Gutteridge JMC Free Radicals in Biology and Medicine 1989 Oxford: Clarendon Press
Kayatekin BM Gonenc S Acikgoz O Uysal N Dayi A Effects of sprint exercise on oxidative stress in skeletal muscle and liver Eur J Appl Physiol 2002 87 141 144 12070624 10.1007/s00421-002-0607-3
Reiser PJ Kline WO Vaghy PL Induction of neuronal type nitric oxide synthase in skeletal muscle by chronic electrical stimulation in vivo J App Physiol 1997 82 1250 1255
Slater M Knowles RG Moncada S Widespread tissue distribution, species distribution and changes in activity of Ca(21)-dependent and Ca(21)-independent nitric oxide synthases FEBS Lett 1991 291 145 149 1718778 10.1016/0014-5793(91)81123-P
Moncada S Higgs A The L-arginine-nitric oxide pathway N Engl J Med 1993 329 2002 2012 7504210 10.1056/NEJM199312303292706
Stull M Wehling M Mechanical loading regulates NOS expression and activity in developing and adult skeletal muscle Am J Physiol 1998 275 C260 C266 9688857
Giulivi C Poderoso JJ Boveris A Production of nitric oxide by mitochondria J Biol Chem 1998 273 11038 11043 9556586 10.1074/jbc.273.18.11038
Ng TB Liu F Wang ZT Antioxidant activity of natural products from plants Life Sci 2000 66 709 723 10680579 10.1016/S0024-3205(99)00642-6
Noguchi N Niki E Phenolic antioxidants: a rationale for design and evaluation of novel antioxidant drug for atherosclerosis Free Rad Biol Med 2000 28 1538 1546 10927179 10.1016/S0891-5849(00)00256-2
Di Mascio PD Medeiros MHG Sies H Bertolotti S Braslavsky SE Piló Veloso D Sales BHLN Magalhães E Braz-Filho R Bechara EJH Quenching of singlet molecular oxygen by natural furan diterpenes J Photochem Photobiol B: Biology 1997 38 169 173 10.1016/S1011-1344(96)07444-1
Lee KJ Protective effect of vegetable extracts on oxidative stress in brain of mice administered with NMDA Food Res Int 2002 35 55 63 10.1016/S0963-9969(01)00119-3
Auvin S Auget M Navet E Harnett JJ Viossat I Schulz J Bigg D Chabrier Pierre-E Novel inhibitors of neuronal nitric oxide sythase with potent antioxidant properties Bioorg Med Chem Let 2003 13 209 212 12482425 10.1016/S0960-894X(02)00883-1
|
16107219
|
PMC1192789
|
CC BY
|
2021-01-04 16:31:46
|
no
|
BMC Complement Altern Med. 2005 Aug 17; 5:17
|
utf-8
|
BMC Complement Altern Med
| 2,005 |
10.1186/1472-6882-5-17
|
oa_comm
|
==== Front
BMC BioinformaticsBMC Bioinformatics1471-2105BioMed Central London 1471-2105-6-1591598517810.1186/1471-2105-6-159Methodology ArticleFull cyclic coordinate descent: solving the protein loop closure problem in Cα space Boomsma Wouter [email protected] Thomas [email protected] Bioinformatics center, Institute of Molecular Biology and Physiology, University of Copenhagen, Universitetsparken 15, Building 10, DK-2100 Copenhagen, Denmark2005 28 6 2005 6 159 159 25 4 2005 28 6 2005 Copyright © 2005 Boomsma and Hamelryck; licensee BioMed Central Ltd.2005Boomsma and Hamelryck; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Various forms of the so-called loop closure problem are crucial to protein structure prediction methods. Given an N- and a C-terminal end, the problem consists of finding a suitable segment of a certain length that bridges the ends seamlessly.
In homology modelling, the problem arises in predicting loop regions. In de novo protein structure prediction, the problem is encountered when implementing local moves for Markov Chain Monte Carlo simulations.
Most loop closure algorithms keep the bond angles fixed or semi-fixed, and only vary the dihedral angles. This is appropriate for a full-atom protein backbone, since the bond angles can be considered as fixed, while the (φ, ψ) dihedral angles are variable. However, many de novo structure prediction methods use protein models that only consist of Cα atoms, or otherwise do not make use of all backbone atoms. These methods require a method that alters both bond and dihedral angles, since the pseudo bond angle between three consecutive Cα atoms also varies considerably.
Results
Here we present a method that solves the loop closure problem for Cα only protein models. We developed a variant of Cyclic Coordinate Descent (CCD), an inverse kinematics method from the field of robotics, which was recently applied to the loop closure problem. Since the method alters both bond and dihedral angles, which is equivalent to applying a full rotation matrix, we call our method Full CCD (FCDD). FCCD replaces CCD's vector-based optimization of a rotation around an axis with a singular value decomposition-based optimization of a general rotation matrix. The method is easy to implement and numerically stable.
Conclusion
We tested the method's performance on sets of random protein Cα segments between 5 and 30 amino acids long, and a number of loops of length 4, 8 and 12. FCCD is fast, has a high success rate and readily generates conformations close to those of real loops. The presence of constraints on the angles only has a small effect on the performance. A reference implementation of FCCD in Python is available as supplementary information.
==== Body
Background
Many protein structure prediction methods require an algorithm that is capable of constructing a new conformation for a short segment of the protein, without affecting the rest of the molecule. In other words, a protein fragment needs to be generated that seamlessly closes the gap between two given, fixed end points. This problem is generally called the loop closure problem, and was introduced in a classic paper by Go and Scheraga more than 30 years ago [1]. It has been the continued subject of intensive research over many years due to its high practical importance in structure prediction.
The loop closure problem arises in at least two different structure prediction contexts. In homology modelling, it is often necessary to rebuild certain loops that differ between the protein being modelled and the template protein [2]. The modelled loop needs to bridge the gap between the end points of the template's loop.
In de novo prediction, local resampling or local moves can be considered as a variant of the loop closure problem. Typically, the conformation of a protein segment needs to be changed without affecting the rest of the protein as a sampling step in a Markov Chain Monte Carlo (MCMC) procedure [3]. In both homology and de novo structure prediction, the problem is however essentially the same.
The classic article by Go and Scheraga [1] describes an analytical solution to finding all possible solutions for a protein backbone of three residues. In this case, the degrees of freedom (DOF) comprise six dihedral angles, ie. the backbone's (φ, ψ) angles. Another approach is to use a fragment library derived from the set of solved protein structures, and look for fragments or combinations of fragments that bridge the given fixed ends [4-6]. More recently, the loop closure problem has been tackled using algorithms borrowed from the field of robotics, in particular inverse kinematics methods [7-9]. Still other methods use various Monte Carlo chain perturbation approaches, often combined with analytical methods [10,3,12]. A good overview of loop closure methods and references can be found in Kolodny et al. (2005) [6].
Most methods assume that one is working with a full-atom protein backbone with fixed bond angles and bond lengths, so the DOF consist solely of the backbone's (φ, ψ) angles. However, in many cases not all the atoms of the protein backbone are present in the model. In particular, a large class of structure prediction, design and in silico folding methods makes use of drastically simplified models of protein structure [13,14].
A protein structure might for example be represented by a chain of Cα atoms or a chain of virtual atoms at the centers of mass of the side chain atoms [15]. In these models, there is obviously no full-atom model of the protein's backbone available.
In the case of Cα-only models, the structure can be described as a sequence of pseudo bonds, pseudo angles θ and pseudo dihedral angles τ [16]. Here, the term 'pseudo' indicates that the consecutive Cα's are not actually connected by chemical bonds. As in the case of the protein's backbone, the pseudo bond lengths can be considered fixed (typically 3.8 Å). In contrast, the pseudo bond angles between three consecutive Cα atoms are most definitely not fixed, but vary between 1.4 and 2.7 radians. Hence, a Cα-only model of N residues can be represented by a sequence of N - 2 pseudo bond angles θ and N - 3 pseudo dihedral angles τ (Figure 1).
Figure 1 A protein segment's Cα trace. The Cα positions are numbered, and the pseudo bond angles θ and pseudo dihedrals τ are indicated. The segment has length 5, and is thus fully described by two pseudo dihedral and three pseudo bond angles.
Most inverse kinematics approaches assume that the DOF consist only of dihedral angles, and keep the bond angles fixed or semi-fixed. Hence, they cannot be readily applied to the Cα-only case without restricting the search space unnecessarily. In principle, fragment library based methods would apply, but here the problem of data sparsity arises [17,18]. Often, no suitable fragments can be found if the number of residues between the fixed ends becomes too high.
In order to solve the loop closure problem in Cα space, we extend a particularly attractive approach that was recently introduced by Canutescu & Dunbrack [8]. The algorithm is called Cyclic Coordinate Descent (CCD), and like many other loop closure algorithms it derives from the field of robotics [19]. As pointed out by Canutescu & Dunbrack, the CCD algorithm is meant as a black box method that generates plausible protein segments that bridge two given, fixed endpoints. The final choice is typically made based upon the occurrence of steric clashes, applicable constraints (for example side chain conformations) and evaluation of the energy.
The CCD algorithm does not directly generate conformations that bridge a given gap, but alters the dihedral angles of a given starting segment that already overlaps at the N-terminus such that it also closes at the C-terminus. The starting segment can be generated in many ways, for example by using a fragment library derived from real structures or by constructing random artificial fragments with reasonable conformations. Surprisingly, most protein loops can be closed efficiently by CCD starting from artificial loops constructed with random (φ, ψ) dihedral angles [8].
The CCD algorithm alters the (φ, ψ) dihedral angles for every residue in the segment in an iterative way. In each step, the RMSD between the chain end and the overlap is minimized by optimizing one dihedral angle. Because only one dihedral angle is optimized at a time, the optimal rotation can be calculated efficiently using simple vector arithmetic.
The list of advantages of CCD is impressive: it is conceptually simple and easy to implement, computationally fast, very flexible (ie. capable of incorporating various restraints and/or constraints) and numerically stable. Therefore, we decided to adopt the CCD algorithm for use with Cα-only models. Here, we describe a new version of CCD that optimizes both dihedral angles and bond angles, while maintaining all the advantages of the CCD method. We call our method Full Cyclic Coordinate Descent (FCCD), where "Full" indicates that both dihedral angles and bond angles are optimized, while only the bond lengths remain fixed. At the heart of the FCCD method lies a procedure to superimpose point sets with minimal Root Mean Square Deviation (RMSD), based on singular value decomposition. As is the case for the CCD algorithm, FCCD is not a modelling method in itself. Rather, it can be used as a method to generate possible conformations that can be evaluted using some kind of energy function.
To test the algorithm, we selected random segments from a protein structure database, and evaluated the efficiency of closing the corresponding gaps starting from artificial segments with protein-like (θ, τ) angles. We show that FCCD is both fast and successful in solving the loop closure problem, even in the presence of angle constraints. Conformations close to those of real protein loops are readily generated. Finally, we discuss possible applications of the FCCD algorithm, and mention some possible disadvantages.
Results and discussion
Overview of the FCCD algorithm
Figure 2 illustrates the essence of the FCCD algorithm, and Table 3 provides detailed pseudo code. Here we define some of the terms that will be used throughout the article, and provide a high level overview of the FCCD algorithm.
Figure 2 The action of the FCCD algorithm in Cα space. The Cα traces of the moving, fixed and closed segments are shown in red, green and blue, respectively. The Cα atoms are represented as spheres. The labels f0, f1 and f2 indicate the three fixed vectors at the N-terminus that are initially common between the fixed and moving segments. The loop is closed when the three C-terminal vectors of the moving segment (labelled mN-3, mN-2, mN-1) superimpose with an RMSD below the given threshold on the three C-terminal vectors of the fixed segment (labelled (fN-3, fN-2, fN-1). This figure and Figure 3 were made with PyMol .
The fixed segment is a list of Cα vector positions that specifies the gap that needs to be bridged. Only the first and last three Cα positions, with corresponding vectors (f0, f1, f2) and (fN-3, fN-2, fN-1) are relevant. We will call these two sets of vectors the N- and C-terminal overlaps, respectively. The moving segment is a list of Cα position vectors that will be manipulated by the FCCD algorithm to bridge the gap. The closed segment is the moving segment after its pseudo bond angles and pseudo dihedral angles were adjusted to bridge the N- and C-terminal overlaps of the fixed segment. The vectors describing the positions of the Cα atoms in a segment of N residues are labelled from 0 to N - 1.
Initially, the first three vectors of the moving loop coincide with the first three vectors of the fixed segment, while the last three vectors are conceivably reasonably close to the last three vectors of the fixed loop. This last condition is however not very critical. The moving segment can be obtained using any algorithm that generates plausible Cα fragments, including deriving them from real protein structures. The fixed segment is typically derived from a real protein of interest, or a model in an MCMC simulation.
The FCCD algorithm changes the pseudo bond angles and pseudo dihedral angles of the moving loop in such a way that the RMSD between the last three vectors of the moving loop (mN-3, mN-2, mN-1) and the last three vectors of the fixed loop (fN-3, fN-2, fN-1) is minimized, thereby seamlessly closing the gap.
Note that we assume that the last three vectors of the moving and fixed segments can be superimposed with an RMSD of 0.0 Å (see Figure 2). In other words, the first and last pseudo bond angles in both segments are equal. It is however perfectly possible to use segments with different pseudo bond angles at these positions. Since the final possible minimum RMSD will be obviously greater than 0 in this case, the RMSD threshold needs to be adjusted accordingly.
The algorithm proceeds in an iterative way. In each iteration, a vector mi in the moving segment is chosen that will serve as a center of rotation. This chosen center of rotation will be called the pivot throughout this article. Then, the rotation matrix that rotates (mN-3, mN-2, mN-1) on (fN-3, fN-2, fN-1) around the pivot and resulting in minimum RMSD is determined, and applied to all the vectors mj downstream i (with i <j <N). In the next iteration, a new pivot is chosen, and the procedure is repeated. The vectors in the chain can be traversed linearly, or they can be chosen at random in each iteration. The difference between FCCD and CCD is that the latter applies a general rotation to the chain using an atom in the chain as a pivot, while the former only applies a rotation around a single axis. The process is stopped when the RMSD falls below a given threshold.
Finding the optimal (with respect to the RMSD) rotation matrix corresponds to finding one optimal pseudo bond angle and pseudo dihedral angle pair. We define θi as the bond angle of the vectors mi-1, mi, mi+1 and τi as the dihedral angle of the vectors mi-2, mi-1, mi, mi+1 (see Figure 1 and [16]). These definitions have the intuitive interpretation that altering (θi, τi) changes the positions of all Cα's downstream from position i. Conversely, using pivot mi and applying a rotation matrix to all the positions downstream from position i corresponds to changing pseudo bond angle θi and pseudo dihedral angle τi.
For a segment of N Cα's (with N > 3), the pseudo angles range from θ1 to θN-2 and the pseudo dihedrals range from τ2 to τN-2. Since the first and last bond angles of the moving segment are fixed, the pivot points range from position 2 to position N - 3 (with N > 4). The pseudo bond angle and pseudo dihedral angle pairs thus range from (θ2, τ2) to (θN-3, τN-3).
Finding the optimal rotation matrix with respect to the RMSD of the C-terminal overlaps can be efficiently solved using singular value decomposition, as described in detail in the following section.
Finding the optimal rotation
In this section we discuss solving the following subproblem arising in the FCCD algorithm: given a chosen pivot point i in the moving segment, find the optimal (θi, τi) pair that minimizes the RMSD between the last three Cα vectors in the moving segment and the last three Cα vectors in the fixed segment. Recall that the (θi, τi) pair at position i corresponds to the pseudo bond angles and pseudo dihedral angles defined by vectors mi-1, mi, mi+1 and mi-2, mi-1, mi, mi+1 respectively.
Finding the optimal (θi, τi) pair simply corresponds to finding the optimal rotation matrix using Cα position i as the center of rotation (see Figure 2). This reformulated problem can be solved by a variant of a well known algorithm to superimpose two point sets with minimum RMSD which makes use of singular value decomposition [20,21]. Below, we describe this adapted version of the algorithm.
First, the C-terminal overlaps of the moving and the fixed segment need to be translated to the new origin that will be used as pivot for the optimal rotation. This new origin is the pivot vector mi at Cα position i in the moving segment. The new vector coordinates of the moving and the fixed segments are put in two matrices (respectively M and F), with the coordinates of the vectors positioned column wise:
M = [mN-3 - mi | mN-2 - mi | mN-1 - mi]
F = [fN-3 - mi | fN-2 - mi | fN-1 - mi]
Then, the correlation matrix Σ is calculated using M and F :
Σ = FMT
Any real n × m matrix A can be written as the product of an orthogonal n × n matrix U, a diagonal n × m matrix D and an orthogonal m × m matrix VT [22]. Such a factorization is called a singular value decomposition of A. The positive diagonal elements of D are called the singular values. Hence, Σ can be written as:
Σ = UDVT
The optimal rotation Γ is then calculated as follows:
Γ = USVT
The value of the diagonal 3 × 3 matrix S is determined by the product det(U)det(VT), which is either 1 or -1. If this product is -1 then S = diag(1, 1, -1), else S is the 3 × 3 unit matrix. The matrix S ensures that Γ is always a pure rotation, and not a rotation-inversion [21].
In order to apply to all the vectors that are downstream from the pivot point i, these vectors are first translated to the origin of the rotation (ie. pivot point mi), left multiplied by Γ and finally translated back to the original origin:
where i <j <N.
Adding angle constraints to FCCD
It is straightforward to constrain the (θ, τ) angles to a given probability distribution. For each rotation matrix Γ, the resulting new pseudo bond angles and dihedral angles can easily be calculated. The new angles can for example be accepted or rejected using a simple rejection sampling Monte Carlo scheme, comparing the probabilities of the previous pair (θprev, τprev) with that of the next pair (θnext, τnext). If P (θnext, τnext) > P (θprev, τprev) the change is accepted, otherwise it is accepted with a chance proportional to P (θnext, τnext) / P (θprev, τprev). A similar approach was used by Canutescu & Dunbrack [8], and we describe its performance in combination with FCCD in the following section.
More advanced methods could take the probability of the sequence of angles into account as well, for example using a Hidden Markov Model of the backbone [23]. The pseudo code in Table 3 illustrates accepting/rejecting rotations using an unspecified 'accept' function, whose details will depend on the application.
FCCD's performance
In order to evaluate the general efficiency of the method, we selected random fragments of various sizes from a representative database of protein structures, and used these fragments as fixed segments. Hence, the evaluation described below is not limited to loops, but extends to random protein segments. This is a relevant test, since local moves in a typical MCMC simulation are indeed performed on random segments.
The fixed segments were sampled from a dataset of fold representatives (see Methods). First we selected a random fold representative, and subsequently extracted a random continuous fragment of suitable length. The lengths varied from 10 to 30 with a step size of 5. It should be noted that the length of the segment here refers to the number of Cα atoms between the ends that need to be bridged.
The moving segments were generated using random dihedral and bond angles in regions accessible to proteins (see previous section). This was done by sampling the (θi, τi) pairs according to a probability distribution derived from a set of representative protein structures (see Methods). The bond length was fixed at 3.8 Å, in tune with the consensus Cα-Cα distance in protein structures. The last bond angle in the moving segment was chosen equal to the last bond angle in the fixed loop to make a final RMSD of 0.0 Å possible. The RMSD threshold was 0.1 Å. The maximum number of iterations was set to 1000, where one iteration is a sweep over all positions. We ran the FCCD program on 1000 different fixed segments. Table 1 summarizes the results.
Table 1 Performance of the FCCD algorithm for various segment lengths. The first and second number in columns 2–4 refer to unconstrained and constrained FCCD, respectively. Columns 2 and 3 respectively show the average time and number of iterations needed for closing a single segment successfully. The percentage of loops successfully closed in under 1000 iterations is shown in the last column.
Segment length Average time (ms) Average iterations % Closed
5 4.5/51.7 14.0/27.0 99.90/86.50
10 5.2/28.3 10.5/16.8 99.40/98.20
15 5.6/28.6 7.8/12.1 99.60/99.40
20 6.2/27.1 6.3/9.0 99.80/99.40
25 7.6/31.7 5.5/7.6 99.00/99.90
30 7.1/31.0 4.4/6.3 99.70/99.40
A first observation is the effect of the angle constraints. These slow down FCCD with a factor of 10 for small segments (5 residues) and roughly a factor of 5 for larger segments (10 residues or more). Nonetheless FCCD including constraints remains quite speed efficient: small five residue segments are on average closed in about 50 ms, while larger segments (from 10 to 30 residues) are closed considerably faster (on average in about 30 ms). The explanation for this is of course that it is easier to close large segments because they have more DOF. Hence, FCCD, like CCD, is fast and easily handles large segments efficiently.
Overall, the success rate of FCCD is excellent, and very little affected by constraints. For 5 residue segments, adding constraints diminishes the number of successfully closed segments from 99.9% to 86.5%. This effect is however much less pronounced for larger segments: more than 98% percent of the moving/fixed segment pairs can be successfully closed. In short, FCCD is both speed efficient and has a high success rate, even in the presence of constraints.
Evaluation of FCCD's sampling space
Does FCCD potentially generate realistic protein conformations? FCCD could be used to propose possible conformations that are subsequently evaluated by an energy function. In this context, it is of course imperative to generate realistic conformations. To answer this question, we evaluate FCCD's ability to generate closed segments that are close to real protein loops. We used 30 real loops with lengths of 4, 8 and 12 residues as fixed segments. The loop length refers to the number of residues between the N- and C-terminal overlaps.
FCCD was applied using (θ, τ) constraints and an RMSD threshold of 0.1 Å. The maximum number of iterations was set to 1000. For each loop, we attempted to generate closed segments from 1000 random moving segments within the allowed number of iterations. The moving segments were generated as described in the previous section. For all 30 loop cases, we then identified the closed segment that resembled the input loop best as judged by the RMSD. For the calculation of the RMSD, we included the N-and C-terminal overlaps. The results are shown in Table 2, and the best fitting loops for each loop size are shown in Figure 3.
Table 2 Minimum RMSD (out of 1000 tries) between a fixed segment derived from a protein structure and a closed segment generated by FCCD. The length of the loops is shown between parentheses in the upper row.
Loop (4) RMSD Loop (8) RMSD Loop (12) RMSD
1dvj, A, 20–23 0.59 1cru, A, 85–92 2.31 1cru, A, 358–369 3.37
1dys, A, 47–50 0.67 1ctq, A, 144–151 2.22 1ctq, A, 26–37 2.40
1egu, A, 404–407 0.61 1d8w, A, 334–341 2.04 1d4o, A, 88–99 3.20
1ej0, A, 74–77 0.61 1ds1, A, 20–27 2.20 1d8w, A, 43–54 2.74
1i0h, A, 123–126 0.73 1gk8, A, 122–129 2.20 1ds1, A, 282–293 3.16
1id0, A, 405–408 0.66 1i0h, A, 145–152 2.42 1dys, A, 291–302 2.90
1qnr, A, 195–198 0.54 1ixh, 106–113 1.98 1egu, A, 508–519 3.06
1qop, A, 44–47 0.58 1lam, 420–427 2.16 1f74, A, 11–22 3.12
1tca, 95–98 0.76 1qop, B, 14–21 2.17 1q1w, A, 31–42 3.04
1thf, D, 121–124 0.56 3chb, D, 51–58 1.97 1qop, A, 175–186 2.97
Average RMSD 0.63 Average RMSD 2.17 Average RMSD 3.00
Table 3 maxit = maximum number of iterations
moving = N × 3 matrix of Cα positions in moving segment
fixed = N × 3 matrix of Cα positions in fixed segment
threshold = desired minimum RMSD
N = length of the segments
M = 3 × 3 matrix (centered coordinates along columns)
F = 3 × 3 matrix (centered coordinates along columns)
S = diag(1, 1, -1)
repeat maxit:
# Start iteration over pivots
for i from 2 to N-3:
pivot = moving[i,:]
# Make pivot point origin
for j from 0 to 2:
M [:,j] = moving [N-3+j,:]-pivot
F [:,j] = fixed [N-3+j,:]-pivot
# Find the rotation Γ that minimizes RMSD
Σ = FMT
U, D, VT = svd(Σ)
# Check for reflection
if det(U)det(VT)<0:
U = US
Γ = UVT
# Evaluate and apply rotation
if accept(Γ):
# Apply the rotation to the moving segment
for j from i+1 to N-1:
moving [j,:] = Γ (moving [j,:]-pivot)+pivot
rmsd = calc_rmsd(moving [N-3,:], fixed [N-3,:])
# Stop if RMSD below threshold
if rmsd<threshold:
return moving, rmsd
# Failed: RMSD threshold not reached before maxit
return 0
The accept function rejects or accepts the proposed rotation, based on the resulting (θ, τ) pair. The svd function performs singular value decomposition, and calc_rmsd calculates the RMSD between two lists of vectors.
Table 4 SABMark identifiers of the 236 structures used as fold representatives
1ew6a_ 1ail__ 1l1la_ 1kid__ 1n8yc1 1gzhb1 1e5da1 1ep3b2 1ihoa_ 1m0wa1
1dhs__ 1gpua2 2lefa_ 1nsta_ 1eaf__ 1iiba_ 1d5ra2 1foha3 1gpua3 1crza2
3pvia_ 1i6pa_ 1e4ft1 1kx5d_ 2pth__ 1lu9a2 1dkla_ 1fsga_ 1m2oa3 2dpma_
1ajsa_ 1fxoa_ 3tgl__ 1bx4a_ 1mtyg_ 1duvg2 1qopb_ 1iata_ 1k2yx2 1f0ka_
1ayl_1 1toaa_ 8abp__ 1nh8a1 1bi5a2 2mhr__ 1a2pa_ 3lzt__ 1dkia_ 1e7la2
1bf4a_ 1bb8__ 1kpf__ 1mu5a2 1lfda_ 1gpea2 1jqca_ 1a2va2 1jfma_ 1ll7a2
1cjxa1 1lo7a_ 1fm0e_ 1fs1b2 1o0wa2 1dtja_ 1k0ra3 1evsa_ 1jpdx2 1qd1a1
1d5ya3 1h3fa2 1iq0a3 1tig__ 1xxaa_ 1ck9a_ 1gyxa_ 1e5qa2 1ivsa2 1qbea_
3grs_3 1f08a_ 1c7ka_ 1lkka_ 1dq3a3 1uox_1 12asa_ 1bob__ 1m4ja_ 1dv5a_
1f5ma_ 1k2ea_ 1ei1a2 1jdw__ 1ln1a_ 2pola2 1f0ia1 1rl6a1 1fvia2 1j7la_
1is2a1 1e8ga2 1qr0a1 2dnja_ 1kuua_ 1qh5a_ 1ii7a_ 1b8pa2 1j7na3 1chua3
1f00i3 1grj_1 1nkd__ 1mwxa3 1jp4a_ 1ih7a2 1eula2 1gnla_ 1maz__ 2por__
4htci_ 1es7b_ 1tocr1 1d1la_ 1fd3a_ 1i8na_ 1h8pa1 4sgbi_ 1fltv_ 1quba1
1d4va3 1tpg_2 1iuaa_ 1fv5a_ 1mdya_ 1zmec1 1fjgn_ 1eska_ 1i50i2 1fbva4
1dmc__ 1e53a_ 1ezvb1 1jeqa1 1k3ea_ 1rec__ 1lm5a_ 1k82a1 1jaja_ 1m0ka_
1c0va_ 1kqfc_ 1ocrk_ 1h67a_ 2cpga_ 1ljra1 1brwa1 1hs7a_ 2cbla2 1jmxa2
1hyp__ 1cuk_2 1ecwa_ 1l9la_ 1g7da_ 1jkw_1 1dgna_ 1iqpa1 1pa2a_ 1ko9a1
1f1za1 1ks9a1 2sqca2 1d2ta_ 1h3la_ 1wer__ 1b3ua_ 1n1ba2 1poc__ 1e79i_
1m1qa_ 1enwa_ 1g4ma1 1e5ba_ 1qhoa2 1kv7a2 1l4ia2 1c8da_ 1amm_1 1ca1_2
1phm_2 1d7pm_ 1jjcb2 1flca1 1gr3a_ 1mjsa_ 1a8d_1 1lf6a2 1fqta_ 1jb0e_
1jh2a_ 1lcya1 1mgqa_ 1hcia1 1b3qa2 1jlxa1 1dar_1 1exma2 1ejea_ 1agja_
1e79d2 2rspa_ 1h0ha1 1gtra1 2erl__ 1btn__ 1lf7a_ 1jmxa5 1crua_ 1m1xa4
1hx0a1 1goia1 1ciy_2 1daba_ 3tdt__ 1gg3a1 1pmi__ 1bdo__ 1h3ia2 1gppa_
1f39a_ 1k6wa1 1jqna_ 1lu9a1 1m6ia1 1o94a3
Figure 3 Loops generated by FCCD (blue) that are close to real protein loops (green). The loops with lowest RMSD to a given loop of length 4 (top), 8 and 12 (bottom) are shown (loops 1qnr, A, 195–198, 3chb, D, 51–58 and 1ctq, A, 26–37). The N- terminus is at the left hand side.
It is clear that FCCD readily generates closed segments that are reasonably close to the real loops, with an average RMSD of about 0.6, 2.2 and 3.0 Å for loops of 4, 8 and 12 residues, respectively. The highest minimum RMSD values for these loop lengths are 0.76, 2.42 and 3.37 Å, respectively, indicating that FCCD in general can come up with a reasonably close conformation. Using more initial moving segments will obviously increase the chance of encountering a close conformation. Additionally, one can also expect an even better performance with a more refined way to constrain the (θ, τ) angles.
Conclusion
In this article, we introduce an algorithm that solves the loop closure problem for Cα only protein models. The method is conceptually similar to the CCD loop closure method introduced by Canutescu and Dunbrack [8], but optimizes dihedral and bond angles simultaneously, while the former method only optimizes one angle at a time. At the heart of the method lies a modified algorithm to superimpose point sets with minimum RMSD, based on singular value decomposition [20,21].
The algorithm is fast, numerically stable and leads to a solution for the great majority of loop closure problems studied here. Importantly, the method remains efficient even in the presence of constraints on the dihedral and bond angles. FCCD readily handles large gaps, and potentially generates realistic conformations. Compared to other loop closure methods, FCCD is surprisingly easy to implement provided a function is available to calculate the singular value decomposition of a matrix.
A possible disadvantage is that FCCD has a tendency to induce large changes to the pseudo angles at the start of the moving segment while angles near the end are less affected, which is also the case for CCD [8]. This can for example be avoided by selecting the pivot points in a random fashion, or by limiting the allowed change in the angles per iteration. Occasionally the method gets stuck, which can be avoided by incorporating stochastic changes away from the encountered local minimum. One can also simply try again with a new random moving segment. We believe that CCD and FCCD despite these disadvantages are among the most efficient loop closure algorithms currently available.
The FCCD algorithm proposed here has great potential for use in structure prediction methods that only make use of Cα atoms, or that otherwise do not include all backbone atoms [15,13,14]. FCCD could be used for example to implement local moves in a MCMC procedure. The moving segments could be derived from a fragment database or generated from a probabilistic model of the protein backbone. The latter model could range from a primitive probability distribution over allowed (θ, τ) angle pairs like we used here to a Hidden Markov Model that also models the sequence of (θ, τ) angle pairs.
We are planning to use the FCCD algorithm in combination with a sophisticated probabilistic model of the protein's backbone, which will steer both the generation of the initial moving loop and the acceptance/rejection of the angles. The performance of FCCD in this context will be the subject of a future publication.
Methods
Implementation
The FCCD algorithm was implemented in C, using the LAPACK [24] function dgesvd for the calculation of the singular value decomposition. Handling PDB files and calculating the (θ, τ) angles [16] was done using Biopython's Bio.PDB module [25]. We used a 2.5 GHz Pentium processor to calculate the benchmarks. A reference implementation of FCCD in Python is available as supplementary information.
Structure databases
For the calculation of the (θ, τ) probability distribution and the generation of random protein fragments, we used the SABMark 1.63 Twilight Zone database [26]. SABMark Twilight Zone contains 2230 high quality protein structures, divided over 236 different folds. All protein pairs have a BLAST E-value below 1, and thus presumably belong to different superfamilies. A dataset of fold representatives was generated by selecting a single structure at random for each fold (see Table 4).
The loops used to evaluate FCCD's sampling space were derived from Canutescu & Dunbrack [8]. We shifted two loops (1d8w, A, 46–57 and 1qop, A, 178–189) by three residues to ensure that all loops had three flanking residues on each side.
Calculation of the (θ, τ) probability distribution
The bond angle θ was subdivided in 18 bins and the dihedral angle τ in 36 bins, in both cases starting at 0 degrees and with a bin width of 10 degrees. All (θ, τ) angles were extracted from all structures in the SABMark Twilight Zone database that consisted of a polypeptide chain without breaks. In total, 257534 angle pairs were extracted. Each such (θ, τ) angle pair was assigned to a bin pair, and the number of angle pairs assigned to each bin pair was stored in a 18 × 36 count matrix. Finally, the normalized count matrix was used to assign a probability to any given (θ, τ) angle pair.
List of abbreviations
• CCD: Cyclic Coordinate Descent
• DOF: Degrees Of Freedom
• FCCD: Full Cyclic Coordinate Descent
• MCMC: Markov Chain Monte Carlo
• RMSD: Root Mean Square Deviation
Authors' contributions
TH conceived the FCCD algorithm. WB implemented FCCD in the C language, and introduced various refinements and optimizations. Both authors read and approved the article.
Supplementary Material
Additional File 1
The file FCCD.py contains an implementation of the FCCD algorithm. The program was implemented in the interpreted, object oriented language Python . The Numeric Python package , a Python module that implements many advanced mathematical operations efficiently in C and FORTRAN, provided implementations of singular value decomposition and various matrix operations. In addition, the Biopython toolkit, a set of Bioinformatics modules implemented in Python, was used to represent atomic coordinates as vector objects [25]. The core of the FCCD implementation comprises only 50 lines of Python code. Numeric Python and Biopython (version 1.4b) are needed to execute the sample code.
Click here for file
Acknowledgements
Wouter Boomsma is supported by the Lundbeckfond . Thomas Hamelryck is supported by a Marie Curie Intra-European Fellowship within the 6th European Community Framework Programme. We acknowledge encouragement and support from Prof. Anders Krogh, Bioinformatics Center, Institute of Molecular Biology and Physiology, University of Copenhagen.
==== Refs
Go N Scheraga H Ring closure and local conformational deformations of chain molecules Macromolecules 1970 3 178 187 10.1021/ma60014a012
Koehl P Delarue M A self consistent mean field approach to simultaneous gap closure and side-chain positioning in homology modelling Nat Struct Biol 1995 2 163 70 7538429 10.1038/nsb0295-163
da Silva R Degreve L Caliri A LMProt: an efficient algorithm for Monte Carlo sampling of protein conformational space Biophys J 2004 87 1567 77 15345537 10.1529/biophysj.104.041541
Jones T Thirup S Using known substructures in protein model building and crystallography EMBO J 1986 5 819 22 3709525
Rohl C Strauss C Chivian D Baker D Modeling structurally variable regions in homologous proteins with rosetta Proteins 2004 55 656 77 15103629 10.1002/prot.10629
Kolodny R Guibas L Levitt M Koehl P Inverse kinematics in biology: The protein loop closure problem Int J Robotics Research 2005 24 151 163 10.1177/0278364905050352
Manocha D Canny J Efficient inverse kinematics for general 6R manipulators IEEE Trans Rob Aut 1994 10 648 657 10.1109/70.326569
Canutescu A Dunbrack R Jr Cyclic coordinate descent: A robotics algorithm for protein loop closure Protein Sci 2003 12 963 72 12717019 10.1110/ps.0242703
Coutsias E Seok C Jacobson M Dill K A kinematic view of loop closure J Comput Chem 2004 25 510 28 14735570 10.1002/jcc.10416
Favrin G Irbäck A Sjunnesson F Monte carlo update for chain molecules: Biased Gaussian steps in torsional space J Chem Phys 2001 114 8154 8158 10.1063/1.1364637
Cahill M Cahill S Cahill K Proteins wriggle Biophys J 2002 82 2665 70 11964253
Singh R Bergert B Chaintweak: sampling from the neighbourhood of a protein conformation Pac Symp Biocomput 2005
Buchete N Straub J Thirumalai D Development of novel statistical potentials for protein fold recognition Curr Opin Struct Biol 2004 14 225 32 15093838 10.1016/j.sbi.2004.03.002
Tozzini V Coarse-grained models for proteins Curr Opin Struct Biol 2005 15 144 50 15837171 10.1016/j.sbi.2005.02.005
Kihara D Lu H Kolinski A Skolnick J TOUCHSTONE: an ab initio protein structure prediction method that uses threading-based tertiary restraints Proc Natl Acad Sci USA 2001 98 10125 30 11504922 10.1073/pnas.181328398
Oldfield T Hubbard R Analysis of C alpha geometry in protein structures Proteins 1994 18 324 37 8208725 10.1002/prot.340180404
Fidelis K Stern P Bacon D Moult J Comparison of systematic search and database methods for constructing segments of protein structure Protein Eng 1994 7 953 60 7809034
van Vlijmen H Karplus M PDB-based protein loop prediction: parameters for selection and methods for optimization J Mol Biol 1997 267 975 1001 9135125 10.1006/jmbi.1996.0857
Wang L Chen C A combined optimization method for solving the inverse kinematics problem of mechanical manipulators IEEE Trans Rob Aut 1991 7 489 499 10.1109/70.86079
Kabsch W A discussion of the solution for the best rotation to relate two sets of vectors Acta Cryst 1978 A34 827 828
Umeyama S Least squares estimation of transformation parameters between two point patterns IEEE Trans Pattern Anal Mach Intell 1991 13 376 80 10.1109/34.88573
Golub GH Loan CFV Matrix Computations 1996 3 Baltimore, Maryland: Johns Hopkins University Press
Bystroff C Thorsson V Baker D HMMSTR: a hidden Markov model for local sequence-structure correlations in proteins J Mol Biol 2000 301 173 90 10926500 10.1006/jmbi.2000.3837
Anderson E Bai Z Bischof C Demmel J Dongarra J Croz JD Greenbaum A Hammarling S McKenney A Ostrouchov S Sorensen D LAPACK's user's guide 1992 Philadelphia, PA, USA: Society for Industrial and Applied Mathematics
Hamelryck T Manderick B PDB file parser and structure class implemented in Python Bioinformatics 2003 19 2308 10 14630660 10.1093/bioinformatics/btg299
Van Walle I Lasters I Wyns L SABmark-a benchmark for sequence alignment that covers the entire known fold space Bioinformatics 2005 21 1267 8 15333456 10.1093/bioinformatics/bth493
|
15985178
|
PMC1192790
|
CC BY
|
2021-01-04 16:27:27
|
no
|
BMC Bioinformatics. 2005 Jun 28; 6:159
|
utf-8
|
BMC Bioinformatics
| 2,005 |
10.1186/1471-2105-6-159
|
oa_comm
|
==== Front
BMC BiotechnolBMC Biotechnology1472-6750BioMed Central London 1472-6750-5-201599627010.1186/1472-6750-5-20Methodology ArticleGenetic and spectrally distinct in vivo imaging: embryonic stem cells and mice with widespread expression of a monomeric red fluorescent protein Long Jonathan Z [email protected] Chantal S [email protected] Anna-Katerina [email protected] Developmental Biology Program, Sloan-Kettering Institute, New York, NY 10021, USA2005 4 7 2005 5 20 20 10 5 2005 4 7 2005 Copyright © 2005 Long et al; licensee BioMed Central Ltd.2005Long et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
DsRed the red fluorescent protein (RFP) isolated from Discosoma sp. coral holds much promise as a genetically and spectrally distinct alternative to green fluorescent protein (GFP) for application in mice. Widespread use of DsRed has been hampered by several issues resulting in the inability to establish and maintain lines of red fluorescent protein expressing embryonic stem cells and mice. This has been attributed to the non-viability, or toxicity, of the protein, probably as a result of its obligate tetramerization. A mutagenesis approach directing the stepwise evolution of DsRed has produced mRFP1, the first true monomer. mRFP1 currently represents an attractive autofluorescent reporter for use in heterologous systems.
Results
We have used embryonic stem cell-mediated transgenesis to evaluate mRFP1 in embryonic stem cells and mice. We find that mRFP1 exhibits the most spatially homogenous expression when compared to the native (tetrameric) and variant dimeric forms of DsRed. High levels of mRFP1 expression do not affect cell morphology, developmental potential or viability and fertility of animals. High levels of widespread mRFP1 expression are maintained in a constitutive manner in embryonic stem cells in culture and in transgenic animals. We have used various optical imaging modalities to visualize mRFP1 expressing cells in culture, in embryos and adult mice. Moreover co-visualization of red, green and cyan fluorescent cells within a sample is easily achieved without the need for specialized methodologies, such as spectral deconvolution or linear unmixing.
Conclusion
Fluorescent proteins with excitation and/or emission profiles in the red part of the visible spectrum represent distinct partners, or longer wavelength substitutes for GFP. Not only do DsRed-based RFPs provide a genetically and spectrally distinct addition to the available repertoire of autoflorescent proteins, but by virtue of their spectral properties they permit deeper tissue imaging. Our work in generating CAG::mRFP1 transgenic ES cells and mice demonstrates the developmental neutrality of mRFP1 in an organismal context. It paves the way for the use of DsRed-based monomeric RFPs in transgenic and gene targeted approaches which often necessitate spatially and/or temporally restricted reporter expression. Moreover animals of the CAG::mRFP1 transgenic strain serve as a source of RFP tagged tissue for the derivation of cell lines and explant, transplant and embryo chimera experiments.
==== Body
Background
GFP and many of its spectral variants have found widespread use as genetically encoded indicators in cell biological applications, including investigating protein expression, localization and interactions [1]. They have also been used for marking cells in complex tissues in chimeras, and for tagging gene expression in targeted or transgenic regimes in whole embryos, adult animals, explants and transplants [2,3]. In addition to the many cell biological benefits provided by increasing the number of genes encoding fluorescent protein species and extending the spectrum of available colors to longer (red) wavelengths of the spectrum, access to red fluorescent protein (RFP) variants would offer benefits for genetically and spectrally-distinct imaging of multiple cell populations in complex tissues [2,4].
All genetically-encoded coelenterate-derived fluorescent proteins cloned to date display some form of quaternary structure [5]. The weak tendency of Aequorea victoria green fluorescent protein (GFP) to dimerize has not impeded its use in exogenous systems. In contrast the obligate tetramerization of the most popular RFP DsRed [6], isolated from Discosoma sp., has impeded its widespread use in mice [3]. In addition to its obligate tetramerization, the evolution of DsRed to a generally applicable and robust tool has been hampered by several critical problems, including its slow and incomplete maturation [7].
An alternative approach to overcoming the shortcomings of DsRed has been to continue the search for DsRed homologues in sea corals and anemones [8-10]. This approach has yielded several novel red-shifted proteins, however the more fundamental problem of tetramerization still exists and needs to be overcome. Therefore it appears likely that a mutagenesis approach focused on a single tetrameric RFP and designed to sequentially render it dimeric and then monomeric while retaining fluorescence quantum yield and extinction coefficient, is key to achieving progress in the field [11].
Attempts, to mutagenize the DsRed coding sequence to improve the rate and or extent of maturation have resulted in several variants including DsRed2. Unfortunately DsRed2 has provided only modest improvements and has not effectively eradicated the apparent 'toxicity' of DsRed [12]. Additional engineered variants, including DsRed.T1, DsRed.T3 and dimer2 have superceded DsRed2, as they have faster more complete maturation [11,13]. However, since these proteins still form obligate dimers, the heterogeneity of protein localization (observed as 'clumps' within cells) has not been overcome [13,14].
Although many dimeric DsRed variants have found widespread use in other organisms including yeast, worms, fruitflies and zebrafish, several investigators, including ourselves, have not previously been successful in using them in mice [3]. Recently mice with widespread conditionally activated DsRed.T3 expression have been reported [15]. Interestingly though high-resolution imaging of embryonic stem cells and cells in chimeras that exhibit constitutive expression of DsRed.T3, or of related dimeric DsRed variants, reveals heterogeneous expression within cells, and in particular, punctate staining in a perinuclear region, possibly the Golgi (our unpublished observations). Although reduced in intensity, this subcellular localization is reminiscent of that observed with tetrameric DsRed variants. It is therefore probable that dimeric DsRed variants are not the optimal starting point for RFPs used in mouse transgenic and gene targeting regimes.
Moreover the available dimeric DsRed variants have not resolved the issue of incorporating RFPs into protein fusions that can, for example, be used for in vivo imaging at subcellular resolution [12], therefore necessitating the development of monomeric RFPs which may provide improved performance and yield developmental neutrality when incorporated into fusions.
The first true monomer designated mRFP1 (monomeric RFP 1), was reported a few years ago [11]. The stepwise evolution of the DsRed sequence to generate mRFP1 involved mutations that first increased the speed of maturation, then mutations which resulted in the breaking of each subunit interface, and then a further round of mutations resulting in restoration of fluorescence. This directed evolution of an RFP to a monomer resulted in the introduction of a total of 33 amino acid substitutions [11]. mRFP1 has been reported to participate in, but not impair, the function of various fusion proteins in cases where both the tetrameric and dimeric DsRed variants were unable to do so. This makes mRFP1 attractive for use in mice both in its native form and as part of a fusion protein.
Furthermore, mRFP1 has recently served as the template for further mutagenesis directed at improving its performance in N-terminal protein fusions, fluorescence quantum yield and extinction coefficient, photostability and/or in providing a variety spectral variants [16]. This has resulted in the next generation of monomers, which include a battery of spectrally-distinct DsRed variants spanning the spectrum from green (mHoneydew), through yellow (mBanana) to a range of reds (mOrange, mStrawberry, mCherry). Therefore monomeric RFPs based on DsRed currently represent the most attractive genetically-encoded red fluorescent proteins for use in mice. However, the developmental neutrality and fluorescent intensity of monomeric RFPs has not been investigated in vivo in transgenic or gene targeting applications mice.
Results
We have investigated the expression and germline transmission of native mRFP1 in embryonic stem cells and mice. This is an essential prerequisite to using the available suite of DsRed-derived monomeric RFPs in fusions as reporters for high-resolution in vivo imaging, and in particular, as fusion proteins exhibiting subcellular localization and acting as segmental markers of 3-dimensional space.
As a first step we cloned the mRFP1 gene into a vector permitting widespread expression in a variety of cells types in culture and in vivo in mice. The mRFP1 coding sequence was engineered to contain a Kozak consensus sequence at its 5' end and subsequently introduced into pCAGGS a vector utilizing the chicken beta actin promoter and SV40 immediate early enhancer combination, designed to drive high-level constitutive gene expression in ES cells, embryos and adult mice [17,18]. Standard protocols were used to establish stable CAG::mRFP1 transgenic lines of ES cells constitutively expressing mRFP1.
Fluorescent colonies (CAG::mRFP1 transgenic) were identified and picked under an epifluorescence stereo dissecting microscope. Clones were passaged in 96-well plates, and scored for the maintenance and extent of red fluorescence with maintenance in culture. Clones that failed to meet these criteria were discarded from further analysis. Thereafter only clones exhibiting robust transgene expression in vitro under both stem cell and differentiation conditions were further analyzed for level and heterogeneity of red fluorescence by flow cytometry as described previously for GFP-variants [19]. Only those clones exhibiting high levels of homogenous expression were selected for further analysis. Sustained strong homogenous red fluorescence and retention of normal ES cell morphology when grown on either gelatin-coated plates or mouse embryo fibroblast feeders was a prerequisite feature of these cells
We then co-cultured three transgenic ES cell lines expressing different fluorescent proteins, namely; CAG::ECFP, CAG::EGFP and CAG::mRFP1 along with wild type (non-fluorescent) ES cells, and then visualized fluorescence in resulting chimeric colonies (Fig. 1). The different fluorophores could easily be distinguished both with standard filter sets using epifluorescence optics and confocal microscopy, without the need for special image processing such as spectral deconvolution or linear unmixing [20]. We also noted that comparable levels of fluorescence were produced as all three color variants produced fluorescent signal within the same dynamic range.
Figure 1 Co-visualization of multiple fluorescent proteins in mouse embryonic stem (ES) cells. Mixed colony of embryonic stem (ES) cells comprised of wild type (untagged) cells, CAG::ECFP transgenic cells exhibiting widespread expression of ECFP, CAG::EGFP transgenic cells exhibiting widespread expression of EGFP and CAG::mRFP1 transgenic cells exhibiting widespread expression of mRFP1. Bright field image (A), CFP channel (B), GFP channel (C), RFP channel (D), merge of all three fluorescent channels overlayed on the bright field image (E), merge of all three fluorescent channels (F), color-coded depth projection of the three fluorescent channel merge with the color-coded scale shown on the right of the image (G). A second mixed transgenic ES cell colony with bright field image (H), merge of three fluorescent channel merge overlayed on the bright field image (I) and dark field three fluorescent channel merge (J). In all panels except G ECFP fluorescence is shown in blue, EGFP fluorescence is in green and mRFP1 fluorescence is in red. bf, bright field; df, dark field.
Having established the neutrality of widespread mRFP1 expression in ES cells, we went on to use ES cell mediated transgenesis through the generation of germline transmitting chimeras to introduce the CAG::mRFP1 transgene into mice. As a first step to test the extent of expression of the transgene in embryos, we generated 4n (tetraploid) wild type <-> mRFP1 ES cell derived chimeras [21], exactly as described previously [3]. Resulting, completely ES cell-derived embryos exhibited widespread mRFP1 expression, indicating that the level of expression was sufficiently strong to be visualized, that the transgene was not silenced and that development was able to proceed normally to midgestation (data not shown). To produce germline transmitting chimeric adult animals, we next generated diploid wild type <-> mRFP1 ES cell chimeras that were allowed to go to term. The CAG::mRFP1 transgene was transmitted to F1 offspring in a Mendelian fashion, suggesting that widespread mRFP1 expressing is compatible with normal development and fertility. Two ES cell lines were taken germline and shown to produce an equivalent intensity and range of fluorescence, therefore data from only one line are shown.
Wide field epifluorescent and laser scanning confocal microscopy was used to image this constitutively expressed transgene reporter in preimplantation stage mouse embryos hemizygous (Tg/+) for the CAG::mRFP1 transgene (Fig. 2 and additional files). Such non-invasive visualization in living preparations allowed us to acquire high-magnification, sequential optical sections (z-stacks) that were used to generate high-resolution anatomical, volumetric images of embryos. To do this, stacks of sequential optical sections were computationally reconstructed into 3-dimensional (3D) projections. This methodology was used to generate 3D image sets, and is illustrated here by imaging whole mouse embryos at the 1-cell stage and blastocyst stage. These data sets can be computationally manipulated in various ways including for the visualization of individual xy slices from a z-stack (Fig. 2B,C,K and 2L and additional files), or of rendered images from the full (Fig. 2D), or partial z-stack (Fig. 2E and 2M). It should be noted that even using epifluorescence imaging CAG::mRFP1 embryos were clearly distinguished from stage-matched CAG::EGFP embryos (Fig. 2G–I).
Figure 2 RFP expression in CAG::mRFP1 preimplantation stage embryos. Single CAG::mRFP1 Tg/+ zygote including the second polar body (A–E). A single x-y section taken from a z-stack, bright field image (A), overlay single confocal section of red fluorescence and bright field (B), red fluorescence channel only (left panel in C). Representative x-y sections taken from the same z-stack that was used to render volumes shown D and E. Rendered z-stack (3D reconstruction) of the whole zygote and second polar body (PB) shown in the previous panels and rotated through 180 degrees counter-clockwise (D). Rendered z-stack (3D reconstruction) of a computationally bisected zygote shown in the previous panels and rotated through 180 degrees counter-clockwise (E). Note that the zygote is not spherical, it has a clear short and long axis (lines with arrows), and the fertilization cone resulting from the site of sperm entry is also clearly evident as a protrusion (asterix). Non-transgenic, CAG::EGFP Tg/+, CAG::mRFP1 Tg/+ embryos recovered at E3.0 and representing compacted morulae through to blastocyst stages (F–I). Bright filed (F), red fluorescence channel (G), green fluorescence channel (H) and green and red fluorescence channel overlay (I). Single CAG::mRFP1 Tg/+ blastocyst (J–M). A single x-y section taken from a z-stack, bright field image (J), overlay single confocal section of red fluorescence and bright field (K), red fluorescence channel only (left panel in L). Representative x-y sections taken from the same z-stack that was used to render the volume shown in M. Rendered z-stack (3D reconstruction) of a computationally bisected blastocyst and rotated through 180 degrees counter-clockwise (M). Note that individual cells of the trophectoderm can be distinguished. The RGB colored vector on the bottom left of the 3D reconstruction rotations depicts the x-axis in green, y-axis in red and z-axis in blue.
Wide field microscopic imaging of later stage hemizygous embryos illustrated the robust, homogenous and widespread expression of mRFP1 from early postimplantation, embryonic day (E) 6.5 (Fig. 3A) to later fetal stages in both the embryo proper and extraembryonic lineages, including the placenta (Fig. 3E–F). Dissection of organs from fetuses confirmed widespread red fluorescence in CAG::mRFP1/+ organs contrasted with a lack of signal observed in non-transgenic littermates (Fig. 3G and 3H).
Figure 3 RFP expression in CAG::mRFP1 postimplantation embryos. Bright field and dark field epifluorescent images of CAG::mRFP1 Tg/+ embryos and non-transgenic littermates at E6.5 (A), E7.75 (B), E8.75 (C). CAG::mRFP1 Tg/+ embryos, CAG::EGFP Tg/+ embryos and non-transgenic littermates at E11.5 (D). E15.5 CAG::mRFP1 Tg/+ fetuses and non-transgenic littermates at (E) demonstrating widespread homogenous red fluorescence throughout later development in whole embryos, dissected embryonic tissues (G and H) and extraembryonic tissues including the placenta (F). (H), cardiothoracic organs from three embryos of different ages, E13.5 (left), E14.5 (center) and E15.5 (right), only two of which are hemizygous for the transgene). Ad, adrenal gland; bl, bladder; h, heart; ki, kidney; lu, lung; te, testis; th, thymus.
Examination of newborn animals revealed strong widespread expression of mRFP1 in the skin of hemizygous CAG::mRFP/+ transgenics and demonstrated that this red fluorophore can be distinguished from green fluorescence observed in CAG::EGFP/+ transgenics, and from non-transgenic littermates by standard macroscopic visualization using standard filter sets and epiflourescent excitation (Fig. 4).
Figure 4 Transgenic CAG::mRFP1 mice can be distinguished from CAG::EGFP animals. Macroscopic images of non-transgenic, CAG::mRFP1 Tg/+ and CAG::EGFP Tg/+ newborn (P5) mouse pups demonstrating that red fluorescence can clearly be distinguished from green fluorescence using conventional epifluorescent illumination and macroscopic observation. Dorsal view (A), and ventral view (B), high magnification of tails of 3 month old animals (C). Inspection of fluorescence in the tails is the method used for routine genotyping of these strains.
Further analysis of various adult organs, including the peritoneum, heart, lung, eye, brain, liver, pancreas, spleen and kidney, that were freshly obtained from CAG::mRFP1/+ adult animals revealed robust and widespread fluorescence (Fig. 5), as has been reported for animals expressing GFP-based fluorescent proteins under the regulation of the CAG promoter. We also noted that newborn pups, or bald skin and tissues expressing mRFP1, exhibit a pink coloration under normal light when compared to non-transgenic or CAG::EGFP transgenics [18]. This is particularly evident in albino animals (tails in Fig. 4C), and in unpigmented organs such as the brain and pancreas (Fig. 5E and 5G). We believe this results from mRFP1 being a red fluorphore with a spectrum closer to the visible range, so that it can be visualized as a pink pigmentation under daylight or bright field conditions. Consequently, newborn CAG::mRFP1 pups can be genotyped based on their pink pigmentation, alleviating the need to image them under epifluorescent conditions.
Figure 5 Widespread RFP expression in adult organs. Panels of bright field and corresponding dark field epifluorescent images of organs taken from a 4 week old CAG::mRFP1 Tg/+ mouse and a non-transgenic littermate. Peritoneum (A), heart (B), lung (C), eye (D), brain (E), liver (F), pancreas (G), spleen (H) and kidney (I). In addition to the fluorescence observed under epifluorescent illumination, RFP expressing tissues exhibit a pink color under bright field illumination. This is particularly evident in panels A, B, C, E and G. This allows for genotyping of newborn pups (by virtue of their pink color) in the absence of fluorescence illumination.
Moreover we succeeded in breeding the CAG::mRFP1 transgene to homozygosity. Homozygous animals retained developmental potential and therefore colonies of the strain are routinely maintained in the homozygous state. To date, we have observed no noticeable reduction in fluorescence or fertility for over four generations on both ICR outbred and 129 inbred backgrounds. From routine daily observations no behavioral differences are distinguished between CAG::mRFP1 animals and wild type age matched animals, though no behavioral tests have been implemented.
Discussion
We have established the applicability of monomeric RFPs as an alternative spectrally-distinct genetically-encoded fluorescent reporter to GFP and its variants. Now monomeric DsRed variants can be used in vivo in mice along side GFP variants to differentially label single cells or different populations of cells.
Advances in optical imaging modalities, and the development of ex vivo embryo or explant culture systems, have evolved alongside the increasing availability of genetically-encoded fluorescent proteins permitting dynamic in vivo imaging of living specimens. Two major limitations in the application of genetically-encoded fluorescent protein technology in mice, have been the lack of fluorescent proteins in the longer wavelength (red) part of the spectrum, and the absence of a fluorescent protein that is genetically distinguishable from GFP. DsRed, the coelenterate derived RFP, has offered the potential to solve both these problems, but original variants, which function as obligate tetramers and dimers, have exhibited 'toxicity', limiting their use in heterologous systems.
Our work demonstrates the developmental neutrality of mRFP1, the first true monomeric RFP, in ES cells and mice. The strain of mice we have generated exhibits widespread mRFP1 expression, providing a novel reagent for live imaging of red fluorescent cells in a genetically tractable mammalian model organism. CAG::mRFP1 animals represent a resource for analyzing development and disease in mouse embryos and adults. They can also be used as tagged populations of cells in chimeras, in addition to transplantation and cell isolation experiments. Moreover RFP cells can easily be distinguished from GFP, or GFP spectral variant expressing cells (including ECFP and EYFP) by both imaging using standard optics, and molecular analyses, due to the different genetic origin.
To date all fluorescent proteins isolated from diverse anthozoan species suffer from obligate tetramerization and will require efforts similar to the evolution of mRFP1 to produce widely useful tools. Our demonstration of the developmental neutrality of widespread mRFP1 expression paves the way for the incorporation of monomeric RFPs into mouse transgenic and gene targeting approaches [2]. Our data on mRFP1 suggest that its mutagenized derivatives will also be amenable to use in mice. If so, there should now be at least six new spectral variant monomeric FPs based on DsRed that can be used in mice. Key applications of the proliferation of spectral variant FPs will include discriminating different cells, transcriptional activities and/or fusion proteins.
Conclusion
Our work demonstrates the use of monomeric RFPs in mice and provides the basis for future efforts designed to incorporate RFPs into protein fusions that can be used as spectrally-distinct subcellularly-localized tags permitting high-resolution live imaging in vivo [12,22]. The ongoing development of mouse strains expressing subcellularly-localized protein fusions incorporating RFPs, and contrasting with GFP-variants will be essential for visualizing 4-dimensional (3-dimensions over time) anatomy and tracking cell position, morphology and behavior in vivo.
Methods
Vector construction
The mRFP1 coding sequence was amplified by PCR using primers 5'RFP (5'-cgtagaattcgccaccaatggctagcatgactgg) and 3'RFP (5'-gcacgaattcgggcgccggtggagtggcggcc) using Pfx Polymerase (Invitrogen). The resulting product was cloned into the EcoRI site of pCAGGS to generate pCX-mRFP1. Transient transfection of Cos-7 cells using Fugene 6 Transfection Reagent as per manufacturer's recommendations (Roche) was used to evaluate pCX-mRFP1, and verify that it produced robust red fluorescence.
Generating transgenic ES cells
Transgenic ES cell lines constitutively expressing mRFP1 were generated by co-electroporation of SalI linearized pCX-mRFP1 construct and a circular PGK-Puro-pA plasmid conferring transient puromycin resistance. Puromycin selection was carried out exactly as described previously [18]. Thereafter cells were passaged according to standard protocols.
Mouse breeding
Two lines of CAG::mRFP1 ES cells were used for chimera generation by injection into C57BL/6 blastocysts using standard procedures [23]. Chimeras were mated to outbred ICR and inbred 129/Tac mice (Taconic, Germantown, NY) for germline transmission and subsequent maintenance of the lines. After germline transmission, both transgenes were bred to, and maintained at, homozygosity, suggesting that the sites of transgene integration were not perturbing essential gene function and that high levels of protein expression were non-toxic. Animals from both lines retained widespread homogenous fluorescence in all subsequent generations tested and therefore data was pooled.
Embryo collection
Preimplantation embryos were recovered in M2 media and subsequently cultured under oil in a tissue culture incubator gased at 5% CO2 in KSOM media. Postimplantation embryos and organs were dissected either in HEPES buffered DMEM containing 10% fetal calf serum or PBS containing 0.1% BSA, and cultured in media comprising 50% rat serum, 50% DMEM/F12 supplemented with L-glutamine.
Image acquisition
Also although mRFP1 has been shown to have a lower extinction coefficient, quantum yield, and photostability than native DsRed or DsRed.T3, mRFP1 has been shown to mature over ten fold faster, so that it shows similar brightness in living cells. In addition, the excitation and emission peaks of mRFP1 (584 and 607 nm respectively) are 25 nm red-shifted from DsRed, which should confer greater tissue penetration and spectral separation from autofluorescence and other fluorescent proteins. All images presented in the figures are of living embryos or freshly dissected (unfixed) tissues maintained under physiological conditions. Wide-field images were acquired with an AxioCam MRc camera attached to a Leica MZ16FA stereo-dissecting microscope or a Zeiss Axiovert 200M inverted microscope equipped with epifluorescent illumination using appropriate filter sets (Chroma). Laser scanning confocal data was acquired in single-track mode using a Zeiss LSM510 META on a Zeiss Axiovert 200M. Fluorophores were excited with a 405 nm diode laser (ECFP), 488 nm Argon laser line (EGFP) and a 543 nm HeNe laser (mRFP1). Objectives used were a C-apochromat 40×/NA1.2, plan-apochromat 20×/0.75 and a fluar 5×/0.25. Confocal images were acquired as z-stacks comprising sequential optical x-y sections taken at 1–2 μm z-intervals.
Image processing
Raw data was processed using Zeiss AIM software (Carl Zeiss Microsystems at ), and Volocity (Improvision at ). Re-animation of data to generate movies of time-lapses or rotations was performed using QuickTime Pro (Apple Computer, Inc at ).
Authors' contributions
JZL and CSL carried out the preimplantation embryo experiments including all imaging and the generation of the aggregation chimeras. AKH planned the study, carried out the experiments in ES cells and postimplantation embryos, and wrote the manuscript.
Footnote
The CAG::mRFP1 mice described in this paper will be made available from the Jackson Laboratories Induced Mutant Resource () as stock number #5645.
Supplementary Material
Additional File 1
Sequential x-y images of a z-stack taken through a CAG::mRFP1 Tg/+ zygote (1-cell stage). The first part of the sequence includes bright field (DIC) images merged with the red fluorescence channel, while the second part depicts only the red fluorescence channel.
Click here for file
Additional File 2
3D reconstruction and rotation of the complete raw data set shown in Movie 1 depicting a view of a full zygote. The sequence depicts the rendered image being rotated around each axes (z, y and z). The RGB colored vector on the bottom left depicts the x-axis in green, y-axis in red and z-axis in blue.
Click here for file
Additional File 3
3D reconstruction and rotation of half the raw data shown in Movie 1 depicting a view of a computationally bisected zygote. The sequence depicts the rendered image being rotated around each axes (z, y and z). The RGB colored vector on the bottom left depicts the x-axis in green, y-axis in red and z-axis in blue.
Click here for file
Additional File 4
Sequential x-y images of a z-stack taken through a CAG::mRFP1 Tg/+ blastocyst stage embryo. The first part of the sequence includes bright field (DIC) images merged with the red fluorescence, while the second part depicts only the red fluorescence channel.
Click here for file
Additional File 5
3D reconstruction and rotation of the complete raw data set shown in Movie 4 depicting a view of half a blastocyst. The sequence depicts the rendered image being rotated around each axes (z, y and z). The RGB colored vector on the bottom left depicts the x-axis in green, y-axis in red and z-axis in blue.
Click here for file
Acknowledgements
We thank J.-I. Miyazaki and R. Tsien for the CAGGS and mRFP1 plasmids respectively, the Memorial Sloan-Kettering Cancer Center (MSKCC) Transgenic Core Facility for production of germline transmitting chimeras, and E. Lacy for comments on the manuscript. This work was supported by laboratory start-up funds from the MSKCC to AKH.
==== Refs
Tsien RY The green fluorescent protein Annu Rev Biochem 1998 67 509 544 9759496 10.1146/annurev.biochem.67.1.509
Hadjantonakis AK Dickinson ME Fraser SE Papaioannou VE Technicolour transgenics: imaging tools for functional genomics in the mouse Nat Rev Genet 2003 4 613 625 12897773 10.1038/nrg1126
Hadjantonakis AK Macmaster S Nagy A Embryonic stem cells and mice expressing different GFP variants for multiple non-invasive reporter usage within a single animal BMC Biotechnol 2002 2 11 12079497 10.1186/1472-6750-2-11
Zhang J Campbell RE Ting AY Tsien RY Creating new fluorescent probes for cell biology Nat Rev Mol Cell Biol 2002 3 906 918 12461557 10.1038/nrm976
Verkhusha VV Lukyanov KA The molecular properties and applications of Anthozoa fluorescent proteins and chromoproteins Nat Biotechnol 2004 22 289 296 14990950 10.1038/nbt943
Matz MV Fradkov AF Labas YA Savitsky AP Zaraisky AG Markelov ML Lukyanov SA Fluorescent proteins from nonbioluminescent Anthozoa species Nat Biotechnol 1999 17 969 973 10504696 10.1038/13657
Gross LA Baird GS Hoffman RC Baldridge KK Tsien RY The structure of the chromophore within DsRed, a red fluorescent protein from coral Proc Natl Acad Sci U S A 2000 97 11990 11995 11050230 10.1073/pnas.97.22.11990
Gurskaya NG Fradkov AF Terskikh A Matz MV Labas YA Martynov VI Yanushevich YG Lukyanov KA Lukyanov SA GFP-like chromoproteins as a source of far-red fluorescent proteins FEBS Lett 2001 507 16 20 11682051 10.1016/S0014-5793(01)02930-1
Wiedenmann J Schenk A Rocker C Girod A Spindler KD Nienhaus GU A far-red fluorescent protein with fast maturation and reduced oligomerization tendency from Entacmaea quadricolor (Anthozoa, Actinaria) Proc Natl Acad Sci U S A 2002 99 11646 11651 12185250 10.1073/pnas.182157199
Miyawaki A Green fluorescent protein-like proteins in reef Anthozoa animals Cell Struct Funct 2002 27 343 347 12502888 10.1247/csf.27.343
Campbell RE Tour O Palmer AE Steinbach PA Baird GS Zacharias DA Tsien RY A monomeric red fluorescent protein Proc Natl Acad Sci U S A 2002 99 7877 7882 12060735 10.1073/pnas.082243699
Hadjantonakis AK Papaioannou VE Dynamic in vivo imaging and cell tracking using a histone fluorescent protein fusion in mice BMC Biotechnol 2004 4 33 15619330 10.1186/1472-6750-4-33
Bevis BJ Glick BS Rapidly maturing variants of the Discosoma red fluorescent protein (DsRed) Nat Biotechnol 2002 20 83 87 11753367 10.1038/nbt0102-83
Baird GS Zacharias DA Tsien RY Biochemistry, mutagenesis, and oligomerization of DsRed, a red fluorescent protein from coral Proc Natl Acad Sci U S A 2000 97 11984 11989 11050229 10.1073/pnas.97.22.11984
Vintersten K Monetti C Gertsenstein M Zhang P Laszlo L Biechele S Nagy A Mouse in red: red fluorescent protein expression in mouse ES cells, embryos, and adult animals Genesis 2004 40 241 246 15593332 10.1002/gene.20095
Shaner NC Campbell RE Steinbach PA Giepmans BN Palmer AE Tsien RY Improved monomeric red, orange and yellow fluorescent proteins derived from Discosoma sp. red fluorescent protein Nat Biotechnol 2004 22 1567 1572 15558047 10.1038/nbt1037
Niwa H Yamamura K Miyazaki J Efficient selection for high-expression transfectants with a novel eukaryotic vector Gene 1991 108 193 199 1660837 10.1016/0378-1119(91)90434-D
Hadjantonakis AK Gertsenstein M Ikawa M Okabe M Nagy A Generating green fluorescent mice by germline transmission of green fluorescent ES cells Mech Dev 1998 76 79 90 9867352 10.1016/S0925-4773(98)00093-8
Hadjantonakis AK Nagy A FACS for the isolation of individual cells from transgenic mice harboring a fluorescent protein reporter Genesis 2000 27 95 98 10951501 10.1002/1526-968X(200007)27:3<95::AID-GENE10>3.0.CO;2-A
Dickinson ME Bearman G Tille S Lansford R Fraser SE Multi-spectral imaging and linear unmixing add a whole new dimension to laser scanning fluorescence microscopy Biotechniques 2001 31 1272, 1274 6, 1278 11768655
Nagy A Gocza E Diaz EM Prideaux V Ivanyi E Markkula M Rossant J Embryonic stem cells alone are able to support fetal development in the mouse. Development 1990 110 815 821 2088722
Plusa B Hadjantonakis AK Gray D Piotrowska-Nitsche K Jedrusik A Papaioannou VE Glover DM Zernicka-Goetz M The first cleavage of the mouse zygote predicts the blastocyst axis Nature 2005 434 391 395 15772664 10.1038/nature03388
Nagy A Gertsenstein M Vintersten K Behringer R Manipulating the mouse embryo: A laboratory manual 2002 , Cold Spring Harbor Press
|
15996270
|
PMC1192791
|
CC BY
|
2021-01-04 16:38:25
|
no
|
BMC Biotechnol. 2005 Jul 4; 5:20
|
utf-8
|
BMC Biotechnol
| 2,005 |
10.1186/1472-6750-5-20
|
oa_comm
|
==== Front
BMC BiotechnolBMC Biotechnology1472-6750BioMed Central London 1472-6750-5-201599627010.1186/1472-6750-5-20Methodology ArticleGenetic and spectrally distinct in vivo imaging: embryonic stem cells and mice with widespread expression of a monomeric red fluorescent protein Long Jonathan Z [email protected] Chantal S [email protected] Anna-Katerina [email protected] Developmental Biology Program, Sloan-Kettering Institute, New York, NY 10021, USA2005 4 7 2005 5 20 20 10 5 2005 4 7 2005 Copyright © 2005 Long et al; licensee BioMed Central Ltd.2005Long et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
DsRed the red fluorescent protein (RFP) isolated from Discosoma sp. coral holds much promise as a genetically and spectrally distinct alternative to green fluorescent protein (GFP) for application in mice. Widespread use of DsRed has been hampered by several issues resulting in the inability to establish and maintain lines of red fluorescent protein expressing embryonic stem cells and mice. This has been attributed to the non-viability, or toxicity, of the protein, probably as a result of its obligate tetramerization. A mutagenesis approach directing the stepwise evolution of DsRed has produced mRFP1, the first true monomer. mRFP1 currently represents an attractive autofluorescent reporter for use in heterologous systems.
Results
We have used embryonic stem cell-mediated transgenesis to evaluate mRFP1 in embryonic stem cells and mice. We find that mRFP1 exhibits the most spatially homogenous expression when compared to the native (tetrameric) and variant dimeric forms of DsRed. High levels of mRFP1 expression do not affect cell morphology, developmental potential or viability and fertility of animals. High levels of widespread mRFP1 expression are maintained in a constitutive manner in embryonic stem cells in culture and in transgenic animals. We have used various optical imaging modalities to visualize mRFP1 expressing cells in culture, in embryos and adult mice. Moreover co-visualization of red, green and cyan fluorescent cells within a sample is easily achieved without the need for specialized methodologies, such as spectral deconvolution or linear unmixing.
Conclusion
Fluorescent proteins with excitation and/or emission profiles in the red part of the visible spectrum represent distinct partners, or longer wavelength substitutes for GFP. Not only do DsRed-based RFPs provide a genetically and spectrally distinct addition to the available repertoire of autoflorescent proteins, but by virtue of their spectral properties they permit deeper tissue imaging. Our work in generating CAG::mRFP1 transgenic ES cells and mice demonstrates the developmental neutrality of mRFP1 in an organismal context. It paves the way for the use of DsRed-based monomeric RFPs in transgenic and gene targeted approaches which often necessitate spatially and/or temporally restricted reporter expression. Moreover animals of the CAG::mRFP1 transgenic strain serve as a source of RFP tagged tissue for the derivation of cell lines and explant, transplant and embryo chimera experiments.
==== Body
Background
GFP and many of its spectral variants have found widespread use as genetically encoded indicators in cell biological applications, including investigating protein expression, localization and interactions [1]. They have also been used for marking cells in complex tissues in chimeras, and for tagging gene expression in targeted or transgenic regimes in whole embryos, adult animals, explants and transplants [2,3]. In addition to the many cell biological benefits provided by increasing the number of genes encoding fluorescent protein species and extending the spectrum of available colors to longer (red) wavelengths of the spectrum, access to red fluorescent protein (RFP) variants would offer benefits for genetically and spectrally-distinct imaging of multiple cell populations in complex tissues [2,4].
All genetically-encoded coelenterate-derived fluorescent proteins cloned to date display some form of quaternary structure [5]. The weak tendency of Aequorea victoria green fluorescent protein (GFP) to dimerize has not impeded its use in exogenous systems. In contrast the obligate tetramerization of the most popular RFP DsRed [6], isolated from Discosoma sp., has impeded its widespread use in mice [3]. In addition to its obligate tetramerization, the evolution of DsRed to a generally applicable and robust tool has been hampered by several critical problems, including its slow and incomplete maturation [7].
An alternative approach to overcoming the shortcomings of DsRed has been to continue the search for DsRed homologues in sea corals and anemones [8-10]. This approach has yielded several novel red-shifted proteins, however the more fundamental problem of tetramerization still exists and needs to be overcome. Therefore it appears likely that a mutagenesis approach focused on a single tetrameric RFP and designed to sequentially render it dimeric and then monomeric while retaining fluorescence quantum yield and extinction coefficient, is key to achieving progress in the field [11].
Attempts, to mutagenize the DsRed coding sequence to improve the rate and or extent of maturation have resulted in several variants including DsRed2. Unfortunately DsRed2 has provided only modest improvements and has not effectively eradicated the apparent 'toxicity' of DsRed [12]. Additional engineered variants, including DsRed.T1, DsRed.T3 and dimer2 have superceded DsRed2, as they have faster more complete maturation [11,13]. However, since these proteins still form obligate dimers, the heterogeneity of protein localization (observed as 'clumps' within cells) has not been overcome [13,14].
Although many dimeric DsRed variants have found widespread use in other organisms including yeast, worms, fruitflies and zebrafish, several investigators, including ourselves, have not previously been successful in using them in mice [3]. Recently mice with widespread conditionally activated DsRed.T3 expression have been reported [15]. Interestingly though high-resolution imaging of embryonic stem cells and cells in chimeras that exhibit constitutive expression of DsRed.T3, or of related dimeric DsRed variants, reveals heterogeneous expression within cells, and in particular, punctate staining in a perinuclear region, possibly the Golgi (our unpublished observations). Although reduced in intensity, this subcellular localization is reminiscent of that observed with tetrameric DsRed variants. It is therefore probable that dimeric DsRed variants are not the optimal starting point for RFPs used in mouse transgenic and gene targeting regimes.
Moreover the available dimeric DsRed variants have not resolved the issue of incorporating RFPs into protein fusions that can, for example, be used for in vivo imaging at subcellular resolution [12], therefore necessitating the development of monomeric RFPs which may provide improved performance and yield developmental neutrality when incorporated into fusions.
The first true monomer designated mRFP1 (monomeric RFP 1), was reported a few years ago [11]. The stepwise evolution of the DsRed sequence to generate mRFP1 involved mutations that first increased the speed of maturation, then mutations which resulted in the breaking of each subunit interface, and then a further round of mutations resulting in restoration of fluorescence. This directed evolution of an RFP to a monomer resulted in the introduction of a total of 33 amino acid substitutions [11]. mRFP1 has been reported to participate in, but not impair, the function of various fusion proteins in cases where both the tetrameric and dimeric DsRed variants were unable to do so. This makes mRFP1 attractive for use in mice both in its native form and as part of a fusion protein.
Furthermore, mRFP1 has recently served as the template for further mutagenesis directed at improving its performance in N-terminal protein fusions, fluorescence quantum yield and extinction coefficient, photostability and/or in providing a variety spectral variants [16]. This has resulted in the next generation of monomers, which include a battery of spectrally-distinct DsRed variants spanning the spectrum from green (mHoneydew), through yellow (mBanana) to a range of reds (mOrange, mStrawberry, mCherry). Therefore monomeric RFPs based on DsRed currently represent the most attractive genetically-encoded red fluorescent proteins for use in mice. However, the developmental neutrality and fluorescent intensity of monomeric RFPs has not been investigated in vivo in transgenic or gene targeting applications mice.
Results
We have investigated the expression and germline transmission of native mRFP1 in embryonic stem cells and mice. This is an essential prerequisite to using the available suite of DsRed-derived monomeric RFPs in fusions as reporters for high-resolution in vivo imaging, and in particular, as fusion proteins exhibiting subcellular localization and acting as segmental markers of 3-dimensional space.
As a first step we cloned the mRFP1 gene into a vector permitting widespread expression in a variety of cells types in culture and in vivo in mice. The mRFP1 coding sequence was engineered to contain a Kozak consensus sequence at its 5' end and subsequently introduced into pCAGGS a vector utilizing the chicken beta actin promoter and SV40 immediate early enhancer combination, designed to drive high-level constitutive gene expression in ES cells, embryos and adult mice [17,18]. Standard protocols were used to establish stable CAG::mRFP1 transgenic lines of ES cells constitutively expressing mRFP1.
Fluorescent colonies (CAG::mRFP1 transgenic) were identified and picked under an epifluorescence stereo dissecting microscope. Clones were passaged in 96-well plates, and scored for the maintenance and extent of red fluorescence with maintenance in culture. Clones that failed to meet these criteria were discarded from further analysis. Thereafter only clones exhibiting robust transgene expression in vitro under both stem cell and differentiation conditions were further analyzed for level and heterogeneity of red fluorescence by flow cytometry as described previously for GFP-variants [19]. Only those clones exhibiting high levels of homogenous expression were selected for further analysis. Sustained strong homogenous red fluorescence and retention of normal ES cell morphology when grown on either gelatin-coated plates or mouse embryo fibroblast feeders was a prerequisite feature of these cells
We then co-cultured three transgenic ES cell lines expressing different fluorescent proteins, namely; CAG::ECFP, CAG::EGFP and CAG::mRFP1 along with wild type (non-fluorescent) ES cells, and then visualized fluorescence in resulting chimeric colonies (Fig. 1). The different fluorophores could easily be distinguished both with standard filter sets using epifluorescence optics and confocal microscopy, without the need for special image processing such as spectral deconvolution or linear unmixing [20]. We also noted that comparable levels of fluorescence were produced as all three color variants produced fluorescent signal within the same dynamic range.
Figure 1 Co-visualization of multiple fluorescent proteins in mouse embryonic stem (ES) cells. Mixed colony of embryonic stem (ES) cells comprised of wild type (untagged) cells, CAG::ECFP transgenic cells exhibiting widespread expression of ECFP, CAG::EGFP transgenic cells exhibiting widespread expression of EGFP and CAG::mRFP1 transgenic cells exhibiting widespread expression of mRFP1. Bright field image (A), CFP channel (B), GFP channel (C), RFP channel (D), merge of all three fluorescent channels overlayed on the bright field image (E), merge of all three fluorescent channels (F), color-coded depth projection of the three fluorescent channel merge with the color-coded scale shown on the right of the image (G). A second mixed transgenic ES cell colony with bright field image (H), merge of three fluorescent channel merge overlayed on the bright field image (I) and dark field three fluorescent channel merge (J). In all panels except G ECFP fluorescence is shown in blue, EGFP fluorescence is in green and mRFP1 fluorescence is in red. bf, bright field; df, dark field.
Having established the neutrality of widespread mRFP1 expression in ES cells, we went on to use ES cell mediated transgenesis through the generation of germline transmitting chimeras to introduce the CAG::mRFP1 transgene into mice. As a first step to test the extent of expression of the transgene in embryos, we generated 4n (tetraploid) wild type <-> mRFP1 ES cell derived chimeras [21], exactly as described previously [3]. Resulting, completely ES cell-derived embryos exhibited widespread mRFP1 expression, indicating that the level of expression was sufficiently strong to be visualized, that the transgene was not silenced and that development was able to proceed normally to midgestation (data not shown). To produce germline transmitting chimeric adult animals, we next generated diploid wild type <-> mRFP1 ES cell chimeras that were allowed to go to term. The CAG::mRFP1 transgene was transmitted to F1 offspring in a Mendelian fashion, suggesting that widespread mRFP1 expressing is compatible with normal development and fertility. Two ES cell lines were taken germline and shown to produce an equivalent intensity and range of fluorescence, therefore data from only one line are shown.
Wide field epifluorescent and laser scanning confocal microscopy was used to image this constitutively expressed transgene reporter in preimplantation stage mouse embryos hemizygous (Tg/+) for the CAG::mRFP1 transgene (Fig. 2 and additional files). Such non-invasive visualization in living preparations allowed us to acquire high-magnification, sequential optical sections (z-stacks) that were used to generate high-resolution anatomical, volumetric images of embryos. To do this, stacks of sequential optical sections were computationally reconstructed into 3-dimensional (3D) projections. This methodology was used to generate 3D image sets, and is illustrated here by imaging whole mouse embryos at the 1-cell stage and blastocyst stage. These data sets can be computationally manipulated in various ways including for the visualization of individual xy slices from a z-stack (Fig. 2B,C,K and 2L and additional files), or of rendered images from the full (Fig. 2D), or partial z-stack (Fig. 2E and 2M). It should be noted that even using epifluorescence imaging CAG::mRFP1 embryos were clearly distinguished from stage-matched CAG::EGFP embryos (Fig. 2G–I).
Figure 2 RFP expression in CAG::mRFP1 preimplantation stage embryos. Single CAG::mRFP1 Tg/+ zygote including the second polar body (A–E). A single x-y section taken from a z-stack, bright field image (A), overlay single confocal section of red fluorescence and bright field (B), red fluorescence channel only (left panel in C). Representative x-y sections taken from the same z-stack that was used to render volumes shown D and E. Rendered z-stack (3D reconstruction) of the whole zygote and second polar body (PB) shown in the previous panels and rotated through 180 degrees counter-clockwise (D). Rendered z-stack (3D reconstruction) of a computationally bisected zygote shown in the previous panels and rotated through 180 degrees counter-clockwise (E). Note that the zygote is not spherical, it has a clear short and long axis (lines with arrows), and the fertilization cone resulting from the site of sperm entry is also clearly evident as a protrusion (asterix). Non-transgenic, CAG::EGFP Tg/+, CAG::mRFP1 Tg/+ embryos recovered at E3.0 and representing compacted morulae through to blastocyst stages (F–I). Bright filed (F), red fluorescence channel (G), green fluorescence channel (H) and green and red fluorescence channel overlay (I). Single CAG::mRFP1 Tg/+ blastocyst (J–M). A single x-y section taken from a z-stack, bright field image (J), overlay single confocal section of red fluorescence and bright field (K), red fluorescence channel only (left panel in L). Representative x-y sections taken from the same z-stack that was used to render the volume shown in M. Rendered z-stack (3D reconstruction) of a computationally bisected blastocyst and rotated through 180 degrees counter-clockwise (M). Note that individual cells of the trophectoderm can be distinguished. The RGB colored vector on the bottom left of the 3D reconstruction rotations depicts the x-axis in green, y-axis in red and z-axis in blue.
Wide field microscopic imaging of later stage hemizygous embryos illustrated the robust, homogenous and widespread expression of mRFP1 from early postimplantation, embryonic day (E) 6.5 (Fig. 3A) to later fetal stages in both the embryo proper and extraembryonic lineages, including the placenta (Fig. 3E–F). Dissection of organs from fetuses confirmed widespread red fluorescence in CAG::mRFP1/+ organs contrasted with a lack of signal observed in non-transgenic littermates (Fig. 3G and 3H).
Figure 3 RFP expression in CAG::mRFP1 postimplantation embryos. Bright field and dark field epifluorescent images of CAG::mRFP1 Tg/+ embryos and non-transgenic littermates at E6.5 (A), E7.75 (B), E8.75 (C). CAG::mRFP1 Tg/+ embryos, CAG::EGFP Tg/+ embryos and non-transgenic littermates at E11.5 (D). E15.5 CAG::mRFP1 Tg/+ fetuses and non-transgenic littermates at (E) demonstrating widespread homogenous red fluorescence throughout later development in whole embryos, dissected embryonic tissues (G and H) and extraembryonic tissues including the placenta (F). (H), cardiothoracic organs from three embryos of different ages, E13.5 (left), E14.5 (center) and E15.5 (right), only two of which are hemizygous for the transgene). Ad, adrenal gland; bl, bladder; h, heart; ki, kidney; lu, lung; te, testis; th, thymus.
Examination of newborn animals revealed strong widespread expression of mRFP1 in the skin of hemizygous CAG::mRFP/+ transgenics and demonstrated that this red fluorophore can be distinguished from green fluorescence observed in CAG::EGFP/+ transgenics, and from non-transgenic littermates by standard macroscopic visualization using standard filter sets and epiflourescent excitation (Fig. 4).
Figure 4 Transgenic CAG::mRFP1 mice can be distinguished from CAG::EGFP animals. Macroscopic images of non-transgenic, CAG::mRFP1 Tg/+ and CAG::EGFP Tg/+ newborn (P5) mouse pups demonstrating that red fluorescence can clearly be distinguished from green fluorescence using conventional epifluorescent illumination and macroscopic observation. Dorsal view (A), and ventral view (B), high magnification of tails of 3 month old animals (C). Inspection of fluorescence in the tails is the method used for routine genotyping of these strains.
Further analysis of various adult organs, including the peritoneum, heart, lung, eye, brain, liver, pancreas, spleen and kidney, that were freshly obtained from CAG::mRFP1/+ adult animals revealed robust and widespread fluorescence (Fig. 5), as has been reported for animals expressing GFP-based fluorescent proteins under the regulation of the CAG promoter. We also noted that newborn pups, or bald skin and tissues expressing mRFP1, exhibit a pink coloration under normal light when compared to non-transgenic or CAG::EGFP transgenics [18]. This is particularly evident in albino animals (tails in Fig. 4C), and in unpigmented organs such as the brain and pancreas (Fig. 5E and 5G). We believe this results from mRFP1 being a red fluorphore with a spectrum closer to the visible range, so that it can be visualized as a pink pigmentation under daylight or bright field conditions. Consequently, newborn CAG::mRFP1 pups can be genotyped based on their pink pigmentation, alleviating the need to image them under epifluorescent conditions.
Figure 5 Widespread RFP expression in adult organs. Panels of bright field and corresponding dark field epifluorescent images of organs taken from a 4 week old CAG::mRFP1 Tg/+ mouse and a non-transgenic littermate. Peritoneum (A), heart (B), lung (C), eye (D), brain (E), liver (F), pancreas (G), spleen (H) and kidney (I). In addition to the fluorescence observed under epifluorescent illumination, RFP expressing tissues exhibit a pink color under bright field illumination. This is particularly evident in panels A, B, C, E and G. This allows for genotyping of newborn pups (by virtue of their pink color) in the absence of fluorescence illumination.
Moreover we succeeded in breeding the CAG::mRFP1 transgene to homozygosity. Homozygous animals retained developmental potential and therefore colonies of the strain are routinely maintained in the homozygous state. To date, we have observed no noticeable reduction in fluorescence or fertility for over four generations on both ICR outbred and 129 inbred backgrounds. From routine daily observations no behavioral differences are distinguished between CAG::mRFP1 animals and wild type age matched animals, though no behavioral tests have been implemented.
Discussion
We have established the applicability of monomeric RFPs as an alternative spectrally-distinct genetically-encoded fluorescent reporter to GFP and its variants. Now monomeric DsRed variants can be used in vivo in mice along side GFP variants to differentially label single cells or different populations of cells.
Advances in optical imaging modalities, and the development of ex vivo embryo or explant culture systems, have evolved alongside the increasing availability of genetically-encoded fluorescent proteins permitting dynamic in vivo imaging of living specimens. Two major limitations in the application of genetically-encoded fluorescent protein technology in mice, have been the lack of fluorescent proteins in the longer wavelength (red) part of the spectrum, and the absence of a fluorescent protein that is genetically distinguishable from GFP. DsRed, the coelenterate derived RFP, has offered the potential to solve both these problems, but original variants, which function as obligate tetramers and dimers, have exhibited 'toxicity', limiting their use in heterologous systems.
Our work demonstrates the developmental neutrality of mRFP1, the first true monomeric RFP, in ES cells and mice. The strain of mice we have generated exhibits widespread mRFP1 expression, providing a novel reagent for live imaging of red fluorescent cells in a genetically tractable mammalian model organism. CAG::mRFP1 animals represent a resource for analyzing development and disease in mouse embryos and adults. They can also be used as tagged populations of cells in chimeras, in addition to transplantation and cell isolation experiments. Moreover RFP cells can easily be distinguished from GFP, or GFP spectral variant expressing cells (including ECFP and EYFP) by both imaging using standard optics, and molecular analyses, due to the different genetic origin.
To date all fluorescent proteins isolated from diverse anthozoan species suffer from obligate tetramerization and will require efforts similar to the evolution of mRFP1 to produce widely useful tools. Our demonstration of the developmental neutrality of widespread mRFP1 expression paves the way for the incorporation of monomeric RFPs into mouse transgenic and gene targeting approaches [2]. Our data on mRFP1 suggest that its mutagenized derivatives will also be amenable to use in mice. If so, there should now be at least six new spectral variant monomeric FPs based on DsRed that can be used in mice. Key applications of the proliferation of spectral variant FPs will include discriminating different cells, transcriptional activities and/or fusion proteins.
Conclusion
Our work demonstrates the use of monomeric RFPs in mice and provides the basis for future efforts designed to incorporate RFPs into protein fusions that can be used as spectrally-distinct subcellularly-localized tags permitting high-resolution live imaging in vivo [12,22]. The ongoing development of mouse strains expressing subcellularly-localized protein fusions incorporating RFPs, and contrasting with GFP-variants will be essential for visualizing 4-dimensional (3-dimensions over time) anatomy and tracking cell position, morphology and behavior in vivo.
Methods
Vector construction
The mRFP1 coding sequence was amplified by PCR using primers 5'RFP (5'-cgtagaattcgccaccaatggctagcatgactgg) and 3'RFP (5'-gcacgaattcgggcgccggtggagtggcggcc) using Pfx Polymerase (Invitrogen). The resulting product was cloned into the EcoRI site of pCAGGS to generate pCX-mRFP1. Transient transfection of Cos-7 cells using Fugene 6 Transfection Reagent as per manufacturer's recommendations (Roche) was used to evaluate pCX-mRFP1, and verify that it produced robust red fluorescence.
Generating transgenic ES cells
Transgenic ES cell lines constitutively expressing mRFP1 were generated by co-electroporation of SalI linearized pCX-mRFP1 construct and a circular PGK-Puro-pA plasmid conferring transient puromycin resistance. Puromycin selection was carried out exactly as described previously [18]. Thereafter cells were passaged according to standard protocols.
Mouse breeding
Two lines of CAG::mRFP1 ES cells were used for chimera generation by injection into C57BL/6 blastocysts using standard procedures [23]. Chimeras were mated to outbred ICR and inbred 129/Tac mice (Taconic, Germantown, NY) for germline transmission and subsequent maintenance of the lines. After germline transmission, both transgenes were bred to, and maintained at, homozygosity, suggesting that the sites of transgene integration were not perturbing essential gene function and that high levels of protein expression were non-toxic. Animals from both lines retained widespread homogenous fluorescence in all subsequent generations tested and therefore data was pooled.
Embryo collection
Preimplantation embryos were recovered in M2 media and subsequently cultured under oil in a tissue culture incubator gased at 5% CO2 in KSOM media. Postimplantation embryos and organs were dissected either in HEPES buffered DMEM containing 10% fetal calf serum or PBS containing 0.1% BSA, and cultured in media comprising 50% rat serum, 50% DMEM/F12 supplemented with L-glutamine.
Image acquisition
Also although mRFP1 has been shown to have a lower extinction coefficient, quantum yield, and photostability than native DsRed or DsRed.T3, mRFP1 has been shown to mature over ten fold faster, so that it shows similar brightness in living cells. In addition, the excitation and emission peaks of mRFP1 (584 and 607 nm respectively) are 25 nm red-shifted from DsRed, which should confer greater tissue penetration and spectral separation from autofluorescence and other fluorescent proteins. All images presented in the figures are of living embryos or freshly dissected (unfixed) tissues maintained under physiological conditions. Wide-field images were acquired with an AxioCam MRc camera attached to a Leica MZ16FA stereo-dissecting microscope or a Zeiss Axiovert 200M inverted microscope equipped with epifluorescent illumination using appropriate filter sets (Chroma). Laser scanning confocal data was acquired in single-track mode using a Zeiss LSM510 META on a Zeiss Axiovert 200M. Fluorophores were excited with a 405 nm diode laser (ECFP), 488 nm Argon laser line (EGFP) and a 543 nm HeNe laser (mRFP1). Objectives used were a C-apochromat 40×/NA1.2, plan-apochromat 20×/0.75 and a fluar 5×/0.25. Confocal images were acquired as z-stacks comprising sequential optical x-y sections taken at 1–2 μm z-intervals.
Image processing
Raw data was processed using Zeiss AIM software (Carl Zeiss Microsystems at ), and Volocity (Improvision at ). Re-animation of data to generate movies of time-lapses or rotations was performed using QuickTime Pro (Apple Computer, Inc at ).
Authors' contributions
JZL and CSL carried out the preimplantation embryo experiments including all imaging and the generation of the aggregation chimeras. AKH planned the study, carried out the experiments in ES cells and postimplantation embryos, and wrote the manuscript.
Footnote
The CAG::mRFP1 mice described in this paper will be made available from the Jackson Laboratories Induced Mutant Resource () as stock number #5645.
Supplementary Material
Additional File 1
Sequential x-y images of a z-stack taken through a CAG::mRFP1 Tg/+ zygote (1-cell stage). The first part of the sequence includes bright field (DIC) images merged with the red fluorescence channel, while the second part depicts only the red fluorescence channel.
Click here for file
Additional File 2
3D reconstruction and rotation of the complete raw data set shown in Movie 1 depicting a view of a full zygote. The sequence depicts the rendered image being rotated around each axes (z, y and z). The RGB colored vector on the bottom left depicts the x-axis in green, y-axis in red and z-axis in blue.
Click here for file
Additional File 3
3D reconstruction and rotation of half the raw data shown in Movie 1 depicting a view of a computationally bisected zygote. The sequence depicts the rendered image being rotated around each axes (z, y and z). The RGB colored vector on the bottom left depicts the x-axis in green, y-axis in red and z-axis in blue.
Click here for file
Additional File 4
Sequential x-y images of a z-stack taken through a CAG::mRFP1 Tg/+ blastocyst stage embryo. The first part of the sequence includes bright field (DIC) images merged with the red fluorescence, while the second part depicts only the red fluorescence channel.
Click here for file
Additional File 5
3D reconstruction and rotation of the complete raw data set shown in Movie 4 depicting a view of half a blastocyst. The sequence depicts the rendered image being rotated around each axes (z, y and z). The RGB colored vector on the bottom left depicts the x-axis in green, y-axis in red and z-axis in blue.
Click here for file
Acknowledgements
We thank J.-I. Miyazaki and R. Tsien for the CAGGS and mRFP1 plasmids respectively, the Memorial Sloan-Kettering Cancer Center (MSKCC) Transgenic Core Facility for production of germline transmitting chimeras, and E. Lacy for comments on the manuscript. This work was supported by laboratory start-up funds from the MSKCC to AKH.
==== Refs
Tsien RY The green fluorescent protein Annu Rev Biochem 1998 67 509 544 9759496 10.1146/annurev.biochem.67.1.509
Hadjantonakis AK Dickinson ME Fraser SE Papaioannou VE Technicolour transgenics: imaging tools for functional genomics in the mouse Nat Rev Genet 2003 4 613 625 12897773 10.1038/nrg1126
Hadjantonakis AK Macmaster S Nagy A Embryonic stem cells and mice expressing different GFP variants for multiple non-invasive reporter usage within a single animal BMC Biotechnol 2002 2 11 12079497 10.1186/1472-6750-2-11
Zhang J Campbell RE Ting AY Tsien RY Creating new fluorescent probes for cell biology Nat Rev Mol Cell Biol 2002 3 906 918 12461557 10.1038/nrm976
Verkhusha VV Lukyanov KA The molecular properties and applications of Anthozoa fluorescent proteins and chromoproteins Nat Biotechnol 2004 22 289 296 14990950 10.1038/nbt943
Matz MV Fradkov AF Labas YA Savitsky AP Zaraisky AG Markelov ML Lukyanov SA Fluorescent proteins from nonbioluminescent Anthozoa species Nat Biotechnol 1999 17 969 973 10504696 10.1038/13657
Gross LA Baird GS Hoffman RC Baldridge KK Tsien RY The structure of the chromophore within DsRed, a red fluorescent protein from coral Proc Natl Acad Sci U S A 2000 97 11990 11995 11050230 10.1073/pnas.97.22.11990
Gurskaya NG Fradkov AF Terskikh A Matz MV Labas YA Martynov VI Yanushevich YG Lukyanov KA Lukyanov SA GFP-like chromoproteins as a source of far-red fluorescent proteins FEBS Lett 2001 507 16 20 11682051 10.1016/S0014-5793(01)02930-1
Wiedenmann J Schenk A Rocker C Girod A Spindler KD Nienhaus GU A far-red fluorescent protein with fast maturation and reduced oligomerization tendency from Entacmaea quadricolor (Anthozoa, Actinaria) Proc Natl Acad Sci U S A 2002 99 11646 11651 12185250 10.1073/pnas.182157199
Miyawaki A Green fluorescent protein-like proteins in reef Anthozoa animals Cell Struct Funct 2002 27 343 347 12502888 10.1247/csf.27.343
Campbell RE Tour O Palmer AE Steinbach PA Baird GS Zacharias DA Tsien RY A monomeric red fluorescent protein Proc Natl Acad Sci U S A 2002 99 7877 7882 12060735 10.1073/pnas.082243699
Hadjantonakis AK Papaioannou VE Dynamic in vivo imaging and cell tracking using a histone fluorescent protein fusion in mice BMC Biotechnol 2004 4 33 15619330 10.1186/1472-6750-4-33
Bevis BJ Glick BS Rapidly maturing variants of the Discosoma red fluorescent protein (DsRed) Nat Biotechnol 2002 20 83 87 11753367 10.1038/nbt0102-83
Baird GS Zacharias DA Tsien RY Biochemistry, mutagenesis, and oligomerization of DsRed, a red fluorescent protein from coral Proc Natl Acad Sci U S A 2000 97 11984 11989 11050229 10.1073/pnas.97.22.11984
Vintersten K Monetti C Gertsenstein M Zhang P Laszlo L Biechele S Nagy A Mouse in red: red fluorescent protein expression in mouse ES cells, embryos, and adult animals Genesis 2004 40 241 246 15593332 10.1002/gene.20095
Shaner NC Campbell RE Steinbach PA Giepmans BN Palmer AE Tsien RY Improved monomeric red, orange and yellow fluorescent proteins derived from Discosoma sp. red fluorescent protein Nat Biotechnol 2004 22 1567 1572 15558047 10.1038/nbt1037
Niwa H Yamamura K Miyazaki J Efficient selection for high-expression transfectants with a novel eukaryotic vector Gene 1991 108 193 199 1660837 10.1016/0378-1119(91)90434-D
Hadjantonakis AK Gertsenstein M Ikawa M Okabe M Nagy A Generating green fluorescent mice by germline transmission of green fluorescent ES cells Mech Dev 1998 76 79 90 9867352 10.1016/S0925-4773(98)00093-8
Hadjantonakis AK Nagy A FACS for the isolation of individual cells from transgenic mice harboring a fluorescent protein reporter Genesis 2000 27 95 98 10951501 10.1002/1526-968X(200007)27:3<95::AID-GENE10>3.0.CO;2-A
Dickinson ME Bearman G Tille S Lansford R Fraser SE Multi-spectral imaging and linear unmixing add a whole new dimension to laser scanning fluorescence microscopy Biotechniques 2001 31 1272, 1274 6, 1278 11768655
Nagy A Gocza E Diaz EM Prideaux V Ivanyi E Markkula M Rossant J Embryonic stem cells alone are able to support fetal development in the mouse. Development 1990 110 815 821 2088722
Plusa B Hadjantonakis AK Gray D Piotrowska-Nitsche K Jedrusik A Papaioannou VE Glover DM Zernicka-Goetz M The first cleavage of the mouse zygote predicts the blastocyst axis Nature 2005 434 391 395 15772664 10.1038/nature03388
Nagy A Gertsenstein M Vintersten K Behringer R Manipulating the mouse embryo: A laboratory manual 2002 , Cold Spring Harbor Press
|
15989692
|
PMC1192792
|
CC BY
|
2021-01-04 16:03:05
|
no
|
BMC Cancer. 2005 Jun 30; 5:67
|
latin-1
|
BMC Cancer
| 2,005 |
10.1186/1471-2407-5-67
|
oa_comm
|
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-301604864510.1186/1471-2296-6-30Research ArticlePatients' views on outcome following head injury: a qualitative study Morris Paul Graham [email protected] Lindsay [email protected] Shoumitro [email protected] Glyn [email protected] Wendy [email protected] Caroline E [email protected] Eleanor [email protected] Section of Clinical & Health Psychology, Old Medical School, Teviot Place, Edinburgh EH8 9AG, Scotland2 School of Social Sciences, Cardiff University, Cardiff CF10 3WT, Wales3 Department of Psychiatry, University of Birmingham, B15 2QZ, England4 University of Bristol, Cotham House, Cotham Hill, Bristol BS6 6JL, England5 Headway Cardiff, Rookwood Hospital, Cardiff, Wales2005 27 7 2005 6 30 30 19 1 2005 27 7 2005 Copyright © 2005 Morris et al; licensee BioMed Central Ltd.2005Morris et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Head injuries are a common occurrence, with continuing care in the years following injury being provided by primary care teams and a variety of speciality services. The literature on outcome currently reflects areas considered important by health-care professionals, though these may differ in some respects from the views of head injured individuals themselves. Our study aimed to identify aspects of outcome considered important by survivors of traumatic head injury.
Methods
Thirty-two individuals were interviewed, each of whom had suffered head injury between one and ten years previously from which they still had residual difficulties. Purposive sampling was used in order to ensure that views were represented from individuals of differing age, gender and level of disability. These interviews were fully transcribed and analysed qualitatively by a psychologist, a sociologist and a psychiatrist with regular meetings to discuss the coding.
Results
Aspects of outcome mentioned by head injury survivors which have received less attention previously included: specific difficulties with group conversations; changes in physical appearance due to scarring or weight change; a sense of loss for the life and sense of self that they had before the injury; and negative reactions of others, often due to lack of understanding of the consequences of injury amongst both family and general public.
Conclusion
Some aspects of outcome viewed as important by survivors of head injury may be overlooked by health professionals. Consideration of these areas of outcome and the development of suitable interventions should help to improve functional outcome for patients.
==== Body
Background
The consequences of brain injury upon an individual are many and varied, with some individuals fortunately making a good recovery whilst others continue to be affected in the longer term. Residual difficulties one year post injury often include physical effects such as fatigue, paralysis, seizures and headaches and/or cognitive effects such as difficulties with attention, short-term memory, planning or language. Increasing recognition in recent years has been given to social and psychological aspects of outcome such as anxiety, depression and social isolation [1-5].
Most studies of outcome following head injury have been based upon questionnaires, which can severely constrain responses and often focus upon the presence of symptoms rather than the degree to which they are causing problems. These outcome measures are invariably devised by health care professionals, reflecting aspects of outcome that they consider important, but with little or no input from head injured individuals themselves. It is reasonable to suppose that professionals and patients will differ in some respects in the aspects of outcome that they consider important and the language used to describe these areas of outcome. However there has been little previous research aimed at eliciting the unconstrained views of head injured individuals themselves upon aspects of outcome.
Qualitative studies of outcome following stroke have enabled a greater understanding of many aspects of individuals' responses to stroke [6], including how patients recognise and respond to their stroke [7]; their experiences whilst in hospital [8]; the strategies they use to manage their illness [9] and their information needs following stroke [10,11]. One such study highlighted consequences of stroke that were important to stroke survivors, but which probably would not have been identified using standardised outcome measures [12].
In order to identify consequences of head injury considered important by patients, we conducted qualitative interviews with survivors of head injury who had residual symptoms at least one year post injury.
Methods
Design of study
Semi-structured interviews were conducted with 32 individuals who had suffered head injury and subsequently returned to a home environment.
Participants
A purposive sampling strategy was employed in order to ensure that views were represented from individuals of differing age, gender and level of disability (Table 1). Potential interviewees with probable moderate or severe disability resulting from a traumatic head injury sustained whilst aged over 16 were identified via local head injury services. Actual level of disability was subsequently determined using the extended Glasgow Outcome Scale, based upon information obtained at interview [13]. Eighty-nine were contacted by letter to explain the study, of these 42 replied and 40 were willing to be interviewed about the consequences of their injury. Thirty-two of these were interviewed, by which time no further themes were being generated and the analysis was deemed to have reached saturation.
Table 1 Number of Participants by Age, Gender and Disability Level
Age at Injury Gender Upper Moderate
Disability Lower Moderate
Disability Severe Disability
16 – 29 Male 6 6 4
Female 1 3 1
30+ Male 3 3 2
Female 0 3 0
Disability levels are based on extended Glasgow Outcome Scale [13]
Ethical Approval
Approval was obtained from the local Bro Taff NHS Health Ethics committee and informed consent was given by all patients who participated in the study.
Interviews
Those who agreed to participate were visited in their own homes and again had the opportunity to ask questions about the study before being asked to complete consent forms. All interviews were conducted at least one year post injury (range 1–10 years) and all interviewees had returned to a home environment for at least six months prior to being interviewed. Where possible the head injured individual was interviewed alone, though in two cases a carer was present. The interviews were semi-structured, with interviewees asked to describe their lives prior to the injury and then to describe the consequences of head injury that had been most important to them. These interviews were all recorded onto minidisk and transcribed in full.
Analysis
The analysis was a continual process in parallel with data collection, involving the repeated reading of recent transcripts in combination with listening to the recorded interviews. Emergent themes reported by participants as being important in their outcome following head injury were then coded by a health psychologist with experience of working with brain injured individuals (PGM). A random selection of over 50% of the transcripts were read and coded separately by a sociologist (LP) and a neuropsychiatrist (SD), with regular meetings held to discuss the coding. A sampling to saturation strategy was employed, with 139 themes generated by the time of the 32nd interview, by which time no further themes were emerging from new interviews. Through a process of discussion and comparison of transcripts, this list of codes was merged into 43 broader representative categories which were then grouped into six domains (Figure 1). These outcome categories were then discussed with members of a local head injury support group.
Figure 1 Interacting domains and categories of outcome.
Results
The distribution of participants by age at injury, gender and level of disability is shown in Table 1. Representative categories of outcome divided into six domains are shown in Figure 1. There is considerable interaction and overlap between these domains and they are not intended to be viewed as independent from each other. For example, difficulties with group conversations often led to reduced social interaction, which in turn is often a cause of depression. Also fatigue may be viewed as much as a cognitive or emotional problem as a physical one. The domains are thus intended as interacting and interdependent and, as would be expected, most of these categories reflect areas of outcome that are well recognised in the current literature. Our results focus upon those categories of outcome which are less well documented and less often measured in current outcome measures.
Group conversations
Particular difficulties in attending to more than one voice at a time were reported by eleven participants (Figure 2). In each case the individuals concerned had no difficulties understanding a single voice, but were unable to make sense of conversations where several people were speaking at once: "I only listen to one at a time" (patient 18); "I found it very difficult to cope with two voices" (patient 14). As a result, some reported feeling isolated or uncomfortable in these situations and consequently avoided situations, such as pubs, where they were likely to have to deal with conversations involving more than one other person. This naturally resulted in less social interaction, which itself was identified by most participants as an important aspect of their outcome.
Figure 2 Group conversations and physical appearance.
Physical appearance
Eight participants mentioned changes in their physical appearance as important aspects of their outcome (Figure 2). These were generally either due to scars or weight change, resulting in concerns about self-image and the perceptions of others. Scars were caused either by neurosurgery or by other injuries which were sustained at the same time as the head injury. Whilst concerns about gains in weight due to less activity were mentioned by both males and females, loss of weight and strength was also mentioned by one male as an important aspect of his outcome.
Loss
Most participants mentioned feelings of loss relating to aspects of the life that they had before the head injury: "a head injury should have a warning of great losses," (patient 11); "I just seem to have lost everything" (patient 24) (Figure 3). These losses included loss of work, of friends, of partners and of abilities. Some felt that they had lost the person they were before the injury, feeling that they were a totally different person now: "I don't think I will ever go back to the young woman I was before the accident" (patient 23); "...it was like a different life after the accident" (patient 28). Others mentioned feelings of loss for the life that they would have had now if the head injury hadn't happened: "I start thinking about what would have happened if I hadn't been knocked over and what life would have been like..." (patient 2).
Figure 3 Loss.
Negative reactions of others
Sixteen participants mentioned difficulties which were directly related to the reactions of others, some examples of which are given in Figure 4. Often these negative reactions were due to a lack of understanding of the consequences of head injury, both amongst the general public and some health care professionals. Thus, for example, some people couldn't understand why head injured individuals were often depressed or exhausted. Sometimes such lack of understanding seemed in part to be due to the absence of external signs of injury. "They put a tie on you and cut your hair and you look ok don't you" (patient 14); "physically I might look alright, but mentally ..." (patient 16); "I looked exactly the same as I did before, so what's really changed is in here (pointing at head), not outside" (patient 17). Consequently, as others could see no external sign of injury they expected them to cope as well as anyone else. Others over-compensated and inadvertently caused offence by treating head injured individuals as if they were far less capable than was actually the case.
Figure 4 Negative reactions of others.
Other participants mentioned being ignored or over-looked by others. This was particularly a problem amongst those who used wheelchairs since their injury or had some speech impairment (Figure 4). Other negative reactions included discrimination which had resulted in difficulties finding suitable employment or in being excluded from social groups.
Discussion
The results highlight areas of outcome which are important to survivors of head injury but which have previously received less attention than they perhaps deserve. This may be due in part to their being less easily identified either in clinical interviews or via current outcome measures. The areas of outcome and means of description derived from these patient interviews have subsequently been used to develop a patient centred outcome measure which focuses upon the effects that particular symptoms have upon the individual, rather than simply whether they are present.
Amongst those areas of outcome perhaps worthy of further attention, difficulties with group conversations were mentioned as an important aspect of outcome by eleven participants. However as these individuals had no difficulties understanding single voices, these difficulties might not be identified by interview or questionnaires unless they asked specifically about situations where there were several voices. Whilst cognitive difficulties with divided attention following head injury are reasonably well established [14], the translation to how this affects day-to-day life for survivors in terms of group conversations appears less well known. Although these individuals would not have any input from speech and language therapy, one individual (patient 3) mentioned that he focused on people's lips more in order to cope when there were other distracting voices. Thus it is possible that some training in lip-reading may be of benefit to individuals with difficulties following group conversations.
It is possible that concerns about physical appearance are under-reported to health care professionals, as they are viewed as cosmetic rather than 'medical'. Also males, who are several times more likely to suffer head injury, may be particularly reticent to mention such concerns. However the interviews highlight that these changes are of concern to survivors and may influence self-esteem and sense of identity, particularly amongst individuals feeling more vulnerable due to other consequences of their injury.
A sense of loss was mentioned by most interviewees. Whilst some had begun to come to terms with these losses and accept their new self and abilities, others were still unable to accept their losses and these continued to be a cause of depression and frustration for them. As such, some aspects of bereavement counselling may be beneficial in supporting individuals with head injury.
Misconceptions and negative reactions were sometimes present amongst family as well as amongst the general public, with similar misconceptions previously reported in a study of carers and health care professionals [15]. Whilst misunderstandings amongst the general public may be more difficult to address, the provision of suitable information to family and friends may help to reduce difficulties due to lack of understanding. Videos have a number of potential benefits in conveying such information, including being able to be watched by the whole family at the same time, enabling illustration of case examples by 'real' people rather than just quotes and being easier to attend to than written or verbal information.
Strengths and limitations of this study
This study reports the views of survivors of head injury themselves, validating the importance of some areas of outcome and highlighting aspects of outcome whose importance has either been under-estimated or viewed differently by health care professionals. The sample was selected in order to obtain a range of views from individuals of differing age, gender and disability, with sampling continuing until no further themes were being generated. The study focused upon those with residual difficulties at least one year after head injury, excluding those who had made a good recovery as measured by the extended Glasgow Outcome Scale. Whilst it is appreciated that some of these individuals have residual difficulties, it is likely that these would be fewer and less pronounced and would also be present amongst some of those with moderate or severe disability and thus accounted for in our study.
All of the interviews were recorded and fully transcribed, with these interviews coded by three investigators, each of whom could bring differing relevant training and experience to facilitate the coding. This process reduces, but naturally cannot exclude, the potential for bias in the analysis of the interviews. However some validation for the findings was provided by head injury survivors who attended the local Headway group, who agreed that the key areas of outcome identified by the study accurately reflected aspects of outcome that were important following head injury.
Conclusion
Head injury leads to a wide range of emotional, social, cognitive and physical difficulties, with primary care teams and various speciality services providing support and care to those who have returned to a home environment. Whilst most areas of outcome mentioned by survivors are already well known, areas of outcome where more attention might be deserved include: difficulties with group conversations in individuals who have no problems with one-to-one conversations, sensitivity to concerns relating to changes in physical appearance, the consideration of sense of loss amongst individuals following injury and the various misconceptions and negative reactions from others. Consideration of these aspects of outcome should help to enhance understanding of the difficulties faced by head injury survivors and improve functional outcome for these patients.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
PGM was involved in study design, conducted and analysed interviews, wrote the paper and will serve as guarantor for the integrity of the data. LP and SD were both involved in study design, analysed interviews and contributed to editing the paper. GL and WM were both involved in study design and contributed to editing the paper. CEB and EB both contributed to editing the paper.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
We are very grateful to Headway and all who agreed to be interviewed for this study.
==== Refs
Kersel DA Marsh NV Havill JH Sleigh JW Psychosocial functioning during the year following severe traumatic brain injury Brain Inj 2001 15 683 696 11485609 10.1080/02699050010013662
Wilson JT Pettigrew LE Teasdale GM Emotional and cognitive consequences of head injury in relation to the glasgow outcome scale J Neurol Neurosurg Psychiatry 2000 69 204 209 10896694 10.1136/jnnp.69.2.204
Milders M Fuchs S Crawford J Neuropsychological impairments and changes in emotional and social behaviour following severe traumatic brain injury J Clin Exp Neuropsychol 2003 25 157 172 12754675
Morris PG Wilson JT Dunn L Anxiety and depression after spontaneous subarachnoid hemorrhage Neurosurgery 2004 54 47 52; discussion 52-4 14683540 10.1227/01.NEU.0000097198.94828.E1
Deb S Lyons I Koutzoukis C Neurobehavioural symptoms one year after a head injury Br J Psychiatry 1999 174 360 365 10533556
McKevitt C Redfern J Mold F Wolfe C Qualitative studies of stroke: a systematic review Stroke 2004 35 1499 1505 15105517 10.1161/01.STR.0000127532.64840.36
Yoon SS Byles J Perceptions of stroke in the general public and patients with stroke: a qualitative study Bmj 2002 324 1065 1068 11991910 10.1136/bmj.324.7345.1065
Pound P Bury M Gompertz P Ebrahim S Stroke patients' views on their admission to hospital Bmj 1995 311 18 22 7613315
Pound P Gompertz P Ebrahim S Social and practical strategies described by people living at home with stroke Health Soc Care Community 1999 7 120 128 11560628 10.1046/j.1365-2524.1999.00168.x
Wiles R Pain H Buckland S McLellan L Providing appropriate information to patients and carers following a stroke J Adv Nurs 1998 28 794 801 9829668 10.1046/j.1365-2648.1998.00709.x
Hanger HC Walker G Paterson LA McBride S Sainsbury R What do patients and their carers want to know about stroke? A two-year follow-up study Clin Rehabil 1998 12 45 52 9549025 10.1191/026921598668677675
Pound P Gompertz P Ebrahim S A patient-centred study of the consequences of stroke Clin Rehabil 1998 12 338 347 9744669 10.1191/026921598677661555
Wilson JT Pettigrew LE Teasdale GM Structured interviews for the Glasgow Outcome Scale and the extended Glasgow Outcome Scale: guidelines for their use J Neurotrauma 1998 15 573 585 9726257
Park NW Moscovitch M Robertson IH Divided attention impairments after traumatic brain injury Neuropsychologia 1999 37 1119 1133 10509834 10.1016/S0028-3932(99)00034-2
Swift TL Wilson SL Misconceptions about brain injury among the general public and non-expert health professionals: an exploratory study Brain Inj 2001 15 149 165 11260765 10.1080/026990501458380
|
16048645
|
PMC1192793
|
CC BY
|
2021-01-04 16:29:12
|
no
|
BMC Fam Pract. 2005 Jul 27; 6:30
|
utf-8
|
BMC Fam Pract
| 2,005 |
10.1186/1471-2296-6-30
|
oa_comm
|
==== Front
BMC Fam PractBMC Family Practice1471-2296BioMed Central London 1471-2296-6-331609113010.1186/1471-2296-6-33Research ArticleThe definition of disabling fatigue in children and adolescents Fowler Tom [email protected] Pamela [email protected] Anita [email protected] Anne [email protected] Department of Psychological Medicine, Wales College of Medicine, Cardiff University, UK2 Brynffynnon Child & Family Service Unit, Pontypridd, UK3 MRC Social, Genetic, Developmental Psychiatric Research Centre, Institute of Psychiatry, London, UK2005 9 8 2005 6 33 33 14 2 2005 9 8 2005 Copyright © 2005 Fowler et al; licensee BioMed Central Ltd.2005Fowler et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Disabling fatigue is the main illness related reason for prolonged absence from school. Although there are accepted criteria for diagnosing chronic fatigue in adults, it remains uncertain as to how best to define disabling fatigue and Chronic Fatigue Syndrome (CFS) in children and adolescents. In this population-based study, the aim was to identify children who had experienced an episode of disabling fatigue and examine the clinical and demographic differences between those individuals who fulfilled a narrow definition of disabling fatigue and those who fulfilled broader definitions of disabling fatigue.
Methods
Participants (aged 8–17 years) were identified from a population-based twin register. Parent report was used to identify children who had ever experienced a period of disabling fatigue. Standardised telephone interviews were then conducted with the parents of these affected children. Data on clinical and demographic characteristics, including age of onset, gender, days per week affected, hours per day spent resting, absence from school, comorbidity with depression and a global measure of impairment due to the fatigue, were examined. A narrow definition was defined as a minimum of 6 months disabling fatigue plus at least 4 associated symptoms, which is comparable to the operational criteria for CFS in adults. Broader definitions included those with at least 3 months of disabling fatigue and 4 or more of the associated symptoms and those with simply a minimum of 3 months of disabling fatigue. Groups were mutually exclusive.
Results
Questionnaires were returned by 1468 families (65% response rate) and telephone interviews were completed on 99 of the 129 participants (77%) who had experienced fatigue. There were no significant differences in demographic and clinical characteristics or levels of impairment between those who fulfilled the narrower definition and those who fulfilled the broader definitions. The only exception was the reported number of days per week that the child was affected by the fatigue. All groups demonstrated evidence of substantial impairment associated with the fatigue.
Conclusion
Children and adolescents who do not fulfil the current narrow definition of CFS but do suffer from disabling fatigue show comparable and substantial impairment. In primary care settings, a broader definition of disabling fatigue would improve the identification of impaired children and adolescents who require support.
==== Body
Background
Disabling fatigue in children and adolescents is an important clinical problem given that it is associated with severe functional impairment that has an impact upon the child's education and social development [1,2]. It is the main illness related reason for prolonged absence from school in young people [3]. It is also a presenting problem for 2% and a background problem in 11% of school children attending general paediatric clinics [4]. There is uncertainty about how to appropriately define disabling fatigue in children and adolescents. Reviews of disabling fatigue in children and adolescents highlight the lack of research in this age group in comparison to adults [2].
There have been several attempts to define disabling fatigue as a disorder in adults and perhaps the most accepted is the Center for Disease Control (CDC) definition [5]. This requires 6 months of disabling fatigue and 4 or more of the following symptoms; 1) impaired memory or concentration, 2) sore throat, 3) tender cervical axillary lymph nodes, 4) muscle pain, 5) multi-joint pain, 6) new headaches, 7) unrefreshing sleep, 8) post-exertion malaise. This has, to date, been the most widely used definition when reporting the prevalence of CFS and CFS like illness in children and adolescents [6-9].
A number of authors have raised the concern that the current criteria for CFS were designed for use in adults and that there has been relatively little work in assessing how appropriate these criteria are for children and adolescents [10-14]. Although a consensus document on CFS by the Royal Colleges [2] suggested that the criteria were applicable to children and adolescents it did also suggest the period of disabling fatigue required for a diagnosis of CFS should be reduced from 6 to 3 months.
The importance that these associated symptoms play in CFS and whether they contribute to defining a homogenous group of individuals with disabling fatigue has been increasingly questioned [15]. As the associated symptoms were originally identified by expert consensus there are growing calls for studies examining individuals with chronic unexplained fatigue to empirically derive a definition of CFS [16].
Given the current discussion about the role of the associated symptoms in CFS in adults, the lack of empirical testing about the appropriateness of the adult CFS criteria in children it seems pertinent to examine whether the clinical and demographic features of children and adolescents differ in those who fulfil CFS criteria and those who fulfil broader criteria for disabling fatigue. Evidence that those individuals with "adult like CFS" did differ significantly would give credence to the idea that adult CFS criteria do identify a specific group of fatigued individuals.
There is also evidence to suggest that the majority (two thirds) of adults who attend GP clinics regarding disabling fatigue lasting more then 6 months do not fulfil the CDC definition of CFS, but do have substantial associated impairment [17]. Given this it is also important to examine and compare the levels of impairment in children and adolescents who fulfil CFS criteria and those who fulfil broader definitions of disabling fatigue. The latter may represent an important group of individuals who require treatment and primary care support.
In this paper, we set out to identify children and adolescents with a lifetime ever episode of disabling fatigue in a population-based sample and test whether those who fulfilled a narrow definition were more severely impaired and different in terms of clinical and demographic characteristics than those who only fulfilled broader definitions of chronic disabling fatigue. Specifically, three groups of individuals with disabling fatigue were defined; those who had a similar symptom profile, duration and impairment as the CDC operational criteria for CFS in adults ("adult like CFS"), those who also had a similar symptom profile and impairment but whose fatigue, at time of interview, had lasted for a period of between 3–6 months disabling fatigue ("child like CFS") and those that did not fulfil the symptom profile for operationally defined CFS but who had had more than 3 months disabling fatigue ("3 months plus disabling fatigue").
Methods
Participants
Participants were identified from the population based twin register CaStANET (Cardiff Study of All Wales and North West England Twins). Previous studies have shown that the register is representative of the local population [18]. Initially 2259 twin pairs aged between 8–17 years old were identified from the register and a questionnaire package was sent to these twins and their parents. Families in which the parents reported that a child had experienced a period of disabling fatigue which had lasted for more than a few days and had caused interference with activities such as being able to go to school, or see friends or family, were selected for a standardised parental telephone interview. Ethical approval was granted by the Multi-Centre Research Ethics Committee (MREC) for Wales.
Measures
The parental telephone interview is described in detail elsewhere [10]. The interview consisted of two parts. The first part obtained further details about the period(s) of fatigue.
To assess whether the young person's fatigue could be classified as "adult like CFS", "child like CFS" or "3 months plus disabling fatigue" enquiry was made about the associated symptoms that sometimes occur with fatigue [5]. These were poor memory and concentration, difficulty thinking, sore throats, tender lymph nodes, muscle pain, multiple joint pains, headaches, unrefreshing sleep, fatigubility, or post-exertion malaise associated with the period of fatigue. Items were rated as 'less than usual', 'same as usual', 'more than usual' or 'a lot more than usual' for the episode of fatigue. Those children whose parents reported "more than usual" or "a lot more than usual" for 4 or more of the associated symptoms required by the CDC definition of CFS in adults were classed as fulfilling the associated symptom requirement.
Questions were also asked about the duration of the fatigue, whether it was ongoing at the time of interview and whether it had been continuous or episodic (where the periods of fatigue were described as episodic the parent was asked to focus on the longest most debilitating period of fatigue for the remainder of the questions). The duration of the fatigue at the time of interview was used to classify the twins into the different categories regardless of whether the fatigue was currently ongoing. The nature and degree of impairment associated with the fatigue was assessed by enquiring about whether the twin needed to rest for at least 1 hour daily, and whether there was interference with school attendance, and/or usual leisure activities and with family and peer relationships. To be classed as suffering from "disabling fatigue" the twin was required to need to rest for at least 1 hr daily and for there to be a report of some interference in at least one of these areas.
Individuals who did not fulfil the associated symptom requirement but had experienced a period of disabling fatigue of greater than 3 months at the time of interview were classed as suffering from "three months plus disabling fatigue". Those who did fulfil the associated symptom requirement but had only experienced a period of between 3–6 months of the disabling fatigue at the time of interview were classed as suffering from "child like CFS" and those who has experienced a period of more than 6 months disabling fatigue were classed as suffering from "adult like CFS"
Information was also sought about the age of onset of the fatigue, the number of days per week affected by the fatigue, the number of hours per day that the affected child required to rest/sleep, whether the child or adolescent was absent from school due to the fatigue, and whether they were taken to visit a General Practioner with regard to the fatigue. A global measure of impairment was created by asking whether the fatigue had interfered with the following 4 areas; i) the child's schoolwork, ii) peer relations, iii) family relations and iv) their usual leisure pursuits. The possible response's to these questions were on a discrete adjective scale consisting of "Not much", "A little" or "A lot" and were scored as 0, 1, or 2 respectively. The answers to these questions were then summed to give the overall score which ranged from 0 to 8.
Parents were asked whether any diagnoses or explanation for the disabling fatigue had been offered by GP's or hospital specialists, when these had been visited. Their responses were recorded verbatim. This information was then reviewed by one of the authors (PD) who was blind to other information from the data collection and who determined those cases where the diagnosis could entirely explain the chronic fatigue in the twin. These cases were then excluded from further analysis.
The second part of the interview consisted of the depression section of the parent version Child and Adolescent Psychiatric Assessment (CAPA) [19]. This is a standardised, reliable psychiatric interview that assesses psychopathology in children and adolescents. The responses to this were used to generate DSM-IV defined diagnoses of major depression during the episode of fatigue.
Analysis
As it was determined that the most important analysis to undertake was to assess whether those individuals who fulfilled just the broad definitions of disabling fatigue ("child like CFS" and "3 months plus disabling fatigue") differed from those who fulfilled the narrow definition ("adult like CFS") and to reduce multiple testing no comparisons were made between individual classed as suffering from "child like CFS" and those who had "3 months plus disabling fatigue". Two sets of comparisons were therefore made, the first between those individuals who had "adult like CFS" and those who had "child like CFS", the second between those individuals who had "adult like CFS" and "3 months plus disabling fatigue". Independent t-tests and chi-square tests were used as appropriate.
However, although the twin register is population based, each twin pair could potentially contributes two observations, and where this is the case the individual twin cannot be classed as statistically independent. Reduced variance due to the correlation between twins' scores can cause high false positive rates [20]. To adjust for this bias the data was treated as equivalent to a 2-stage cluster design with the twin pairs as the primary sampling unit [21]. Consequently, the survey analysis procedures of the statistical analysis package STATA Release 6 [22] were used to adjust the variances of all analyses to be equivalent to independently sampled pairs. Each family unit was classed as a clustering unit, with some clusters containing information from both twins, while others contained information from just one.
Results
Participants
Parents from 1468 families returned questionnaires (65% response rate). There were no significant socio-demographic differences between the responding and non-responding families [18]. The screening questionnaire identified one hundred and twenty nine children and adolescents who had experienced more than a few days of disabling fatigue and telephone interviews were undertaken on 99 (77%). For the remaining individuals parents had either not given permission to be contacted for a telephone interview when returning the questionnaires or when contacted did not wish to take part. Of the families who met the selection critieria there were 11 twin pairs where interviews were conducted about both twins. Following the interview, 3 participants whose parents reported diagnoses of Cerebral Palsy, Nephrotic Syndrome and Thalassaemia, were excluded from the analysis as it was felt that these disorders could entirely explain the presence of the disabling fatigue.
Analysis
Twins classed as suffering from "3 months plus disabling fatigue" had an average duration of fatigue of 25.8 months (standard error of mean, SEM, 14.0) of which 27% still had ongoing fatigue at the time of interview. Twins classed as suffering from "adult like CFS" had an average duration of 23.5 months of fatigue (SEM 6.1) with 41% with ongoing fatigue at the time of interview. Those classed as suffering "child like CFS" had by definition between 3–6 months disabling fatigue, of which 27% had ongoing fatigue at the time of interview. Table 1 shows the male: female ratio, age of onset, days per week impaired during the worst period of fatigue, hours per day impaired by the fatigue, number of days absent from school per term during the worst period of fatigue and the global impairment score. It also presents the percentage in each group who visited a GP with regard to the fatigue and who fulfilled DSM-1V criteria for depression during the period of fatigue.
Table 1 Characteristics of individuals who fulfil different duration and associated symptom criteria of disabling fatigue
Characteristic "3 months plus disabling fatigue" "Child like CFS" "Adult like CFS"
Number of affected individuals 11 15 34
Male: Female ratio 1:2.3 1:2.0 1:2.7
Age of onset in months (SEM) 117.5 (20.4) 130.5 (11.1) 146.6 (8.1)
No of Days per week affected (SEM) 6.2 (0.5) 4.6 (0.5) 5.8 (0.3)
Total No of hrs resting and sleeping per day (SEM) 13.7 (0.8) 13.3 (0.8) 13.1 (1.1)
Global impairment score (SEM) 5.9 (0.4) 6.5 (0.3) 6.5 (0.2)
No of days per term absent from school during worst period 18.8 (13.2) 17.3 (4.7) 12.9 (2.7)
% who contacted their GP with regard to the fatigue 55% 73% 71%
% with DSM-IV clinical depression 44% 52% 53%
SEM = standard error of mean
The results of the statistical analysis suggested no significant difference (P > 0.1) between the groups. The only exception to this was the number of days that the child was affected during the episode of fatigue, for which there is a significant difference between individuals who were classed as suffering from "adult like CFS" and those classed as suffering from "child like CFS" (t = -2.58, df 47, p = 0.027). These analyses were repeated using non-parametric tests, however there was no difference in results.
The descriptive statistics suggested that there was greater variability in individuals classed as suffering from "3 months plus disabling fatigue" than in individuals in the other two groups, particularly for the age of onset and number days per term absent from school. Less individuals from this group had visited a GP with regard to the fatigue. Although no significant differences or trends were found between individuals classed as suffering from "3 months plus disabling fatigue" and those classed as suffering from "adult like CFS" for these 3 variables, because of the descriptive statistics, a post hoc set of analyses was also conducted. For the post hoc analyses individuals classed as suffering from "adult like CFS" and "child like CFS" (i.e. all those individuals who had more than 3 months disabling fatigue and whose symptom profile resembled that of the CDC operational criteria for CFS) were combined into one group and compared to those classed as suffering from "3 months plus disabling fatigue". However no significant differences were found between these groups for age of onset, number of days absent from school per term or visits to the GP. As in the previous statistical analyses, these tests were also repeated using non-parametric tests however there was no difference in the pattern of results.
Discussion
The results indicate that the clinical and demographic characteristics and level of impairment of children with narrowly defined disabling fatigue and more broadly defined fatigue are similar. There was no significant difference in gender ratio, the average age of onset or the rate of comorbidity with depression between the groups. Likewise affected young people seemed to spend a similar number of hours resting and/or sleeping, there was parental report of similar levels of global impairment, they were absent for a similar number of days from school during the term in which they were most affected by the fatigue and a similar percentage of individuals were contacting their GP about the fatigue.
A lack of significant findings cannot necessarily be interpreted as meaning there is no differences between these groups but there are still a number of implications of these results. Specifically, the results suggest that a longer duration of fatigue and accompanying symptoms are not necessary in defining disabling, impairing fatigue in children and adolescents. Children who present with shorter histories of fatigue and/or in the absence of CFS associated symptoms have high levels of parent reported impairment, both in terms of family and peer relationships and school work. They also appear to be absent for a substantial number of days from school and it causes enough worry for the parents to take the majority of these children and adolescents to a GP. Even though their was a significant difference for the number of days affected by the disabling fatigue between those individuals classed as suffering from "adult like CFS" and "child like CFS", all groups reported that they were affected by the disabling fatigue for most days of the week.
The descriptive statistics also suggested the possibility that individuals within the "three months plus disabling fatigue group" may be more heterogeneous, given that there is a relatively large amount of variance in this group in comparison to the other groups. Although this may have nosological implications it does not negate the high level of impairment this group also appears to be experiencing.
One important limitation of this study is that the individual's symptoms were identified by telephone interview rather than medical examination. Although every effort was made to exclude individuals who may have had an alternative diagnosis that would explain the fatigue this does not guarantee that if an in depth medical assessment were undertaken with these individuals that a number would not be classed as fulfilling the CDC definition of CFS. This may in some way explain why over half individuals in this population based sample showed a similar symptom profile to the CDC criteria for CFS whereas in the adult population only one third were categorised as fulfilling this criteria [17], although this may also be due to differences in presentation and/or aetiology of disabling fatigue between adults and children and adolescents. However this does not change the fact that all these individuals are reporting high levels of impairment and interference with their life due to the fatigue.
A further possible limitation is that parent report of disabling fatigue was used to identify individuals for further data collection and the analysis is based on information from parental interview only. There are often low levels of agreement between parent and child/adolescent report of behavioural/psychiatric symptoms [23] and there is some evidence that this is also the case with CFS symptoms [24]. However as life time ever disabling fatigue was being examined and as there is evidence that adolescents and young adults may often not recall key symptoms (for example, over 50% of individuals with a diagnosis of depression between 15–21 failed to recall a key symptom at age 25 [25]) it seemed most appropriate to use parent report. Further interviews with twins over the age of 12 were also conducted and on prompting 97% were able to recall a period of disabling fatigue about which it was possible to conduct the interview.
Conclusion
These findings have clinical significance in that they imply episodes of fatigue in children and adolescents which last at least 3 months warrant investigation regardless of whether or not a diagnosis of CFS can be made. This is because the degree of impairment is substantial in terms of interference with usual activities, family and peer relationships, school absence and school work, and the impairment does not appear to differ from the level of impairment found in more narrowly defined disabling fatigue.
There was no indication that the group of children with shorter periods of fatigue differed on a number of associated characteristics (in terms of gender ratio, comorbidity with depression and age of onset) or in terms of impairment from the group of children who had a symptom profile similar to the CDC criteria for CFS. Although this does not prove that there is no differences between individuals in these groups it does indicate that within a primary care setting it may be appropriate to consider a broader definition of disabling fatigue in children for clinical and research purposes.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
TF interviewed the families, performed the statistical analysis and drafted the manuscript. PM provided clinical advice when conducting the analysis and helped drafted the manuscript. AT participated in the design and coordination of the study. AF conceived of the study, and participated in its design and coordination.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Acknowledgements
Funded by a grant from PPP Charitable Trust (AF & AT).
==== Refs
Garralda ME Rangel L Annotation: Chronic Fatigue Syndrome in children and adolescents J Child Psychol Psychiatry 2002 43 169 176 11902596 10.1111/1469-7610.00010
Joint Working Groups of Royal Colleges of Physicians, Psychiatrists and General Practioners Chronic Fatigue Syndrome 1996 London, Royal College of Physicians Publication Unit
Dowsett EG Colby J Long-Term Sickness Absence Due to ME/CFS in UK Schools: An Epidemiological Study with Medical And Educational Implications Journal of Chronic Fatigue Syndrome 1997 3 29 43
Garralda ME Bailey D Psychiatric disorders in general paediatric referrals Arch Dis Child 1989 64 1727 33 2624479
Fukuda K Straus SE Hickie I Sharpe MC Dobbins JG Komaroff A The chronic fatigue syndrome: a comprehensive approach to its definition and study. International Chronic Fatigue Syndrome Study Group Ann Intern Med 1994 121 953 959 7978722
Dobbins JG Randall B Reyes M Steele L Livens EA Reeves WC The Prevelence of Chronic Fatiguing Illnesses Among Adolescents in the United States Journal of Chronic Fatigue Syndrome 1997 3 15 27
Jordan KM Ayers PM Jahn SC Taylor KK Huang C Richman J Jason LA Prevalence of Fatigue and Chronic Fatigue Syndrome Like Illness in Children and Adolescents Journal of Chronic Fatigue Syndrome 2000 6 3 21 10.1300/J092v06n01_02
Chalder T Goodman S Wessely S Hotopf M Meltzer H Epidemiology of chronic fatigue syndrome and self reported myalgic encephalomyelitis in 5–15 year olds: cross sectional study BMJ 2003 327 654 655 14500438 10.1136/bmj.327.7416.654
Farmer A Fowler T Scourfield J Thapar A Prevalence of chronic disabling fatigue in children and adolescents Br J Psychiatry 2004 184 477 81 15172940 10.1192/bjp.184.6.477
Marshall GS Report of a workshop on the epidemiology, natural history, and pathogenesis of chronic fatigue syndrome in adolescents J Pediatr 1999 134 395 405 10190912
Breau LM McGrath PJ Ju LH Review of juvenile primary fibromyalgia and chronic fatigue syndrome J Dev Behav Pediatr 1999 20 278 288 10475602
Wright JB Beverley DW Chronic fatigue syndrome Arch Dis Child 1998 79 368 374 9875054
Jordan KM Landis DA Downey MC Osterman SL Thurm AE Jason LA Chronic fatigue syndrome in children and adolescents: a review J Adolesc Health 1998 22 4 18 9436061 10.1016/S1054-139X(97)00212-7
Mears CJ Taylor RR Jordan KM Binns HJ Pediatric Practice Research Group Sociodemographic and symptom correlates of fatigue in an adolescent primary care sample J Adolesc Health 2004 35 21 6 15581533 10.1016/j.jadohealth.2004.02.012
Wilson A Hickie I Hadzi-Pavlovic D Wakefield D Parker G Straus SE Dale J McCluskey D Hinds G Brickman A Goldenberg D Demitrack M Blakely T Wessely S Sharpe M Lloyd A What is chronic fatigue syndrome? Heterogeneity within an international multicentre study Aust N Z J Psychiatry 2001 35 520 527 11531735 10.1046/j.1440-1614.2001.00888.x
Reeves WC Lloyd A Vernon SD Klimas N Jason LA Bleijenberg G Evengard B White PD Nisenbaum R Unger ER the International Chronic Fatigue Syndrome Study Group Identification of ambiguities in the 1994 chronic fatigue syndrome research case definition and recommendations for resolution BMC Health Serv Res 2003 3 25 14702202 10.1186/1472-6963-3-25
Darbishire L Ridsdale L Seed PT Distinguishing patients with chronic fatigue from those with chronic fatigue syndrome: a diagnostic study in UK primary care Br J Gen Pract 2003 53 441 5 12939888
Rice F Harold GT Thapar A Assessing the effects of age, sex and shared environment on the genetic aetiology of depression in childhood and adolescence J Child Psychol Psychiatry 2002 43 1039 51 12455925 10.1111/1469-7610.00231
Angold A Costello EJ The Child and Adolescent Psychiatric Assessment (CAPA) J Am Acad Child Adolesc Psychiatry 2000 39 39 48 10638066 10.1097/00004583-200001000-00015
Bland JM Kerry SM Statistics notes. Trials randomised in clusters BMJ 1997 315 600 9302962
Simonoff E Pickles A Meyer JM Silberg JL Maes HH Loeber R Rutter M Hewitt JK Eaves LJ The Virginia Twin Study of Adolescent Behavioral Development. Influences of age, sex, and impairment on rates of disorder Arch Gen Psychiatry 1997 54 801 808 9294370
Stata Corporation Stata statistical software. (6.0) 1999 College Station, TX
Cantwell DP Lewinsohn PM Rohde P Seeley JR Correspondence between adolescent report and parent report of psychiatric diagnosticdata J Am Acad Child Adolesc Psychiatry 1997 36 610 619 9136495 10.1097/00004583-199705000-00011
Jones JF Nisenbaum R Solomon L Reyes M Reeves WC Chronic fatigue syndrome and other fatiguing illnesses in adolescents: a population-based study J Adolesc Health 2004 35 34 40 15193572 10.1016/j.jadohealth.2003.09.007
Wells JE Horwood LJ How accurate is recall of key symptoms of depression? A comparison of recall and longitudinal reports Psychol Med 2004 34 1001 11 15554571 10.1017/S0033291703001843
|
16091130
|
PMC1192794
|
CC BY
|
2021-01-04 16:29:13
|
no
|
BMC Fam Pract. 2005 Aug 9; 6:33
|
utf-8
|
BMC Fam Pract
| 2,005 |
10.1186/1471-2296-6-33
|
oa_comm
|
==== Front
BMC GastroenterolBMC Gastroenterology1471-230XBioMed Central London 1471-230X-5-251607899110.1186/1471-230X-5-25Case ReportChylous ascites as the main manifestation of left ventricular dysfunction: a case report Ridruejo Ezequiel [email protected]ó Oscar G [email protected] Hepatology Section, Department of Internal Medicine, Centro de Educación Médica e Investigaciones Clínicas "Norberto Quirno" (CEMIC), Las Heras 2939, (1425), Buenos Aires, Argentina2005 3 8 2005 5 25 25 18 4 2005 3 8 2005 Copyright © 2005 Ridruejo and Mandó; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Ascites is one of the most common complications of liver diseases, even though in 15% of the cases it is related to extrahepatic diseases; 3% are of cardiac nature and they appear associated with signs and symptoms of heart failure.
Case presentation
A 70 year old man was admitted with more than one year history of abdominal distension and a weight gain of 10 kilograms. He is asymptomatic and walks 2000–3000 meters a day without angor or dyspnea. The physical examination shows moderate abdominal distension, with no hepatosplenomegaly or edema, and there is mild jugular vein distension. The studies performed (complete laboratory work up, paracentesis, liver biopsy, echocardiogram, intrahepatic pressure measurements, etc.) showed a chylous ascites related to portal hypertension, and left ventricular dysfunction was the only probable cause found.
Conclusion
Asymptomatic heart dysfunction can mimic liver disease and should be kept in mind as a cause of chylous ascites.
==== Body
Background
Ascites is one of the most common complications of liver diseases, even though in 15% of the cases it is related to extrahepatic diseases; 3% are of cardiac nature and they appear associated with signs and symptoms of heart failure. It is also associated with certain blood laboratory abnormalities. In this cases, the ascitic fluid shows a high albumin gradient (>1.1 g/dl) and high concentrations of proteins [1-4].
We report an asymptomatic patient with chylous ascites as the only symptom of heart failure
Case presentation
A 70 year old man was admitted with more than one year history of abdominal distension and a weight gain of 10 kilograms. He was evaluated in another center and was given the diagnosis of ascites associated with cirrhosis.
His past medical history was significant due to acute myocardial infarction in 1974, mild asthma and paroxysmal atrial fibrillation. Since his last evaluation, he has been receiving diltiazem 180 mg/day, aspirin 100 mg/day, digoxin 0,750 mg/week and spironalactone 50 mg/day. He is asymptomatic and walks 2000–3000 meters a day without angor or dyspnea.
The physical examination shows moderate abdominal distension, with no hepatosplenomegaly or edema, and there is mild jugular vein distension (1/3). The hepatojugular reflux was negative. His blood pressure is 130/70 mmHg, pulse 70/ min and regular. The cardiac auscultation is normal and his lungs are clear. His current weight is 78.8 kg. One year before this admission he weighted 70 kg and before starting diuretics he was weighting 81 kg.
The blood laboratory was normal, except for a slightly increase in gamma-glutamyltransferase and 5'nucleotidase (Table 1). A paracentesis was performed which showed a milky fluid with a high albumin gradient showing portal hypertension (Table 2), all cultures and cytological studies were negative. Other studies showed normal iron values; antiHIV, HBsAg, antiHBs, antiHBc IgM and antiHCV negative, antiHBc IgG positive; ANA, AMA, SMA y ANCA negative.
A helical CT scan of the abdomen showed mild hepatomegaly, mild enlargement of caudate lobe and ascites. An upper endoscopy ruled out esophageal varices and portal hypertensive gastropathy. A Doppler ultrasound showed a patent portal vein, with a normal diameter (6.6 mm) without alterations in the flow. A cardiac ultrasound showed moderate to severe systolic left ventricle dysfunction with global hypokinesia and the inferolateral wall was akinesic. The left ventricle diameters were normal and the left atria was enlarged (52 mm). The estimated left ventricular ejection fraction was 35%. There were no relevant valve signs, nor mitral or tricuspid regurgitation, nor pericardial disease.
A liver biopsy showed engrossment of centrilobular veins walls and fibrosis, with centrilobular hemorrhage and marked sinusoidal dilatation in acinar zone 3, suggesting increased intrahepatic venous pressure. An angiogram showed dilated but patent suprahepatic and portal veins and inferior vena cava, with slow flow. The measure of free suprahepatic, wedge suprahepatic, pulmonary capillary wedge, pulmonary artery, right ventricle and right atrial pressures confirmed the diagnosis of portal hypertension related to heart failure (Table 3). He began treatment with furosemide 40 mg/day and spironalactone 100 mg/day; he continued treatment with aspirin, digoxin and diltiazem. The patient lost 8 kg and the abdominal distension was resolved approximately 2 month after the beginning of the treatment. Given the good response to treatment we decided to withhold other possible treatments (i.e. angiotensin-converting enzyme inhibitors).
Discussion
Chylous ascites is an unusual type of ascites (<1%) featured by the presence of high concentration of triglycerides in the ascitic fluid (>200 mg/dl). The most common causes in adults are disseminated neoplasia and lymphomas, traumatic or surgical rupture of lymphatic vessels, and less frequently it is associated with cirrhosis. In this last case, its pathogenic mechanism is unknown, but it could be related to degenerative changes related to age (it is most frequently seen in elderly patients) and hypertension in lymphatic vessels caused by portal hypertension [1,5,6].
Liver involvement in chronic heart failure is common and it will depend on the type and severity of cardiac disease. In patients with moderate to severe heart failure, 95% show hepatomegaly, 75% peripheral edema, 20–25% pleural effusion and up to 25% show ascites. Ascitic fluid has a high albumin gradient with a high concentration of proteins, usually more than 3 g/dl. Blood laboratory shows mild increases in ALT and AST (5%), mild elevations in bilirubin (20–80%), high alkaline phosphatase (10–20%), prolonged prothrombin time and decrease in albumin levels (30–50%) [7,8].
These disorders appear with a symptomatic cardiac disease which, in general, was previously diagnosed. Even tough there are cases of heart diseases presenting as liver diseases [9-11], this is the first case of chylous ascites caused by heart failure, with no signs or symptoms of cardiac disorder. There are some cases associated with constrictive pericarditis [12-14] and with severe heart failure [15,16]; all these patients showed signs or symptoms manifesting cardiac involvement (pulmonary edema, jugular vein distension, etc.). There are two mechanisms probably involved: an increase in the abdominal lymph production and an ineffective development of collateral flow. High venous pressure increases the abdominal lymph production due to an augmented capillary filtration and, even tough lymphatic flow increases in response, the augmented central venous pressure reduces lymphatic drainage. Unlike mechanical obstruction of the thoracic duct, where the development of lymphaticovenous collaterals channels provides lymphatic drainage, generalized central venous hypertension caused by cardiac disease prevents the development of an effective collateral flow. Given that heart failure is a common disorder and chylous ascites is a very unusual one, other unknown mechanisms should participate in its pathogenesis.
In this case, other probable causes were ruled out: the long evolution (more than a year) and various images studies ruled out the presence of a hidden neoplasia; there is no recent traumatic or surgical history; the angiogram ruled out vascular liver disease and liver biopsy didn't show any specific liver disease or cirrhosis.
Taking into account all the performed diagnostic procedures and the good response to diuretic treatment, left ventricular dysfunction is the main probable cause. Historically, chylous ascites caused by neoplasia or lymphoma, was associated with poor prognosis and was difficult to treat. Other causes of chylous ascites should be kept in mind, because the prognosis and the response to treatment depend on the disease that causes it.
Conclusion
Chylous ascites is an uncommon type of ascites and has various causes. Asymptomatic heart dysfunction should be kept in mind as a possible cause of chylous ascites and, also, as a cause of liver disease of unknown etiology.
Competing interests
The author(s) declare that they have no competing interests.
Authors' contributions
ER and OGM have made substantial contributions to acquisition of data, analysis and interpretation of data; have been involved in drafting the manuscript or revising it critically for important intellectual content; and have given final approval of the version to be published. Each author has participated sufficiently in the work to take public responsibility for appropriate portions of the content. All authors read and approved the final manuscript.
Pre-publication history
The pre-publication history for this paper can be accessed here:
Figures and Tables
Table 1 Laboratory values.
Variable On admission (reference value)
Hematocrit (%) 49.5
Hemoglobin (g/dl) 15.7
Erythrocyte sedimentation rate (mm/h) 2 (3–10)
White cell count (per mm3) 9800 (4000–10000)
Platelet count (per mm3) 219000 (150000–400000)
Prothrombin time (%) 99 (70–100)
Bilirubin total (mg/dl) 0.94 (<1.2)
Bilirubin direct (mg/dl) 0.26 (<0.36)
AST (UI/L) 12 (<23)
ALT (UI/L) 13 (<25)
Alkaline phosphatase (UI/L) 159 (<207)
Gamma-glutamyltransferase (UI/L) 89 (<38)
5'nuceotidase (UI/L) 20 (<9)
Total proteins (g/dl) 7.8 (6.6–8.7)
Albumin (g/dl) 4.7(3.5–5)
LDH (UI/L) 191 (160–320)
Creatinine (mg/dl) 1.17 (<1.2)
Urea (mg/dl) 39 (<50)
Cholesterol total (mg/dl) 200 (<220)
Triglycerides (mg/dl) 135 (<190)
Table 2 Laboratory values in ascitic fluid.
Variable
Aspect Milky, turbid
Proteins (g/dl) 5.6
Albumin (g/dl) 3.4
Albumin gradient (g/dl) 1.3
Glucose (mg/dl) 123
LDH (UI/L) 115
Cholesterol (mg/dl) 141
Triglycerides (mg/dl) 814
Leukocyte count (per mm3) 3000
Neutrophil count <10%
Table 3 Pressures (mm Hg) recorded during catheterisation
Systolic Diastolic Media
Free suprahepatic 13 10 12
Wedge suprahepatic 16 13 15
Pulmonary capillary wedge 30 14 22
Pulmonary artery 43 17 27
Right ventricle 42 11 24
Right atrium 22 10 17
==== Refs
McHutchinson JG Differential diagnosis of ascites Semin Liver Dis 1997 17 191 202 9308124
Arroyo V Ginés P Planas R Rodés J Birchner J, Benhamou JP, McIntyre N, Rizzetto M, Rodes J Pathogenesis, diagnosis and treatment of ascites in cirrhosis Oxford textbook of Clinical Hepatology 1999 2 Oxford Medical Publications 697 731
Runyon B Feldman M, Scharschmidt BF, Sleisenger MH Ascites and spontaneous bacterial peritonitis Sleisinger and Fordtran's Gastrointestinal and Liver Disease 1998 6 Philadelphia: Saunders 1310 33
Mc Intyre N Collins P Birchner J, Benhamou JP, McIntyre N, Rizzetto M, Rodés J The liver in cardiovascular and pulmonary disease Oxford textbook of Clinical Hepatology 1999 2 Oxford Medical Publications 1671 83
Press OW Ottman Press N Kaufman SD Evaluation and management of chylous ascites Ann Intern Med 1982 96 358 64 7059101
Rector WG Spontaneous chylous ascites of cirrhosis J Clin Gastroenterol 1984 6 369 72 6481122
Braunwald E Colucci WS Grossman W En Brauwald E Clinical aspects of heart failure Heart disease 1997 W Saunders 445 70
Kibo SH Walter BA John PHA Clark M Cody RJ Liver function abnormalities in chronic heart failure. Influence of systemic hemodynamics Arch Intern Med 1987 147 1227 30 3606280 10.1001/archinte.147.7.1227
Lowe MD Harcombe AA Grace AA Petch MC Restrictive constrictive heart failure masquerading as liver disease BMJ 1999 318 585 6 10037641
Hoffman BJ Pate MB Marsh WH Lee WM Cardiomyopathy unrecognized as a cause of hepatic failure J Clin Gastroenterol 1990 12 306 9 2362100
Van der Merwe S Dens J Daenen W Desmet V Fevery J Pericardial disease is often not recognised as a cause of chronic severe ascites J Hepatol 2000 32 164 169 10673082 10.1016/S0168-8278(00)80204-4
Savage MP Muñoz SJ Herman WM Kusiak VM Chylous ascites caused by constrictive pericarditis Am J Gastroenterol 1987 82 1088 1090 3661521
Guneri S Nazli C Kinay O Kirimli O Mermut C Hazan E Chylous ascites due to constrictive pericarditis Int J Card Imaging 2000 16 49 54 10832625 10.1023/A:1006379625554
Williams AT Schneider RP Chylous ascites should suggest constrictive pericarditis even in a patient with cirrhosis J Clin Gastroenterol 1990 12 581 584 2230003
Villena V de Pablo A Martin-Escribano P Chylothorax and chylous ascites due to heart failure Eur Respir J 1995 8 1235 1236 7589411 10.1183/09031936.95.08071235
Hurley MK Emiliani VJ Comer GM Patel A Navarro C Maiki CO Dilated cardiomyopathy associated with chylous ascites Am J Gastroenterol 1989 84 1567 1569 2596459
|
16078991
|
PMC1192795
|
CC BY
|
2021-01-04 16:03:27
|
no
|
BMC Gastroenterol. 2005 Aug 3; 5:25
|
utf-8
|
BMC Gastroenterol
| 2,005 |
10.1186/1471-230X-5-25
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-1061609115210.1186/1471-2164-6-106Research ArticleCross genome phylogenetic analysis of human and Drosophila G protein-coupled receptors: application to functional annotation of orphan receptors Metpally Raghu Prasad Rao [email protected] Ramanathan [email protected] National Centre for Biological Sciences, Tata Institute of Fundamental Research, UAS-GKVK Campus, Bellary Road, Bangalore 560065, INDIA2005 10 8 2005 6 106 106 21 3 2005 10 8 2005 Copyright © 2005 Metpally and Sowdhamini; licensee BioMed Central Ltd.2005Metpally and Sowdhamini; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
The cell-membrane G-protein coupled receptors (GPCRs) are one of the largest known superfamilies and are the main focus of intense pharmaceutical research due to their key role in cell physiology and disease. A large number of putative GPCRs are 'orphans' with no identified natural ligands. The first step in understanding the function of orphan GPCRs is to identify their ligands. Phylogenetic clustering methods were used to elucidate the chemical nature of receptor ligands, which led to the identification of natural ligands for many orphan receptors. We have clustered human and Drosophila receptors with known ligands and orphans through cross genome phylogenetic analysis and hypothesized higher relationship of co-clustered members that would ease ligand identification, as related receptors share ligands with similar structure or class.
Results
Cross-genome phylogenetic analyses were performed to identify eight major groups of GPCRs dividing them into 32 clusters of 371 human and 113 Drosophila proteins (excluding olfactory, taste and gustatory receptors) and reveal unexpected levels of evolutionary conservation across human and Drosophila GPCRs. We also observe that members of human chemokine receptors, involved in immune response, and most of nucleotide-lipid receptors (except opsins) do not have counterparts in Drosophila. Similarly, a group of Drosophila GPCRs (methuselah receptors), associated in aging, is not present in humans.
Conclusion
Our analysis suggests ligand class association to 52 unknown Drosophila receptors and 95 unknown human GPCRs. A higher level of phylogenetic organization was revealed in which clusters with common domain architecture or cellular localization or ligand structure or chemistry or a shared function are evident across human and Drosophila genomes. Such analyses will prove valuable for identifying the natural ligands of Drosophila and human orphan receptors that can lead to a better understanding of physiological and pathological roles of these receptors.
==== Body
Background
G protein-coupled receptors (GPCRs) are one of the largest superfamilies of cellular receptor proteins, generally consisting of seven transmembrane helices (TMH) connected by three extracellular and three cytoplasmic loops of varying lengths. Different GPCRs respond to a wide variety of different external stimuli (light, odorants, peptides, lipids, ions, nucleotides etc) and activate a number of different GTP binding proteins (G proteins), there by initiating a wide spectrum of intracellular responses. GPCRs play important roles in cellular signaling networks involving such processes as neurotransmission, taste, smell, vision, cellular metabolism, differentiation and growth, inflammatory and immune responses and secretion. Abnormalities of signaling by GPCRs are the root cause of disorders that affect most tissues and organs in our body, such as color blindness, thrombosis, restenosis, atherosclerosis, hyper functioning thyroid adenoma and nephrogenic diabetes insipidus and precocious puberty. GPCRs are of major importance to the pharmaceutical industry since they play major roles in the pathogenesis of human diseases and are targets for more than half of the current therapeutic agents on the market [1]. Despite the importance of GPCRs in physiology and diseases, only one high-resolution structure has been solved, that of bovine rhodopsin [2]. A majority of the identified GPCRs are with no known ligand specificity (orphan receptors), which presents a challenge for identifying their native ligands and defining their function.
Characterizing the role of any GPCR involves the identification of both the activating ligand and the activated G protein. A diverse range of procedures have led to the identification of ligands for orphan receptors: (1) identifying relationship between receptor and ligand expression patterns [3], (2) testing tissue extracts in receptor-based functional assays and (3) testing ligands for identified GPCRs on orphan GPCRs with high sequence identity [4] and in some cases randomly evaluating orphan GPCRs against arrayed families of known ligands. The physiological role of these receptors can be well understood by the identification of natural ligands, which further advance the design of pharmacologically active surrogate activators or inhibitors of the GPCRs that have defined native ligands. Strategies described above will be facilitated by better prediction of ligand structure or chemical class of orphan GPCRs.
Proteins similar in sequence often exhibit similar functions. Therefore, sequence homology can be used as a primary criterion for functional screening. This powerful principle can be extended to proteins that are homologous in different species. This has led to the identification of many new novel GPCRs across different species [5]. Many orphan GPCRs are conserved among different species suggesting that they should be active and thus bind novel ligands. This led to the idea that orphan GPCRs could be used as targets to identify their natural ligands and consequently led to the discovery of novel transmitters [6]. Those orphan receptors that share more than 45 percent of sequence identity with the GPCRs with known ligands are very likely to also share common ligands [5]. Often, the direct association of ligand class to orphan receptors is non-trivial by simple BLAST searches even at high sequence identity [7]. The top ranking hits constitute GPCRs from diverse ligand classes (Metpally and Sowdhamini unpublished results) and may not suggest a consensus on possible ligand class to be inferred directly. However, if the sequence identity is below the twilight zone (less than 30 percent), predictions using direct sequence search methods often fail. Phylogenetic tree building has shown that receptors that respond to the same, or similar, agonists often cluster together, even with low sequence identity. For example, most members of the prostanoid receptor subfamily share less than 30 percent amino acid identity, yet these receptors are more like one another than any other GPCR [8]. Phylogenetic clustering methods were used to elucidate the chemical nature of receptor ligands, which led to the identification of natural ligands for many orphan receptors [9-14].
GPCRs were previously classified into distinct families by different groups [14-18]. The classifications would include rhodopsin-like receptors, secretin receptor-like receptors, metabotropic glutamate-like receptors, adhesion-like receptors and frizzled/smoothened-like receptors as proposed by Fredriksson and coworkers [16]; in addition, other groups have proposed two more classes, viz., the fungal pheromone receptor like family and cyclic AMP receptors family [17,18]. These classification schemes were generated mostly from individual genome studies [12,16].
Studies in model organisms and cross-genome comparisons have provided major insights in the general understanding of numerous genes and pathways involved in a wide variety of physiological processes and human diseases [19]. Drosophila is a very good model organism owing to the simplicity in the genetic system and a short lifespan enabling the screening of large individuals to identify mutations in new candidate genes that may have human counterparts involved in cellular physiology and diseases [20]. Despite disparity in morphology or phenotype, Drosophila shows similarity with humans in developmental and cellular processes like core aspects of cell cycle, signaling pathways, apoptosis, neuronal signaling, cytoskeleton and core proteome (including main protein domains and families) [21]. We, therefore, sought out to adopt Drosophila GPCRs to study human gene function using comparative genomics [21-23].
A large number of Drosophila GPCRs have no characterized ligands. On the other hand, many human GPCRs are well characterized in their physiology and pharmacology. In this study, we collected a large set of GPCR sequences from human and Drosophila genomes and performed cross-genome multiple phylogenetic analyses. Further analysis reveals unexpected levels of similarity between GPCRs of these two species and phylogenetic association could be employed to predict ligands (chemical structure or class and/or functions) for many of Drosophila and human orphan receptors.
Results and discussion
Cross genome phylogenetic analysis of human and Drosophila non-olfactory receptors resulted in eight major groups. They are i) peptide receptors, ii) chemokine receptors, iii) nucleotide and lipid receptors iv) biogenic amine receptors v) secretin receptors vi) glutamate receptors vii) cell adhesion receptors and viii) frizzled receptors. These were further classified into 32 clusters (Table 1) with eleven clusters of peptide receptors, two clusters of chemokine receptors, six clusters of nucleotide and lipid receptors, five clusters of biogenic amine receptors, two clusters of secretin receptors, four clusters of glutamate receptors and one cluster each of cell adhesion and frizzled receptors (The combined phylogenetic and ligand analyses of human-Drosophila GPCRs are shown in Figures 1, 2, 3, 4, 5, 6, 7, 8, 9). About thirty one GPCR sequences could not be assigned to any of these clusters; these are discussed separately below as unassociated GPCRs. Our method sometimes resulted in clusters with members whose ligands belong to different chemical structure or classes and these results are discussed in detail below.
Table 1 List of GPCRs in each of the 32 clusters derived from phylogenetic analysis. Suffix _Hum and _Dro refers to human and Drosophila sequences respectively. Orphan receptors are shown in bold.
CLUSTER 1 CLUSTER 2 CLUSTER 3 CLUSTER 4 CLUSTER 5 CLUSTER 6 CLUSTER 7 CLUSTER 8
GALR_Hum GPR7_Hum AG22_Hum BRS3_Hum GHSR_Hum Q9VFW6_Dro GP19_Hum FSHR_Hum CCKR_Hum
GALS_Hum GPR8_Hum AG2R_Hum ET1R_Hum GP39_Hum Q9VP15_Dro GRHR_Hum LGR4_Hum GASR_Hum
GALT_Hum OPRD_Hum APJ_Hum ETB2_Hum MTLR_Hum Q9VT27_Dro GRR2_Hum LGR5_Hum NFF1_Hum
GP24_Hum OPRK_Hum BRB1_Hum ETBR_Hum NTR1_Hum Q9W025_Dro GRHR_Dro LGR6_Hum NFF2_Hum
Q969F8_Hum OPRM_Hum BRB2_Hum GP37_Hum NTR2_Hum Q9W027_Dro OXYR_Hum LGR7_Hum OX1R_Hum
Q969V1_Hum OPRX_Hum GP15_Hum GRPR_Hum NMU1R_Hum Q9W4H3_Dro GRHRII_Dro LGR8_Hum OX2R_Hum
Q9NBC8_Dro Q8I943_Dro GP25_Hum NMBR_Hum NMU2R_Hum TRFR_Hum Q8ITD2_Dro LSHR_Hum Q14439_Hum
Q9U721_Dro Q8ISJ9_Dro Q8NGZ8_Hum Q8TDV0_Hum Q8ITC7_Dro Q8NGU9_Hum Q8SX01_Dro GPR103_Hum
SAPR_Hum SSR1_Hum Q9V858_Dro Q8ITC9_Dro V1AR_Hum Q9NDI1_Dro Q8MKU0_Dro
UR2R_Hum SSR2_Hum Q9V9K3_Dro Q8SWR3_Dro V1BR_Hum Q9VBP0_Dro Q9VWQ9_Dro
SSR3_Hum Q9V5T1_Dro V2R_Hum Q9VYG0_Dro Q9VWR3_Dro
SSR4_Hum Q9VDC4_Dro TSHR_Hum
SSR5_Hum Q9VFW5_Dro
CLUSTER 9 CLUSTER 10 CLUSTER 11 CLUSTER 12 CLUSTER 13 CLUSTER 14 CLUSTER 15 CLUSTER 16 CLUSTER 17
C3AR_Hum MAS_Hum GP10_Hum C3X1_Hum ADMR_Hum OPN3_Hum CLT1_Hum GP34_Hum ACTR_Hum
C5AR_Hum MRG_Hum GP72_Hum CKD6_Hum CCR3_Hum OPN4_Hum CLT2_Hum H963_Hum CB1R_Hum
C5L2_Hum MRGF_Hum NK1R_Hum CKR1_Hum CCR4_Hum OPS1_Dro GP17_Hum P2YC_Hum CB2R_Hum
CML1_Hum Q8NGK7_Hum NK2R_Hum CKR2_Hum CCR5_Hum OPS2_Dro GP31_Hum P2YX_Hum EDG2_Hum
FML1_Hum Q8TDD6_Hum NK3R_Hum CKR3_Hum CCR6_Hum OPS3_Dro GP40_Hum PAFR_Hum EDG3_Hum
FML2_Hum Q8TDD8_Hum NK4R_Hum CKR4_Hum CKR6_Hum OPS4_Dro GP41_Hum Q8TDU7_Hum EDG4_Hum
FMLR_Hum Q8TDE0_Hum NY1R_Hum CKR5_Hum CKR7_Hum OPS5_Dro GP43_Hum Q96JZ8_Hum GP12_Hum
GP32_Hum Q96LB1_Hum NY2R_Hum CKR8_Hum CKRA_Hum OPS6_Dro GP82_Hum GPR3_Hum
GP44_Hum NY4R_Hum CXC1_Hum CKRB_Hum OPSB_Hum HM74_Hum GPR6_Hum
GPR1_Hum NY5R_Hum O75307_Hum CML2_Hum OPSD_Hum P2Y2_Hum MC3R_Hum
L4R1_Hum NYR_Dro DUFF_Hum OPSG_Hum P2Y4_Hum MC4R_Hum
L4R2_Hum PKR1_Hum IL8A_Hum OPSX_Hum P2Y6_Hum MC5R_Hum
Q8NGA4_Hum PKR2_Hum IL8B_Hum Q96FC5_Hum P2YB_Hum O95136_Hum
Q8TDT2_Hum Q8SZ35_Dro CKR9_Hum Q9VTU7_Dro P2YR_Hum O95977_Hum
NY6R_Hum Q96CH1_Hum Q8TDQ8_Hum Q8WUL7_Hum
Q9VRM0_Dro RDC1_Hum Q8TDS5_Hum Q9H228_Hum
Q9VW75_Dro Q96P68_Hum Q9NRB8_Hum
Q9W189_Dro Q9BXC0_Hum Q9NYN8_Hum
TLR1_Dro
TLR2_Dro
CLUSTER 18 CLUSTER 19 CLUSTER 20 CLUSTER 21 CLUSTER 22 CLUSTER 23 CLUSTER 24
O00325_Hum EBI2_Hum 5H4_Hum ML1A_Hum 5H1A_Hum AA1R_Hum GP63_Hum 5H2A_Hum HH2R_Hum
O75228_Hum FK79_Hum O14804_Hum ML1B_Hum 5H1B_Hum AA2A_Hum GP85_Hum 5H2B_Hum O61730_Dro
PD2R_Hum GP18_Hum Q969N4_Hum ML1X_Hum 5H1D_Hum AA2B_Hum HH1R_Hum 5H2C_Hum O97171_Dro
PE21_Hum GP20_Hum Q96RI8_Hum O77269_Dro 5H1E_Hum AA3R_Hum HH3R_Hum 5H6_Hum OAR_Dro
PE22_Hum GP35_Hum Q96RI9_Hum O77270_Dro 5H1F_Hum ACM1_Dro HH4R_Hum A1AB_Hum Q13675_Hum
PE24_Hum GP68_Hum Q96RJ0_Hum Q9NQS5_Hum 5H5A_Hum ACM1_Hum O43898_Hum A1AD_Hum Q8IPN2_Dro
PF2R_Hum GPR4_Hum Q9P1P4_Hum 5H7_Hum ACM2_Hum Q8NDV2_Hum A2AA_Hum Q8IS45_Dro
PI2R_Hum O75819_Hum Q9P1P5_Hum 5HT1_Dro ACM3_Hum Q8TDV4_Hum A2AB_Hum Q8N6U8_Hum
Q9VVJ1_Dro P2Y5_Hum Q9VCZ3_Dro 5HTA_Dro ACM4_Hum Q9VAA2_Dro A2AC_Hum Q8NGU3_Hum
P2Y9_Hum Q9VG54_Dro 5HTB_Dro ACM5_Hum Q9VHW1_Dro B1AR_Hum Q8TDV5_Hum
P2YA_Hum Q16538_Hum GP21_Hum Q9VMI4_Dro B2AR_Hum Q96P66_Hum
PAR1_Hum Q8TDV2_Hum GP27_Hum SRB3_Hum B3AR_Hum Q9GZN0_Hum
PAR2_Hum Q9VEG1_Dro GP52_Hum D3DR_Hum Q9NZR3_Hum
PAR3_Hum Q9VEG2_Dro GP62_Hum D4DR_Hum Q9VBG4_Dro
PAR4_Hum DADR_Hum Q9VE32_Dro
Q8N580_Hum DBDR_Hum Q9VHP6_Dro
Q9H1C0_Hum DOP1_Dro Q9W3V5_Dro
Q9UNW8_Hum DOP2_Dro
CLUSTER 25 CLUSTER 26 CLUSTER 27 CLUSTER 28 CLUSTER 29 CLUSTER 30 CLUSTER 31 CLUSTER 32
CALR_Hum MTH_Dro BAI1_Hum Q8NG96_Hum MGR_Dro CASR_Hum O75205_Hum GBR1_Hum FRIZ_Dro
CGRR_Hum MTH1_Dro BAI3_Hum Q8NGA7_Hum MGR1_Hum Q8NGV9_Hum O95357_Hum GBR2_Hum FRZ2_Dro
CRF2_Hum MTH2_Dro CD97_Hum Q8NGB3_Hum MGR2_Hum Q8NGW9_Hum Q9NQ84_Hum Q8NFN8_Hum FRZ3_Dro
GIPR_Hum MTH3_Dro CLR1_Hum Q8NGW8_Hum MGR3_Hum Q8NGZ7_Hum Q9NZD1_Hum Q9BML5_Dro FRZ4_Dro
GLP1_Hum MTH4_Dro CLR2_Hum Q8NH12_Hum MGR4_Hum Q8NHZ9_Hum BOSS_Dro Q9BML7_Dro FZ10_Hum
GLP2_Hum MTH5_Dro CLR3_Hum Q8SZ78_Dro MGR5_Hum Q9V3Q9_Dro FZD1_Hum
GLR_Hum MTH6_Dro O94910_Hum Q8T4B2_Dro MGR6_Hum Q9VKA4_Dro FZD2_Hum
GRFR_Hum MTH7_Dro O95490_Hum Q8WXG9_Hum MGR8_Hum Q9VNZ5_Dro FZD3_Hum
PACR_Hum MTH8_Dro Q8IXE3_Hum Q96JW0_Hum Q8NFS4_Hum Q9VR40_Dro FZD4_Hum
PTR2_Hum MTH9_Dro Q8IZF1_Hum Q96K78_Hum Q9V4U4_Dro Q9Y133_Dro FZD5_Hum
PTRR_Hum MTHA_Dro Q8IZF2_Hum Q96PE1_Hum FZD6_Hum
Q8NG71_Hum MTHC_Dro Q8IZF3_Hum Q9BY15_Hum FZD7_Hum
Q8NHB4_Hum Q8INM0_Dro Q8IZF4_Hum Q9HAR2_Hum FZD8_Hum
Q9V6C7_Dro Q8IPD0_Dro Q8IZF5_Hum Q9HBW9_Hum FZD9_Hum
Q9V6N4_Dro Q8IZF6_Hum Q9V4V8_Dro SMO_Dro
Q9V716_Dro Q8IZF7_Hum STAN_Dro SMO_Hum
SCRC_Hum Q8IZP9_Hum
VIPR_Hum
VIPS_Hum
Figure 1 Phylogenetic trees of peptide receptors (clusters 1–11). Trees were inferred as described in Methods (using TREE-PUZZLE 5.1 corrected using JTT substitution frequency matrix. Quartet-puzzling support percentage values from 10,000 puzzling steps are shown). Out-group not showed in the figure. The scale bars indicate a maximum likelihood branch length of 0.1 inferred substitutions per site. Orphan receptors are shown in bold letters. Cluster numbers are marked in the top left corner.
Figure 2 Representative multiple sequence alignment of GPCR clusters. GPCR sequences of ET1R_Hum, ETAR_Hum, ETBR_Hum, ETB2_Hum, GRPR_Hum, NMBR_Hum, BRS3_Hum, GP37_Hum, Q8TDV0_Hum, Q9V858_Dro and Q9V9K3_Dro belonging to cluster 4 were aligned with ClustalX. Sequence region comprising of TMH-1 to TMH-7 alone were considered for the analysis (Alignment was modified by deleting the extremely variable amino termini upstream of the first transmembrane helix and carboxyl termini downstream of the seventh transmembrane helix). Identical amino-acid residues in all aligned sequences are shaded in black and similar residues in gray and consensus residues are indicated below. Transmembrane helices (TMH) identified by the HMMTOP program are indicated.
Figure 3 Phylogenetic trees of chemokine receptors (clusters 12 and 13). The mode of deriving phylogenetic trees is as described in Methods and indications are as in Figure 2.
Figure 4 Phylogenetic trees of nucleotide and lipid receptors (clusters 14–19). The mode of deriving phylogenetic trees is as described in Methods and indications are as in Figure 2.
Figure 5 Phylogenetic trees of biogenic amine receptors (clusters 20–24). The mode of deriving phylogenetic trees is as described in Methods and indications are as in the Figure 2 except for the cluster 22, where scale bar indicates a maximum likelihood branch length of 1.0 inferred substitutions per site.
Figure 6 Phylogenetic trees of class B (secretin) receptors (clusters 25 and 26). The mode of deriving phylogenetic trees is as described in Methods and indications are as in Figure 2.
Figure 7 Phylogenetic tree of cell adhesion receptors (cluster 27). The mode of deriving phylogenetic tree is as described in Methods and indications are as in Figure 2.
Figure 8 Phylogenetic trees of class C (glutamate) receptors (clusters 28–31). The mode of deriving phylogenetic trees is as described in Methods and indications are as in Figure 2.
Figure 9 Phylogenetic tree of frizzled/smoothened receptors (cluster 32). The mode of deriving phylogenetic tree is as described in Methods and indications are as in Figure 2.
Peptide receptors
Clusters 1 to 11 comprise of peptide receptors (Figure 1). The size of peptide ligands can vary from two amino acids to as many as 50. Some of the natural peptide ligands include apelin, bombesin, calcitonin, endothelin, galanin, gastrin, ghrelin, neurotensin, neuropeptide B, W, Y, orexin, oxytocin, relaxin, somatostatin, urocortins, etc. These receptors are involved in many human diseases including chronic inflammatory diseases, degenerative diseases, autoimmune diseases, cancer, cardiovascular diseases etc, thus they could be of new therapeutic targets [24,25].
Receptors with known ligands in cluster 1 binds to galanins or kisspeptins or cyclic peptides. Drosophila allostatin receptors (DARs) (Q9NBC8_Dro and Q9U721_Dro) are very closely related to galanin receptors [26]. Receptors, Q969V1_Hum and Q96S47_Hum, are closely related to GP24_Hum receptor that bind to melanin-concentrating hormone and may have similar cyclic peptides as their ligands. As the name suggests, orphan receptor, SAPR_Hum, does not bind to somatostatins and angiotensins [27] since it is distantly related to GP24_Hum and UR2R_Hum receptors in this tree. Instead, this receptor may bind to similar cyclic peptides.
Cluster 2 consists of receptors for opioid, somatostatin and neuropeptide (NPB or NPW) ligands forming different branches. Opioids and somatostatins are obtained from preprocessing of larger precursor peptides. It is known that GPR7_Hum and GPR8_Hum bind to NPB/W ligands [28]. Drosophila orphan receptors, Q8ISJ9_DRo and Q8I943_Dro branch is close to somatostatin receptors and might bind to ligands similar to somatostatins. Small peptide (apelin, angiotensin, and bradykinin) receptors comprise of cluster 3. The human orphan receptors encoded by GPR15_Hum, GPR25_Hum and Q8NGZ8_Hum are related to APJ_Hum and show significant amino acid identity suggesting these might bind to small peptide endogenous ligands.
Cluster 4 comprises of endothelin and bombesin receptors with known ligands (ET1R_Hum, ETAR_Hum and ETBR_Hum, gastrin-releasing peptide receptor (GRPR_Hum), the neuromedin B receptor (NMBR_Hum) and bombesin receptor (BRS3)). Drosophila orphan receptors, Q9V9K3_Dro and Q9V858_Dro, share the branch with bombesin, GRPR and NMBR receptors. They share many conserved amino acids, known to be important for high affinity binding of gastrin-releasing peptide (GRP) and bombesin to GRPR and NMB binding to NMB-R [29-31] (Figure 2). This suggests Q9V9K3_Dro and Q9V858_Dro might bind to similar neuropeptide(s) for its activation. Human orphan receptor GPR37_Hum is closely related to ETB2_Hum suggesting it may bind to endothelin-like peptides. Q8TDV0_Hum is sequentially similar to both galanin (cluster 1) and bombesin receptors but sub-clustering of peptide receptors by maximum likelihood method has placed it in this cluster suggesting closer association of these two clusters.
Cluster 5 is composed of receptors for neurotensin (NT), neuromedin U (NMU), motilin, growth hormone secretagogue, thyrotropin-releasing hormone and some of PRX-amide peptides. GPR39_Hum is closely related to NT receptors and might bind to neurotensin ligands. Drosophila receptors, Q8ITC7_Dro, Q9VFW5_Dro, Q9VFW6_Dro, Q8ITC9_Dro and Q9VP15_Dro form a separate branch, which are closely related to vertebrate neuromedin receptors and they bind to PRXa pyrokinins or FXPRXamide or Cap2b-like peptides (FPRXamide) or ecdysis triggering hormones (PRXamide) (Park et al. 2002). Q9VDC4_Dro forms a distinct branch and is sequentially close to GHSR_Hum, TRFR_Hum, Q8ITC7_Dro and Q9VFW5_Dro and might bind to neuropeptides. Drosophila orphan receptors, Q9W4H3_Dro, Q9VT27_Dro, Q8SWR3_Dro, Q9V5T1_Dro, Q9W025_Dro and Q9W027_Dro, branch out from that of TRFR_Hum and might form a separate family of receptors binding to novel neuropeptide ligands. Supporting our analysis, Q9W025_Dro and Q9W027_DRo were reported as first receptors specific for Drosophila myosuppressins (Drome-MS) [32] and Q9W4H3_Dro was reported as neuropeptide proctolin binding receptor [33]. Q9VT27_Dro is very closely related to Q9W4H3_Dro and might bind to proctolin or similar neuropeptide ligands for its activation.
Cluster 6 consists of peptide hormone receptors binding arginine vasopressin (AVP) or growth hormone releasing hormone or oxytocin or gonadotropin-releasing hormone II or crustacean cardioactive peptide (CCAP) or corazonin or adipokinetic hormone (AKH) (Park et al. 2002). GP19_Hum is related to Drosophila CCAP receptor (Q8ITD2_Dro) that is activated by CCAP and AKH, but not by AVP. Thus, CCAP and AKH might as well bind to GP19_Hum for its activation. Drosophila gonadotropin-releasing hormone and/or corazonin receptor (GRHR_Dro) and putative corazonin (GRHR II) receptor clusters well with human counterparts (GRHR_Hum and GRR2_Hum) suggesting early evolution of GRHR receptors. Q8NGU9_Hum forms a separate branch, but shares sequence similarity with AVP receptors and might bind to similar neuropeptide ligands.
Cluster 7 comprises leucine-rich repeat-containing G protein-coupled receptors (LGR) like glycoprotein receptors, follicle stimulating hormone receptor (FSHR_Hum), thyroid-stimulating hormone receptor (TSHR_Hum), luteinizing hormone receptor (LSHR_Hum) and receptors binding to relaxin. These are unique in having a large N-terminal extracellular (ecto) domain containing leucine-rich repeats important for interaction with the glycoprotein ligands and are classified into three sub-groups [34]. Our analysis also shows that there are three LGR subfamilies: (i) the glycoprotein hormone receptors LSHR_Hum, FSHR_Hum, TSHR_Hum, Q8SX01_Dro and Q9NDI1_Dro (ii) LGR4_Hum LGR5_Hum and LGR6_Hum (iii) LGR5_Hum, LGR7_Hum and LGR8_Hum, Q9VBP0_Dro, and Q9VYG0_Dro. Drosophila orphan receptors Q8SX01_Dro and Q9NDI1_Dro are closely related to human glycoprotein hormone receptors and might bind to glycoprotein hormones. Q9VBP0_Dro and Q9VYG0_Dro are very similar in their overall domain architecture to LGRs with long N-termini, but their similar relationship in extracellular domain arrangements are also evident from this phylogenetic analysis without considering the N and C termini.
Cluster 8 consists of peptide receptors with known ligands such as gastrin (GAS), cholecystokinin (CCK), orexin (OXR) and neuropeptide FF (NFF) or morphine modulating peptides. GPR103_Hum (Q96P65) is closely related to neuropeptide FF receptors, as predicted by our phylogenetic analysis and previous prediction on human GPCRs [12]. Subsequently, GPR103 was characterized and a novel RF-amide peptide, P52 was shown to be its ligand [35]. Drosophila orphan receptors, Q9VWR3_Dro (CCKLR-17D1) and Q9VWQ9_Dro (CCKLR-17D3), are related to each other and branch off from the cholecystokinin (CCK) receptors and might have cholecystokinin as its natural ligand. Q14439_Hum branch off orexin receptors that bind to two novel neuropeptides, orexin-A and B, derived from a common prepro-orexin precursor by proteolytic processing [36].
The receptors with known ligands binding to chemotactic substances (hydrophilic peptides, N-formyl-methionyls (FML) and anaphylactic complement factors) are part of cluster 9. These ligands are structurally very diverse but functionally related peptides. Human orphan receptors, GP32_Hum and Q8NGA4_Hum branch out early from FML receptors and may probably bind to smaller hydrophilic peptides. L4R1_Hum, L4R2_Hum and Q8TDT2_Hum form a separate branch distant from other chemotactic peptide receptors with out bootstrap support. CML1_Hum and GPR1_Hum form a separate branch distinct from the other branches, and also GPR44_Hum forming an individual branch. Prediction of ligands for these receptors is not possible using this phylogenetic tree, but these receptors may be activated by chemotactic substances [37].
Mas proto-oncogene, Mas-related genes (MRGs) and sensory neuron-specific G protein-coupled receptors (SNSRs) form cluster 10. Angiotensin (1–7) has been identified as an endogenous ligand for the G protein-coupled receptor Mas [38]. SNSRs are activated by proenkephalin A peptide fragments, like bovine adrenal medulla peptide 22 (BAM22). Some MRGs and SNSRs are expressed in nociceptive sensory neurons suggesting that they could be involved in pain sensation or its modulation. Previous studies also suggest that ligands for MRG receptors may include neuropeptides that modulate pain sensitivity [39]. Human orphan receptor Q8NGK7_Hum is closely related to MRG receptor.
All receptors with known ligands in cluster 11 are neuropeptide receptors. Drosophila tachykinin-like peptide receptors (TLR1_Dro and TLR2_Dro) and human neurokinin receptors (NK1-4R_Hum) form a closely-knit branch. PKR1_Hum (Q8NFJ7) and PKR2_Hum (Q8NFJ6) form a separate branch of receptors that bind to prokineticins [40]. Q9VRM0_Dro is closely related to Drosophila receptor NYR_Dro that bind to neuropeptide Y. Q9VRM0_Dro might probably bind to similar neuropeptides. Neuropeptide Y binding receptors (NY1R_Hum, NY4R_Hum, NY5R_Hum and NY6R_Hum (Q99463)) form a separate branch. The human prolactin-releasing peptide (PrRP) binding GPR10_Hum forms a separate branch in this phylogenetic tree [41]. Drosophila orphan receptors, Q9VW75_Dro and Q8SZ35_Dro constitute a separate branch close to other neuropeptide receptors that might functionally be activated by neuropeptides. Similarly, orphan receptor GP72_Hum forms a new branch. Drosophila orphan receptor Q9W189_Dro is a very distantly related member and was only grouped into this cluster by blastp results.
Chemokine receptors
Chemokine receptors are phylogenetically represented by two clusters 12 and 13 (Figure 3). Chemokines are important molecules in inflammatory responses, as immunomodulators and they also have critical functions in lymphopoiesis [42]. There are no Drosophila members belong to this group of receptors suggesting these receptors might be recent in evolutionary origin. They have been divided into two subfamilies on the basis of the arrangement of the two disulphide-bond forming N-terminal cysteine residues, CXC and CC. Many human CXC chemokines that mainly act on neutrophils are clustered at chromosome 4q12–13, while many CC chemokines that mainly act on monocytes are located in another cluster at chromosome 17q11.2. Our phylogenetic analysis has also divided chemokine receptors into two major clusters, concurrent with that of chemokine classes, suggesting co-evolution of receptors and ligands [43].
Cluster 12 consists of receptors associated with CC type chemokines. As reported previously through earlier approach [12] O75307_Hum (CRAM-A) might bind to CC-type chemokine ligand. Cluster 13 consists of both CXC and CC-type receptors. ADMR_Hum and Q8NE10_Hum (RDC1) form a branch whereas Duff antigen and Q96CH1_Hum are distantly related to CML2_Hum. These two branches are associated to chemokine receptors based on BLASTP similarity at an E-value significance of 5e-04 and 7e-07, respectively, with other members of this cluster.
Nucleotide and lipid receptors
Nucleotide and lipid receptors consists of six clusters (Figure 4), except for cluster 14 (opsins) and cluster 18 (receptors binding ligands are derivatives of arachidonic acid) there are no counter parts from Drosophila. Opsins are included in cluster 14 that are activated by isoprenoid ligands. Drosophila opsins show significantly high homology to human opsins. There is strong conservation of the retinal binding site and other regions suggesting that they are derived from a common ancestor and diverged thereafter retaining structural and functional features [44]. Drosophila receptor Q9VTU7_Dro is closely related to OPS3–5_Dro receptors, which are localized in the inner-cells of the Drosophila eye (either R7 or R8 cells). This suggests Q9VTU7_Dro might be localized in the inner cells of Drosophila eye.
Receptors for pyramidine or purine nucleotides, cysteinyl leukotriene, nicotinic acid (niacin; pellagra preventing factor) and short, medium and long chain fatty acids make up cluster 15. Q9BXC0_Hum (GPR81), Q8TDS5_Hum and GP31_Hum share the branch with closely related nicotinic acid (HM74_Hum) receptor [45] and might have similar carboxylic acids as their ligands. Q8TDQ8_Hum and Q96P68_Hum are related to each other as well as to P2Y receptors and may bind to P2Y nucleotides. GP17_Hum and GP82_Hum receptors are distantly related to other members in this cluster and might represent potential new subfamilies binding to nucleotide or lipids.
Cluster 16 is a heterogeneous group of receptors binding to lipids, nucleotides, modified nucleotides and platelet activating factor (PAF). Orphan receptor Q8TDU7_Hum (GPR86) is closely related to platelet ADP-binding receptor (P2YC_Hum). Q96JZ8_Hum (GPR87) is closely related to UDP-glucose receptor (P2YX_Hum) and might bind to a modified nucleotide ligand. GPR34_Hum forms a separate branch which is distantly related to PAFR_Hum. No prediction of ligands is possible for GPR34_Hum with this phylogenetic tree.
Cluster 17 consists of lipid receptors (cannabinoids, lysophospholipid sphingosine 1-phosphate (S1P)) and exceptionally some of the peptide receptors (melanocortin peptides derived from processing of pro-opiomelanocortin) are represented in different branches. Although they bind to different ligands, they identify each other during sequence searches and display 23–29% sequence identity. The functionally important motifs are fairly conserved [46] (please see Additional data file 2). Indeed, this unusual branching including peptide and lipid receptors has been noted earlier by Methner's and Fredicksson's groups [12,16].
Cluster 18 is composed of receptors binding to prostaglandins, prostacyclins and thromboxanes. All these ligands are derivatives of arachidonic acid (AA), which serves as the precursor via the cyclooxygenase (COX) pathway. Drosophila orphan receptor Q9VVJ1_Dro within this tree might bind to ligands derived from AA by the action of COX.
Cluster 19 is also a heterogeneous group of receptors consisting of protease-activated receptors, psychosine receptors, lysophosphatidylcholine and sphingosylphosphorylcholine. Ovarian cancer G-protein-coupled receptor 1 (OGR1), previously described as a receptor for sphingosylphosphorylcholine, acts as a proton-sensing receptor stimulating inositol phosphate formation [47], whereas GPR4 is also involved in pH homeostasis, but elicits cyclic AMP formation [48]. OGR1 (GPR68) and GPR4 are different from other sphingosylphosphorylcholine binding endothelial differentiation gene (EDG) receptors. Orphan P2Y receptors in this cluster are misnomers since they do not cluster with the classical neuropeptide receptors (cluster 15 and 16) and instead appear to be co-clustered with members of this heterogeneous cluster. Either they may have uncommon nucleotide(s) as natural ligand or despite their structural similarity to the P2Y family they may not be nucleotide receptors [49]. GP35_Hum and Q8N580_Hum, EBI2_Hum and GP18_Hum and GP20_Hum cluster as separate branches and are distantly related to members of other branches but probably bind to lipids as their natural ligands.
Biogenic amine receptors
Biogenic amine receptors consists of five clusters mainly consisting of trace amine; melatonin; serotonin receptors; histamines, muscarinic acetylcholine, adenosine and histamine; dopamine, octopamine and adrenaline receptors (Figure 5). In these clusters fairly good intermixing of human and Drosophila receptors is observed. This suggests biogenic amine receptors have ancient evolutionary origin as they are observed in invertebrates to higher vertebrates. Cluster 20 is represented mainly by trace amine (TA) receptors (Figure 5). Trace amines binding these receptors are believed to play an important role in human disorders such as depression, attention deficit disorder, schizophrenia and parkinson's disease [50]. They form a subfamily of GPCRs, distinct from, but related to serotonin (5-HT), Norepinephrine (NE) and dopamine (DA) receptors. Drosophila orphan receptors Q9VG54_Dro and Q9VCZ3_Dro are closely related to 5H4_Hum. Q9P1P4_Hum (GPR57) and Q9P1P5_Hum (GPR58) are closely related to Q96RJ0_Hum (TA1). Similarly O14804_Hum, a putative neurotransmitter receptor (PNR) is closely related to trace amine (Q969N4_Hum, Q96RI8_Hum, and Q96RI9_Hum) receptors.
Cluster 21 consists of melatonin receptors (ML1A_Hum, ML1B_Hum and ML1X_Hum) and other related orphan receptors (O77269_Dro, O77270_Dro, and Q9NQS5_Hum). Melatonin receptors bind to and are activated by biogenic amine 5-methoxy-N-acetyltryptamine (melatonin). The melatonin-related receptor (ML1X_Hum), despite sharing considerable amino acid sequence identity with other melatonin receptors, does not bind melatonin [51]. The receptors in this cluster show considerable sequence similarity to neuropeptide Y (NPY) receptors than other biogenic amine receptors and were previously grouped along with NPY receptors [12].
All receptors with known ligands of Cluster 22 consist of serotonin receptors. These are structurally distinct from serotonin receptors in cluster 24. Drosophila orphan receptors Q9VEG1_Dro and Q9VEG2_Dro form a separate branch but are closely related to other serotonin receptors in this tree and might have similar ligand (s) for its activation. Q8TDV2_Hum and Q16538_Hum (Protein A-2), however, are distantly related to other receptors in this tree and were placed only based on BLASTP similarity.
Receptors of biogenic amines (muscarinic acetylcholine, adenosine and histamine) and many orphan receptors are all placed in different branches in cluster 23. Drosophila orphan receptor Q9VHW1_Dro branch out along with muscarinic acetylcholine and histamine receptors in this tree and might bind to acetylcholine or histamines for its activation. Q9VAA2_Dro is closely related to that of adenosine receptors. Super conserved receptors expressed in brain (SRB1-3) from vertebrate species form a separate branch and might represent potential novel subfamily of GPCRs binding to undiscovered endogenous biogenic amine ligands [52]. High-affinity lysophosphatidic acid (LPA) receptor homologs O43898_Hum and GPR63_Hum form a distinct branch. Similarly, orphan receptors GP21_Hum and GP51_Hum, GPR62_Hum and Q8TDV4_Hum, Q8NDV2_Hum (GPR26) and Q8NGV3_Hum and Q9VMI4_Dro form a distinct branch, suggesting only distant relationship with other members of the cluster.
Receptors of biogenic amines (dopamine, histamine, octopamine and adrenaline), few serotonergic receptors and many orphan receptors are represented in different branches in cluster 24. Drosophila dopamine 2-like receptor (DD2R), Q8IS45_Dro, groups well with the human counterparts suggesting that their evolution extends much before Drosophila. Interestingly, DOP2_Dro is grouped with the adrenaline receptors instead with dopaminergic receptors and shows similar sequence identity (40–48%) with vertebrate alpha 1-, and beta-adrenergic, and D1-like, D2-like dopaminergic and serotonergic receptors. This Drosophila receptor has been discussed as a novel structural class of dopamine receptors [53]. Drosophila octopamine receptor isoforms in mushroom bodies (OAMB) (O97171_Dro and O61730_Dro) branch out with human alpha 1 adrenergic (A1A (A, B and D) _Hum) receptors since they share high sequence identity (52–55%) in TM regions with alpha 1 adrenergic receptors [54]. Q9VE32_Dro branches out from human alpha 2 adrenergic receptors and may have adrenaline as its ligand for activation. Orphan striatum-specific G protein-coupled receptor (STRG or Q9GZN0_Hum), though grouped with biogenic amine receptors, may represent a novel subtype of GPCR due to the lack of conservation of key functional residues [55]. Orphan receptors, Q9W3V5_Dro and Q8TDV5_Hum, Q96P66_Hum and Q8N6U8_Hum, Q9VHP6_Dro and Q9VBG4_Dro form their own branch sharing distant relationship with other receptors in this tree and might represent potential novel subfamilies of biogenic amine GPCRs.
Class B (secretin) receptors
Class B receptors are represented by two clusters (25 and 26) consisting of classical hormone receptors and Drosophila methuselah (MTH) like proteins (Figure 6). The ligands for receptors of cluster 25 are structurally related polypeptide hormones of 27–141 amino-acid residues (pituitary adenylate cyclase-activating polypeptide (PACAP), secretin, calcitonin, corticotropin-releasing factor (CRF), urocortins, growth-hormone-releasing hormone (GHRH), vasoactive intestinal peptide (VIP), glucagon, glucagon-like peptides (GLP-1, GLP-2) and glucose-dependent insulinotropic polypeptide (GIP). Drosophila orphan receptors, Q9V716_Dro and Q9V6C7_Dro are closely related to the human receptor for Corticotropin releasing factor receptor (CRF) which binds to urocortins. Q9V6N4_Dro, Q9VYH9_Dro and Q9NEF7_Dro are related to calcitonin (CALR_Hum) and calcitonin gene-related peptide type 1 receptors (CGRR_Hum). Three small accessory proteins, called receptor activity-modifying proteins (RAMPs), interact with these calcitonin receptors and can generate six pharmacologically distinct receptors. If this phenomenon of RAMP-enabled receptor diversity exists in other receptors, then it will further complicate the ligand-receptor interactions of GPCRs, assuming they still bind to structurally similar ligands. Human orphan receptor, Q8NHB4_Hum, is very closely related to PTRR_Hum receptor binding to parathyroid hormone and parathyroid hormone-related protein (PTHrP). Methuselah receptors and its paralogs of Drosophila solely represent cluster 26. The Drosophila mutant methuselah (MTH) was identified from a screen for single gene mutations that extended average lifespan of an organism and also increased resistance to several forms of stress, including starvation, heat, and oxidative damage [56]. There are no obvious homologues of these receptors within human or C. elegans genomes. Drosophila receptors, Q8INM0_Dro, Q8IPD0_Dro and Q95NU7_Dro, are closely related to previously identified MTH members and may be new paralogs of these receptors.
Cell adhesion receptors
Large number of GPCRs with long extracellular N-termini, containing GPCR proteolytic site (GPS) domain, are represented in cluster 27 (Figure 7). Several of these receptors also have one or many functional domains such as epidermal growth factor (EGF), leucine rich repeat (LRR), hormone-binding domain (HBD) and immunoglobulin (Ig) domains [16]. These form several distantly related branches. Except CD97_Hum, all the receptors in this cluster are orphans with no known ligands [57]. There are only four Drosophila sequences representing these receptors.
Class C (glutamate) receptors
Receptors of Class C are divided mainly into four clusters (28–31): metabotropic glutamate receptors (MGR), γ-aminobutryic acid (GABA) receptors, calcium-sensing receptors (CASR) and retinoic acid-inducible G-protein-coupled receptors (RAIG) (Figure 8).
Cluster 28 consists of human and Drosophila MGRs. Human MGRs are sub-grouped into three different branches: first contains MGR1_Hum and MGR5_Hum and second contains MGR2_Hum and MGR3_Hum. The third branch, including MGR4_Hum, 6–8 and Drosophila MGRs represent a separate subgroup [58]. Drosophila orphan receptor Q9V4U4_Dro is closely related to MGR_Dro and might bind to glutamate for its activation.
Calcium-sensing receptor (CASR_Hum) forms cluster 29 along with a set of orphan receptors (Q8NHZ9_Hum, Q8NGV9_Hum, Q8NGW9_Hum and Q8NGZ7_Hum). These orphan receptors either may have ligands and/or function similar to that of CASR_Hum or they may act as pheromone/olfactory receptors. Phylogenetic tree of most members (including olfactory, putative pheromone, and sweet and amino acid taste receptors) of family 3 GPCRs across different genomes (Catfish (Ictalurus punctatus), Caenorhabditis elegans, Drosophila melanogaster, Japanese pufferfish (Fugu rubripes), Goldfish (Carassius auratus), human (Homo sapiens sapiens), mouse (Mus musculus), rat (Rattus norvegicus) and Salmon (Oncorhynchus masou)) have shown CASR_Hum forms a separate branch part of pheromone/olfactory cluster of class C GPCRs [59]. To note that olfactory and gustatory/taste receptors are not considered in this work.
Cluster 30 consists of retinoic acid-inducible G-protein-coupled receptors (RAIG). RAIGs have short (30–50 amino acids) extracellular amino-terminal domains (ATDs) as opposed to the other receptors currently assigned to class C. BOSS_Dro also has short ATD and branch out very early with the members of RAIGs and may represent new single member subfamily of class C receptors.
The GABAB receptors are present in cluster 31. It is represented by four sub-branches, of which three are GABABR1-3_Hum type receptors and fourth sub-branch of Drosophila orphan receptors (Q9VKA4 and Q9VR40) related to that of GABA receptors. GABAB3 is exclusively present in Drosophila as separate branch whose function is not yet known. Previous results have only been able to functionally characterize D-GABABR1 and R2 when the two subtypes are co-expressed either in Xenopus laevis oocytes or mammalian cell lines, whilst D-GABABR3 was inactive in any combination. This suggests D-GABABR3 requires a counterpart other than D-GABABR1 and R2 to form a functional heterodimer [60]. Thus the current clustering approach suggests that Q9VKA4_Dro or Q9VR40_Dro may interact with D-GABABR3 and form a functional heterodimer.
Frizzled/smoothened receptors
Cluster 32 comprises receptors with a long (about 200-amino acid) N-terminus and conserved cysteine rich domains (CRD) which are likely to participate in Wnt ligand binding (Figure 9). These receptors control the specification of cell fate, cell adhesion, migration, polarity and proliferation [61]. This cluster is represented by ten human (FZD1-10) and four Drosophila (FRZ1-4) frizzled receptors together with smoothened (SMO_Hum and SMO_Dro) receptors. The topology of the phylogenetic tree shows one smoothened and four frizzled branches. FRZ1_Dro is closely related to human FZD3_Hum and FZD6_Hum. FRZ2_Dro is related to FZD5_Hum and FZD8_Hum, whereas FRZ3_Hum and FRZ4_Hum form separate branches distantly related to other receptors.
Unassociated GPCRs
Thirty one GPCR sequences could not be included in any cluster with appreciable bootstrap values or BLASTP similarity. This can either be viewed as members of single member clusters with certain atypical parts of their sequences that could be a result of chimeric origin of the receptors or due to evolutionary pressure not shared by their closest phylogenetic neighbors [62]. We have therefore placed these receptors separately as unassociated GPCRs, although these receptors clearly do not belong to the same group (see Additional data file 1). Most of the unassociated receptors remain as orphan receptors.
Conclusion
The phylogenetic analyses performed using human and Drosophila GPCRs suggest that the sequences can be divided into 32 clusters and reveals unexpected level of similarity between human and Drosophila GPCRs. 21 clusters group Drosophila and human GPCRs together suggesting high evolutionary conservation across species for GPCR sequences. There are 10 clusters, four of nucleotide-lipid receptors three clusters of peptide receptors and two clusters of chemokine and one cluster of glutamate receptors that do not contain any representation from Drosophila GPCRs in our current dataset of sequences considered. Perhaps the immune-related receptors, such as the chemokine ones, are not either recognized yet or not present in lower organisms such as Drosophila. If there is a clear absence of such classes of receptors, this might also suggest that immune defense is regulated by proteins other than GPCRs in Drosophila. Interestingly, there is one cluster of secretin Drosophila receptors where there is no human representation. These proteins are involved in aging in Drosophila. Furthermore, in this analysis, we also notice that out of the 21 clusters that co-cluster human and Drosophila GPCRs, Drosophila GPCRs remain isolated sub-clusters in 12 of them leaving behind only nine clusters that allow easy inter-mixing of the two sets of sequences. This includes 3 clusters each of peptide and biogenic amine receptors and one cluster each of class B, C and frizzled receptors.
The current clustering analysis provides ligand class association to 52 Drosophila (Table 2) and 95 human orphan receptors could be associated with probable ligand classes using co-clustering principles as earlier observed within human GPCR sequences alone [12]. Further, similar cellular localizations have been suggested for Drosophila orphan receptors that belong to the opsin family (cluster 14). GPCRs with similar extracellular domain architecture also co-cluster suggesting this similarity is encoded even within the GPCR domain. Further this analysis also suggests dimerizing partner (Q9VKA4_Dro or Q9VR40_Dro) for D-GABABR3 that might form a functional heterodimer. We have determined the relationship of the receptors within subgroups of the large GPCR superfamily by means of a cross-genome phylogenetic clustering approach. These studies also revealed a higher-level phylogenetic organization in which clusters with common ligand structure or chemistry, or a shared function, are evident across genomes. We hope that this approach proves valuable for identifying the natural ligands of Drosophila and human orphan receptors.
Table 2 List of Drosophila orphan receptors
Name Swissprot Code Best match receptor with known ligand; % Identity Cluster Description
Peptide receptors
Q8I943_Dro Q8I943 SSR2_HUMAN; 40.2 2 Somatostatin receptor
Q8ISJ9_Dro Q8ISJ9 SSR5_HUMAN; 45.8 2 Orphan GPCR
Q9V858_Dro Q9V858 GRPR_HUMAN; 40.0 4 CG30106 protein
Q9V9K3_Dro Q9V9K3 BRS3_HUMAN; 36.4 4 CG14593 protein
Q8SWR3_Dro Q8SWR3 5 RE15519p; CG16752 protein
Q9V5T1_Dro Q9V5T1 TRFR_HUMAN; 22.5 5 CG13229 protein; AT19640p
Q9VDC4_Dro Q9VDC4 GHSR_HUMAN; 34.2 5 CG5911 protein
Q9VT27_Dro Q9VT27 TRFR_HUMAN; 38.3 5 CG16726 protein
Q9W025_Dro Q9W025 TRFR_HUMAN; 27.4 5 CG8985 protein
Q9W027_Dro Q9W027 TRFR_HUMAN; 29.3 5 CG13803 protein
GRHRII_Dro GRHRII_Dro GRR2_HUMAN; 34.0 6 Putative corazonin receptor
Q8SX01_Dro Q8SX01 LSHR_HUMAN; 50.0 7 RH44949p
Q9NDI1_Dro Q9NDI1 LSHR_HUMAN; 48.9 7 Glycoprotein hormone receptor II
Q9VBP0_Dro Q9VBP0 LGR8_HUMAN; 35.4 7 CG31096-PA
Q9VYG0_Dro Q9VYG0 LGR8_HUMAN; 40.6 7 CG4187 protein
Q8MKU0_Dro Q8MKU0 NFF2_HUMAN; 31.2 8 CG30340-PA
Q9VWQ9_Dro Q9VWQ9 CCKR_HUMAN; 42.0 8 CG32540 protein
Q9VWR3_Dro Q9VWR3 CCKR_HUMAN; 29.6 8 CG6857 protein
Q8SZ35_Dro Q8SZ35 NY2R_HUMAN; 37.0 11 RE18294p
Q9VRM0_Dro Q9VRM0 NYR_DROME; 37.9 11 CG10626 protein
Q9VW75_Dro Q9VW75 NY1R_HUMAN; 36.6 11 CG7395 protein; GH23382p
Q9W189_Dro Q9W189 NYR_DROME; 29.0 11 CG13575 protein
Nucleotide and lipid receptors
Q9VTU7_Dro Q9VTU7 OPS3_DROME; 38.6 14 CG5638 protein; GH14208p
Q9VVJ1_Dro Q9VVJ1 O00325; 26.9 18 CG7497 protein; GH27361p
Biogenic amine receptors
Q9VCZ3_Dro Q9VCZ3 5H4_HUMAN; 44.3 20 CG6919 protein
Q9VG54_Dro Q9VG54 5H4_HUMAN; 39.9 20 CG6989 protein
O77269_Dro O77269 ML1A_HUMAN; 29.2 21 EG:22E5.10 protein
O77270_Dro O77270 ML1A_HUMAN; 28.1 21 EG:22E5.11 protein
Q9VEG1_Dro Q9VEG1 5H1A_HUMAN; 39.6 22 CG7431 protein
Q9VEG2_Dro Q9VEG2 5HT1_DROME; 19.4 22 CG16766 protein
Q9VAA2_Dro Q9VAA2 AA2A_HUMAN; 39.2 23 CG9753 protein
Q9VHW1_Dro Q9VHW1 ACM3_HUMAN; 36.1 23 CG7918 protein
Q9VMI4_Dro Q9VMI4 5HT1_DROME; 22.3 23 CG13995 protein; RE05601p
Q9VBG4_Dro Q9VBG4 HH2R_HUMAN; 32.6 24 CG12290 protein; GH12381P
Q9VE32_Dro Q9VE32 A2AA_HUMAN; 39.5 24 CG18208 protein
Q9W3V5_Dro Q9W3V5 Q13675; 30.3 24 CG12796 protein
Class B (secretin) receptors
Q9NEF7_Dro Q9NEF7 CRF2_HUMAN; 37.0 25 EG:BACR25B3.3 protein
Q9V6C7_Dro Q9V6C7 CRF2_HUMAN; 42.8 25 CG12370 protein
Q9V6N4_Dro Q9V6N4 CGRR_HUMAN; 41.1 25 CG17043 protein
Q9V716_Dro Q9V716 CRF2_HUMAN; 42.8 25 CG8422 protein
Q8INM0_Dro Q8INM0 MTH_DROME; 38.2 26 CG31147-PA
Q8IPD0_Dro Q8IPD0 MTHA_DROME; 29.8 26 CG31720-PB
Cell adhesion receptors
Q8SZ78_Dro Q8SZ78 CD97_HUMAN; 27.4 27 RE14222p
Q8T4B2_Dro Q8T4B2 CD97_HUMAN; 26.6 27 AT07595p
Q9V4V8_Dro Q9V4V8 CD97_HUMAN; 22.6 27 CG8639 protein
STAN_Dro STAN_DROME CD97_HUMAN; 50.8 27 Protocadherin-like wing polarity protein stan precursor; Starry night protein; Flamingo protein
Class C (glutamate) receptors
Q9V4U4_Dro Q9V4U4 Q8NFS4(MGR7_HUMAN); 41.1 28 CG30361 protein
BOSS_Dro BOSS_DROME O95357(RAIG1); 22.3 30 Bride of sevenless protein precursor
Q9VKA4_Dro Q9VKA4 Q9Y133; 27.2 31 CG31760 protein
Q9VNZ5_Dro Q9VNZ5 MGR_DROME; 32.3 31 CG32447 protein
Q9VR40_Dro Q9VR40 GBR2_HUMAN; 31.4 31 CG31660 protein
Methods
Sequence data mining
Human (537) and Drosophila (284) GPCR amino acid sequences were downloaded from GPCRDB (7.0) [18]. The subset of entries containing the keyword 'olfactory receptors (OR)' or 'gustatory receptors (GR)' or 'taste receptors' were extracted by text parsing and were removed as they were extremely diverse sequences and inclusion of them affects badly on alignments quality. Further, we wanted to avoid polymorphism, splice variants, pseudogenes and duplicates of these receptors and sequences above 90% sequence identity were removed from the data set using CD-HIT [63]. This set amounted to 371 human and 113 Drosophila sequences (Additional data file 1). GPCRs without published ligands in the NCBI-PubMed were considered as orphan receptors. The sequences were renamed to add suffix _Hum and _Dro to refer to human and Drosophila sequences respectively.
Transmembrane helix predictions
Transmembrane domains were identified using HMMTOP program [64]. Amino termini upstream of TMH-1 and carboxyl termini downstream of TMH-7 were removed as they show extreme variability in these regions. Sequence comprising of TMH-1 to TMH-7 alone were considered for the analysis (Figure 2).
Multiple sequence alignments
ClustalX 1.83 [65] was used for multiple sequence alignments (MSA) of receptors with a gap penalty of 10, a gap extension penalty of 0.05 and delay divergent sequences of 35% and protein weight matrix was BLOSUM series. The slow-accurate method was used for the initial pairwise alignments. The protein weight matrix was Blossom 30. When necessary, alignments were optimized by manual editing (Figure 2).
Phylogenetic analysis
An overall phylogenetic tree was inferred from the multiple sequence alignment using PHYLIP package (V 3.5) [66]
Sequence bootstrapping
The bootstrapping of multiple sequence alignment was performed 100 times using SEQBOOT to obtain 100 different alignments. Owing to the limitations in the CONSENSE program of Phylip package to handle large datasets, we restricted to 100 bootstrap replication steps [16].
Neighbor-joining tree
Protein distances were calculated using PROTDIST from the PHYLIP package. The trees were calculated using Neighbor-Joining (NJ) method [67,68] on 100 different distance matrices using NEIGHBOR from the PHYLIP 3.5 package, resulting in 100 trees. These were analyzed using CONSENSE from the PHYLIP package to derive a bootstrapped consensus tree. An unrooted tree was plotted using TREEVIEW [69]. Sequences with more than 50% bootstrap support values were confirmed and grouped.
Maximum likelihood trees
MSAs for each of the groups were obtained as described above and were used for building maximum likelihood trees [70] using TREE-PUZZLE 5.1 [71]. It is least affected by sampling errors and robust to many violations of the assumptions in the evolutionary model [72]. Parameters were estimated by Quartet sampling and NJ tree; The jones-taylor-thornton (JTT) substitution model was used for the calculation with amino acid usage estimated from data, site-to-site rate variation modeled on a gamma distribution with eight rate categories plus invariant sites, and the gamma distribution parameters estimated from the data. 10,000 quartet puzzling steps were performed to obtain support values for each internal branch and trees inferred with the highest likelihood. This method outperforms other methods like neighbor joining or parsimony methods except that it is computationally intensive, extremely slow and cannot be applied to very large datasets. Drosophila 5HTA receptor (5HTA_Dro) of family A was used as out-group for secretin, glutamate, cell adhesion and frizzled receptors. Human (O75205_Hum or GPRC5B) receptor of family B was used as out-group for peptide, chemokine, nucleotide and lipid and biogenic amine receptors for tree constructions (out-groups not shown in the figures) using Tree View [69].
BLAST searches
For sequences with lower support values, similarity measures obtained by searching all against all sequences using BLASTP [73] were used to associate them to the clusters identified by PHYLIP and maximum likelihood methods. Manual inspection of the alignments, bit-score, E-Value, and length of pairwise alignments were considered as measures of similarity. Such receptors may be distantly related to members of the groups but may be sharing high structural similarity and common functional role, possibly due to convergent evolution [74]. It is also possible that these sequences are very diverse that the clustering methods were not sensitive enough to measure these changes [17].
Authors' contributions
M.R.P. Rao has carried out the work and has written the first draft of the manuscript. R.S. has initiated the idea and was involved in discussions and drafting of the final manuscript.
Supplementary Material
Additional data file 2
Key residues conserved among the members of cluster 17.
Click here for file
Additional data file 1
Table indicating the cluster, accession numbers, swissprot codes, gene names and description of the GPCR sequences that have been used.
Click here for file
Acknowledgements
R.S. is a recipient of Senior Research Fellowship awarded by the Wellcome Trust, UK. M.R.P. Rao is a recipient of Senior Research fellowship awarded by Council of Scientific and Industrial Research (CSIR), INDIA. We also thank NCBS-TIFR for infrastructural support.
==== Refs
Christopoulos A Allosteric binding sites on cell-surface receptors: novel targets for drug discovery Nat Rev Drug Discou 2002 1 198 210 10.1038/nrd746
Palczewski K Kumasaka T Hori T Behnke CA Motoshima H Fox BA Le Trong I Teller DC Okada T Stenkamp RE Yamamoto M Miyano M Crystal structure of rhodopsin: A G protein-coupled receptor Science 2000 289 739 745 10926528 10.1126/science.289.5480.739
Libert F Schiffmann SN Lefort A Parmentier M Gerard C Dumont JE Vanderhaeghen JJ Vassart G The orphan receptor cDNA RDC7 encodes an A1 adenosine receptor Embo J 1991 10 1677 1682 1646713
Pyne S Pyne NJ Sphingosine 1-phosphate signalling in mammalian cells Biochem J 2000 349 385 402 10880336 10.1042/0264-6021:3490385
Marchese A George SR Kolakowski LFJ Lynch KR O'Dowd BF Novel GPCRs and their endogenous ligands: expanding the boundaries of physiology and pharmacology Trends Pharmacol Sci 1999 20 370 375 10462760 10.1016/S0165-6147(99)01366-8
Civelli O Nothacker HP Reinscheid R Reverse physiology: discovery of the novel neuropeptide, orphanin FQ/nociceptin Crit Rev Neurobiol 1998 12 163 176 9847053
Gaulton A Attwood TK Bioinformatics approaches for the classification of G-protein-coupled receptors Curr Opin Pharmacol 2003 3 114 120 12681231 10.1016/S1471-4892(03)00005-5
Narumiya S Sugimoto Y Ushikubi F Prostanoid receptors: structures, properties, and functions Physiol Rev 1999 79 1193 1226 10508233
An S Bleu T Hallmark OG Goetzl EJ Characterization of a novel subtype of human G protein-coupled receptor for lysophosphatidic acid J Biol Chem 1998 273 7906 7910 9525886 10.1074/jbc.273.14.7906
Im DS Heise CE Ancellin N O'Dowd BF Shei GJ Heavens RP Rigby MR Hla T Mandala S McAllister G George SR Lynch KR Characterization of a novel sphingosine 1-phosphate receptor, Edg-8 J Biol Chem 2000 275 14281 14286 10799507 10.1074/jbc.275.19.14281
Szekeres PG Muir AI Spinage LD Miller JE Butler SI Smith A Rennie GI Murdock PR Fitzgerald LR Wu H McMillan LJ Guerrera S Vawter L Elshourbagy NA Mooney JL Bergsma DJ Wilson S Chambers JK Neuromedin U is a potent agonist at the orphan G protein-coupled receptor FM3 J Biol Chem 2000 275 20247 20250 10811630 10.1074/jbc.C000244200
Joost P Methner A Phylogenetic analysis of 277 human G-protein-coupled receptors as a tool for the prediction of orphan receptor ligands Genome Biol 2002 3 RESEARCH0063 12429062 10.1186/gb-2002-3-11-research0063
Ignatov A Lintzel J Hermans-Borgmeyer I Kreienkamp HJ Joost P Thomsen S Methner A Schaller HC Role of the G-protein-coupled receptor GPR12 as high-affinity receptor for sphingosylphosphorylcholine and its expression and function in brain development J Neurosci 2003 23 907 914 12574419
Metpally RPR Sowdhamini R Genome wide survey of G protein-coupled receptors in Tetraodon nigroviridis BMC Evol Biol 2005 5 41 16022726 10.1186/1471-2148-5-41
Attwood TK Findlay JB Fingerprinting G-protein-coupled receptors Protein Eng 1994 7 195 203 8170923
Fredriksson R Lagerstrom MC Lundin LG Schioth HB The G-protein-coupled receptors in the human genome form five main families. Phylogenetic analysis, paralogon groups, and fingerprints Mol Pharmacol 2003 63 1256 1272 12761335 10.1124/mol.63.6.1256
Josefsson LG Evidence for kinship between diverse G-protein coupled receptors Gene 1999 239 333 340 10548735 10.1016/S0378-1119(99)00392-3
Horn F Weare J Beukers MW Horsch S Bairoch A Chen W Edvardsen O Campagne F Vriend G GPCRDB: an information system for G protein-coupled receptors Nucleic Acids Res 1998 26 275 279 9399852 10.1093/nar/26.1.275
Banfi S Borsani G Rossi E Bernard L Guffanti A Rubboli F Marchitiello A Giglio S Coluccia E Zollo M Zuffardi O Ballabio A Identification and mapping of human cDNAs homologous to Drosophila mutant genes through EST database searching Nat Genet 1996 13 167 174 8640222 10.1038/ng0696-167
Fortini ME Skupski MP Boguski MS Hariharan IK A survey of human disease gene counterparts in the Drosophila genome J Cell Biol 2000 150 F23 30 10908582 10.1083/jcb.150.2.F23
Rubin GM Yandell MD Wortman JR Gabor Miklos GL Nelson CR Hariharan IK Fortini ME Li PW Apweiler R Fleischmann W Cherry JM Henikoff S Skupski MP Misra S Ashburner M Birney E Boguski MS Brody T Brokstein P Celniker SE Chervitz SA Coates D Cravchik A Gabrielian A Galle RF Gelbart WM George RA Goldstein LS Gong F Guan P Harris NL Hay BA Hoskins RA Li J Li Z Hynes RO Jones SJ Kuehl PM Lemaitre B Littleton JT Morrison DK Mungall C O'Farrell PH Pickeral OK Shue C Vosshall LB Zhang J Zhao Q Zheng XH Lewis S Comparative genomics of the eukaryotes Science 2000 287 2204 2215 10731134 10.1126/science.287.5461.2204
Adams MD Celniker SE Holt RA Evans CA Gocayne JD Amanatides PG Scherer SE Li PW Hoskins RA Galle RF George RA Lewis SE Richards S Ashburner M Henderson SN Sutton GG Wortman JR Yandell MD Zhang Q Chen LX Brandon RC Rogers YH Blazej RG Champe M Pfeiffer BD Wan KH Doyle C Baxter EG Helt G Nelson CR Gabor GL Abril JF Agbayani A An HJ Andrews-Pfannkoch C Baldwin D Ballew RM Basu A Baxendale J Bayraktaroglu L Beasley EM Beeson KY Benos PV Berman BP Bhandari D Bolshakov S Borkova D Botchan MR Bouck J Brokstein P Brottier P Burtis KC Busam DA Butler H Cadieu E Center A Chandra I Cherry JM Cawley S Dahlke C Davenport LB Davies P de Pablos B Delcher A Deng Z Mays AD Dew I Dietz SM Dodson K Doup LE Downes M Dugan-Rocha S Dunkov BC Dunn P Durbin KJ Evangelista CC Ferraz C Ferriera S Fleischmann W Fosler C Gabrielian AE Garg NS Gelbart WM Glasser K Glodek A Gong F Gorrell JH Gu Z Guan P Harris M Harris NL Harvey D Heiman TJ Hernandez JR Houck J Hostin D Houston KA Howland TJ Wei MH Ibegwam C Jalali M Kalush F Karpen GH Ke Z Kennison JA Ketchum KA Kimmel BE Kodira CD Kraft C Kravitz S Kulp D Lai Z Lasko P Lei Y Levitsky AA Li J Li Z Liang Y Lin X Liu X Mattei B McIntosh TC McLeod MP McPherson D Merkulov G Milshina NV Mobarry C Morris J Moshrefi A Mount SM Moy M Murphy B Murphy L Muzny DM Nelson DL Nelson DR Nelson KA Nixon K Nusskern DR Pacleb JM Palazzolo M Pittman GS Pan S Pollard J Puri V Reese MG Reinert K Remington K Saunders RD Scheeler F Shen H Shue BC Siden-Kiamos I Simpson M Skupski MP Smith T Spier E Spradling AC Stapleton M Strong R Sun E Svirskas R Tector C Turner R Venter E Wang AH Wang X Wang ZY Wassarman DA Weinstock GM Weissenbach J Williams SM WoodageT Worley KC Wu D Yang S Yao QA Ye J Yeh RF Zaveri JS Zhan M Zhang G Zhao Q Zheng L Zheng XH Zhong FN Zhong W Zhou X Zhu S Zhu X Smith HO Gibbs RA Myers EW Rubin GM Venter JC The genome sequence of Drosophila melanogaster Science 2000 287 2185 2195 10731132 10.1126/science.287.5461.2185
Burdett H van den Heuvel M Fruits and flies: a genomics perspective of an invertebrate model organism Brief Funct Genomic Proteomic 2004 3 257 266 15642188
Davenport AP Peptide and trace amine orphan receptors: prospects for new therapeutic targets Curr Opin Pharmacol 2003 3 127 134 12681233 10.1016/S1471-4892(03)00003-1
Reubi JC Peptide receptors as molecular targets for cancer diagnosis and therapy Endocr Rev 2003 24 389 427 12920149 10.1210/er.2002-0007
Birgul N Weise C Kreienkamp HJ Richter D Reverse physiology in Drosophila: identification of a novel allatostatin-like neuropeptide and its cognate receptor structurally related to the mammalian somatostatin/galanin/opioid receptor family Embo J 1999 18 5892 5900 10545101 10.1093/emboj/18.21.5892
Matsumoto M Kamohara M Sugimoto T Hidaka K Takasaki J Saito T Okada M Yamaguchi T Furuichi K The novel G-protein coupled receptor SALPR shares sequence similarity with somatostatin and angiotensin receptors Gene 2000 248 183 189 10806363 10.1016/S0378-1119(00)00123-2
Tanaka H Yoshida T Miyamoto N Motoike T Kurosu H Shibata K Yamanaka A Williams SC Richardson JA Tsujino N Garry MG Lerner MR King DS O'Dowd BF Sakurai T Yanagisawa M Characterization of a family of endogenous neuropeptide ligands for the G protein-coupled receptors GPR7 and GPR8 Proc Natl Acad Sci U S A 2003 100 6251 6256 12719537 10.1073/pnas.0837789100
Akeson M Sainz E Mantey SA Jensen RT Battey JF Identification of four amino acids in the gastrin-releasing peptide receptor that are required for high affinity agonist binding J Biol Chem 1997 272 17405 17409 9211882 10.1074/jbc.272.28.17405
Sainz E Akeson M Mantey SA Jensen RT Battey JF Four amino acid residues are critical for high affinity binding of neuromedin B to the neuromedin B receptor J Biol Chem 1998 273 15927 15932 9632639 10.1074/jbc.273.26.15927
Lin Y Jian X Lin Z Kroog GS Mantey S Jensen RT Battey J Northup J Two amino acids in the sixth transmembrane segment of the mouse gastrin-releasing peptide receptor are important for receptor activation J Pharmacol Exp Ther 2000 294 1053 1062 10945859
Egerod K Reynisson E Hauser F Cazzamali G Williamson M Grimmelikhuijzen CJ Molecular cloning and functional expression of the first two specific insect myosuppressin receptors Proc Natl Acad Sci U S A 2003 100 9808 9813 12907701 10.1073/pnas.1632197100
Johnson EC Garczynski SF Park D Crim JW Nassel DR Taghert PH Identification and characterization of a G protein-coupled receptor for the neuropeptide proctolin in Drosophilamelanogaster Proc Natl Acad Sci U S A 2003 100 6198 6203 12730362 10.1073/pnas.1030108100
Hsu SY Kudo M Chen T Nakabayashi K Bhalla A van der Spek PJ van Duin M Hsueh AJ The three subfamilies of leucine-rich repeat-containing G protein-coupled receptors (LGR): identification of LGR6 and LGR7 and the signaling mechanism for LGR7 Mol Endocrinol 2000 14 1257 1271 10935549 10.1210/me.14.8.1257
Jiang Y Luo L Gustafson EL Yadav D Laverty M Murgolo N Vassileva G Zeng M Laz TM Behan J Qiu P Wang L Wang S Bayne M Greene J Monsma FJ Zhang FL Identification and characterization of a novel RF-amide peptide ligand for orphan G-protein-coupled receptor SP9155 J Biol Chem 2003 278 27652 27657 12714592 10.1074/jbc.M302945200
Sakurai T Amemiya A Ishii M Matsuzaki I Chemelli RM Tanaka H Williams SC Richardson JA Kozlowski GP Wilson S Arch JR Buckingham RE Haynes AC Carr SA Annan RS McNulty DE Liu WS Terrett JA Elshourbagy NA Bergsma DJ Yanagisawa M Orexins and orexin receptors: a family of hypothalamic neuropeptides and G protein-coupled receptors that regulate feeding behavior Cell 1998 92 573 585 9491897 10.1016/S0092-8674(00)80949-6
Bae YS Park EY Kim Y He R Ye RD Kwak JY Suh PG Ryu SH Novel chemoattractant peptides for human leukocytes Biochem Pharmacol 2003 66 1841 1851 14563494 10.1016/S0006-2952(03)00552-5
Santos RA Simoes e Silva AC Maric C Silva DM Machado RP de Buhr I Heringer-Walther S Pinheiro SV Lopes MT Bader M Mendes EP Lemos VS Campagnole-Santos MJ Schultheiss HP Speth R Walther T Angiotensin-(1-7) is an endogenous ligand for the G protein-coupled receptor Mas Proc Natl Acad Sci U S A 2003 100 8258 8263 12829792 10.1073/pnas.1432869100
Dong X Han S Zylka MJ Simon MI Anderson DJ A diverse family of GPCRs expressed in specific subsets of nociceptive sensory neurons Cell 2001 106 619 632 11551509 10.1016/S0092-8674(01)00483-4
Lin DC Bullock CM Ehlert FJ Chen JL Tian H Zhou QY Identification and molecular characterization of two closely related G protein-coupled receptors activated by prokineticins/endocrine gland vascular endothelial growth factor J Biol Chem 2002 277 19276 19280 11886876 10.1074/jbc.M202139200
Langmead CJ Szekeres PG Chambers JK Ratcliffe SJ Jones DN Hirst WD Price GW Herdon HJ Characterization of the binding of [(125)I]-human prolactin releasing peptide (PrRP) to GPR10, a novel G protein coupled receptor Br J Pharmacol 2000 131 683 688 11030716 10.1038/sj.bjp.0703617
Zlotnik A Yoshie O Chemokines: a new classification system and their role in immunity Immunity 2000 12 121 127 10714678 10.1016/S1074-7613(00)80165-X
Park Y Kim YJ Adams ME Identification of G protein-coupled receptors for Drosophila PRXamide peptides, CCAP, corazonin, and AKH supports a theory of ligand-receptor coevolution Proc Natl Acad Sci U S A 2002 99 11423 11428 12177421 10.1073/pnas.162276199
O'Tousa JE Baehr W Martin RL Hirsh J Pak WL Applebury ML The Drosophila ninaE gene encodes an opsin Cell 1985 40 839 850 2985266 10.1016/0092-8674(85)90343-5
Tunaru S Kero J Schaub A Wufka C Blaukat A Pfeffer K Offermanns S PUMA-G and HM74 are receptors for nicotinic acid and mediate its anti-lipolytic effect Nat Med 2003 9 352 355 12563315 10.1038/nm824
Montero C Campillo NE Goya P Paez JA Homology models of the cannabinoid CB1 and CB2 receptors. A docking analysis study Eur J Med Chem 2005 40 75 83 15642412 10.1016/j.ejmech.2004.10.002
Xu Y Zhu K Hong G Wu W Baudhuin LM Xiao Y Damron DS Sphingosylphosphorylcholine is a ligand for ovarian cancer G-protein-coupled receptor 1 Nat Cell Biol 2000 2 261 267 10806476 10.1038/35010529
Ludwig MG Vanek M Guerini D Gasser JA Jones CE Junker U Hofstetter H Wolf RM Seuwen K Proton-sensing G-protein-coupled receptors Nature 2003 425 93 98 12955148 10.1038/nature01905
Li Q Schachter JB Harden TK Nicholas RA The 6H1 orphan receptor, claimed to be the p2y5 receptor, does not mediate nucleotide-promoted second messenger responses Biochem Biophys Res Commun 1997 236 455 460 9240460 10.1006/bbrc.1997.6984
Branchek TA Blackburn TP Trace amine receptors as targets for novel therapeutics: legend, myth and fact Curr Opin Pharmacol 2003 3 90 97 12550748 10.1016/S1471-4892(02)00028-0
Barrett P Conway S Morgan PJ Digging deep--structure-function relationships in the melatonin receptor family J Pineal Res 2003 35 221 230 14521626 10.1034/j.1600-079X.2003.00090.x
Matsumoto M Saito T Takasaki J Kamohara M Sugimoto T Kobayashi M Tadokoro M Matsumoto S Ohishi T Furuichi K An evolutionarily conserved G-protein coupled receptor family, SREB, expressed in the central nervous system Biochem Biophys Res Commun 2000 272 576 582 10833454 10.1006/bbrc.2000.2829
Feng G Hannan F Reale V Hon YY Kousky CT Evans PD Hall LM Cloning and functional characterization of a novel dopamine receptor from Drosophila melanogaster J Neurosci 1996 16 3925 3933 8656286
Han KA Millar NS Davis RL A novel octopamine receptor with preferential expression in Drosophila mushroom bodies J Neurosci 1998 18 3650 3658 9570796
Mizushima K Miyamoto Y Tsukahara F Hirai M Sakaki Y Ito T A novel G-protein-coupled receptor gene expressed in striatum Genomics 2000 69 314 321 11056049 10.1006/geno.2000.6340
Lin YJ Seroude L Benzer S Extended life-span and stress resistance in the Drosophila mutant methuselah Science 1998 282 943 946 9794765 10.1126/science.282.5390.943
Foord SM Jupe S Holbrook J Bioinformatics and type II G-protein-coupled receptors Biochem Soc Trans 2002 30 473 479 12196118 10.1042/BST0300473
Parmentier ML Galvez T Acher F Peyre B Pellicciari R Grau Y Bockaert J Pin JP Conservation of the ligand recognition site of metabotropic glutamate receptors during evolution Neuropharmacology 2000 39 1119 1131 10760355 10.1016/S0028-3908(99)00204-X
Pin JP Galvez T Prezeau L Evolution, structure, and activation mechanism of family 3/C G-protein-coupled receptors Pharmacol Ther 2003 98 325 354 12782243 10.1016/S0163-7258(03)00038-X
Mezler M Muller T Raming K Cloning and functional expression of GABA(B) receptors from Drosophila Eur J Neurosci 2001 13 477 486 11168554 10.1046/j.1460-9568.2001.01410.x
Wang HY Malbon CC Wnt signaling, Ca2+, and cyclic GMP: visualizing Frizzled functions Science 2003 300 1529 1530 12791979 10.1126/science.1085259
Fredriksson R Gloriam DE Hoglund PJ Lagerstrom MC Schioth HB There exist at least 30 human G-protein-coupled receptors with long Ser/Thr-rich N-termini Biochem Biophys Res Commun 2003 301 725 734 12565841 10.1016/S0006-291X(03)00026-3
Li W Jaroszewski L Godzik A Clustering of highly homologous sequences to reduce the size of large protein databases Bioinformatics 2001 17 282 283 11294794 10.1093/bioinformatics/17.3.282
Tusnady GE Simon I The HMMTOP transmembrane topology prediction server Bioinformatics 2001 17 849 850 11590105 10.1093/bioinformatics/17.9.849
Thompson JD Gibson TJ Plewniak F Jeanmougin F Higgins DG The CLUSTAL_X windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools Nucleic Acids Res 1997 25 4876 4882 9396791 10.1093/nar/25.24.4876
Felsenstein J PHYLIP, phylogenetic inference package, Department of Genetics, University of Washington, Seattle, WA. 2003
Saitou N Nei M The neighbor-joining method: a new method for reconstructing phylogenetic trees Mol Biol Evol 1987 4 406 425 3447015
Kumar S Gadagkar SR Efficiency of the neighbor-joining method in reconstructing deep and shallow evolutionary relationships in large phylogenies J Mol Evol 2000 51 544 553 11116328
Page RD TreeView: an application to display phylogenetic trees on personal computers Comput Appl Biosci 1996 12 357 358 8902363
Strimmer K Haeseler AV Quartet puzzling: a quartet maximum-likelihood method for reconstructing tree topologies. Mol Biol Evol 1996 13 964 969
Schmidt HA Strimmer K Vingron M von Haeseler A TREE-PUZZLE: maximum likelihood phylogenetic analysis using quartets and parallel computing Bioinformatics 2002 18 502 504 11934758 10.1093/bioinformatics/18.3.502
Felsenstein J Phylogenies from molecular sequences: inference and reliability Annu Rev Genet 1988 22 521 565 3071258 10.1146/annurev.ge.22.120188.002513
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Donnelly D Findlay JB Blundell TL The evolution and structure of aminergic G protein-coupled receptors Receptors Channels 1994 2 61 78 8081733
|
16091152
|
PMC1192796
|
CC BY
|
2021-01-04 16:32:48
|
no
|
BMC Genomics. 2005 Aug 10; 6:106
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-106
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-1071609554410.1186/1471-2164-6-107Research ArticleTranscribed Tc1-like transposons in salmonid fish Krasnov Aleksei [email protected] Heikki [email protected] Sergey [email protected]ölsä Hannu [email protected] Institute of Applied Biotechnology, University of Kuopio, P.O.B. 1627, FIN-70211 Kuopio, Finland2 Sechenov Institute of Evolutionary Physiology and Biochemistry, M.Toreza av. 44, Petersburg, 194223, Russia2005 12 8 2005 6 107 107 31 1 2005 12 8 2005 Copyright © 2005 Krasnov et al; licensee BioMed Central Ltd.2005Krasnov et al; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Mobile genetic elements comprise a substantial fraction of vertebrate genomes. These genes are considered to be deleterious, and in vertebrates they are usually inactive. High throughput sequencing of salmonid fish cDNA libraries has revealed a large number of transposons, which remain transcribed despite inactivation of translation. This article reports on the structure and potential role of these genes.
Results
A search of EST showed the ratio of transcribed transposons in salmonid fish (i.e., 0.5% of all unique cDNA sequences) to be 2.4–32 times greater than in other vertebrate species, and 68% of these genes belonged to the Tc1-family of DNA transposons. A phylogenetic analysis of reading frames indicate repeated transposition of distantly related genes into the fish genome over protracted intervals of evolutionary time. Several copies of two new DNA transposons were cloned. These copies showed relatively little divergence (11.4% and 1.9%). The latter gene was transcribed at a high level in rainbow trout tissues, and was present in genomes of many phylogenetically remote fish species. A comparison of synonymous and non-synonymous divergence revealed remnants of divergent evolution in the younger gene, while the older gene evolved in a neutral mode. From a 1.2 MB fragment of genomic DNA, the salmonid genome contains approximately 105 Tc1-like sequences, the major fraction of which is not transcribed. Our microarray studies showed that transcription of rainbow trout transposons is activated by external stimuli, such as toxicity, stress and bacterial antigens. The expression profiles of Tc1-like transposons gave a strong correlation (r2 = 0.63–0.88) with a group of genes implicated in defense response, signal transduction and regulation of transcription.
Conclusion
Salmonid genomes contain a large quantity of transcribed mobile genetic elements. Divergent or neutral evolution within genomes and lateral transmission can account for the diversity and sustained persistence of Tc1-like transposons in lower vertebrates. A small part of transposons remain transcribed and their transcription is enhanced by responses to acute conditions.
==== Body
Background
A large fraction of repetitive sequences originate in eukaryotic genomes from mobile genetic elements (MGEs), which are grouped into 2 classes. Class I transposons require mRNA intermediates, whereas class II elements transpose directly as DNA. Tc1-like class II transposons, named after the founder gene in Caenorhabditis elegans, are probably the most widespread MGEs in nature, and are found in fungi, plant ciliates, nematodes, arthropods, fish, amphibians and mammals (reviewed in [1]). These genes contain a single reading frame that encodes for the enzyme transposase, which is flanked with terminal inverted repeating units. Transposition of class II MGEs is characterized by limited requirements for host cellular factors, which can account for their remarkable ability to undergo horizontal transfer across great taxonomic distances [2]. MGEs are regarded as parasitic genes, and proliferation is deleterious for the host. Therefore, transposition is commonly followed by inactivation. MGEs could play an important role in the evolution of teleost fish, and comprise a substantial fraction of their genome. Multiple copies of Tc1-like transposons were found in several fish species from different orders [3-6], however transcription of teleost Tc1-like genes has not been documented. Recent high-throughput sequencing of salmonid cDNA libraries has revealed surprisingly large number of transposon transcripts. Most if not all these sequences contain incapacitating mutations in the reading frames, and can be regarded as transcribed pseudogenes or null-alleles. At present, rainbow trout (Oncorhynchus mykiss) and Atlantic salmon (Salmo salar) TIGR Gene Indices [7] contain 50773 and 31341 unique cDNA sequences, respectively among which we found several hundreds MGE, Tc1-like genes being most abundant. This wealth of sequence information provides insight into the structure and evolution of transposons. We also cloned several copies of two rainbow trout Tc1-like genes with complete reading frames, which adds to understanding the transposon life cycle. Multiple gene expression analyses with high-density cDNA microarray indicate stimulation of rainbow trout transposons transcription in response to stress, toxicity and pathogens.
Results
In order to search for transcribed transposons in salmonid fish, we compared the unique cDNA sequences from TIGR gene indices with 262 metazoan transposon proteins retrieved from Swissprot. Blastx found matches in 273 rainbow trout and 163 Atlantic salmon sequences at a cutoff value e < 10-20 (Table 1). The ratio of transposons to all cDNA sequences in salmonids was 2.35–31.6 times greater than in other vertebrate species with available gene indices, and a large fraction (68.3%) showed similarity to 11 proteins of the Tc1 family. Tc1-like transcripts were found in the gene indices of 4 other teleost fish species and in the African clawed frog Xenopus laevis, but not in higher vertebrates. To estimate an approximate number of Tc1-like genes, 6 genomic clones of Atlantic salmon were analyzed, covering 1.2 MB [Genbank:AC148723, Genbank:AC149099, Genbank:AC148779, Genbank:AC148618, Genbank:AC148617 and AC148616], and a blastx search found 56 matches at a cutoff value of 10-20. The size of the haploid Atlantic salmon genome is equal to 3 billion base pairs. Assuming a relatively homogenous distribution of Tc1-transposons, about 140,000 copies can be expected, which is 3 orders of magnitude greater than the number of Tc1-like sequences in the salmonid fish gene indices. It is necessary to note that TIGR contigs are produced by automatic assembly of EST sequences that have at least 95% homology in overlaps of minimum units of 40 base pairs [8]. Therefore, transcripts of recently diverged transposon copies could be merged unless they were flanked by differing 5'- and 3'-untranslated sequences. The numbers of transcribed transposons can be greater than the number estimated by searching across gene indices, but it is likely that only a minor fraction of salmonid Tc1-like genes is active.
Table 1 Transcribed vertebrate transposons1.
Species Sequences Transposons Tc1-like
Rainbow trout (Oncorhynchus mykiss) 50773 273 (0.538) 188
Atlantic salmon (Salmo salar) 31341 163 (0.52) 110
Medaka (Oryzias latipes) 26689 59 (0.221) 8
Zebrafish (Danio rerio) 93442 194 (0.208) 84
Killifish (Fundulus heteroclitus) 15538 3 (0.019) 3
Pufferfish (Takifugu rubripes) 11112 11 (0.099) 1
African clawed frog (Xenopus laevis) 77599 164 (0.211) 53
Chicken (Gallus gallus) 116777 20 (0.017) 0
Rat (Rattus norvegicus) 147056 310 (0.211) 0
1Unique cDNA sequences from TIGR Gene Indexes were searched across 262 metazoan transposons from Swissprot and 11 Tc1-like transposases using blastx at a cutoff value of e < 10-20. Percentages for all cDNA sequences are in parentheses.
Most Tc1-like sequences from the rainbow trout gene index contained incomplete reading frames. To analyse the structural relatedness of these genes, we used 38 fragments, which encode at least 170 amino acids at the C-termini. Thus, 31 sequences were from the TIGR database plus 7 more were produced in this study (i.e., newly identified genes named Glan and Barb [Genbank: AY880883-AY880888]). The maximum likelihood (ML) tree consisted of 3 single genes and 11 clades, containing 2 to 5 sequences (Figure 1). Seven clades (I-VII) could be regarded as a part of the multi gene family, as sequence identity with the nearest neighbors was in the range of 35–73%; the remaining 4 clades were highly divergent. Only 1 of 6 clades containing more than 3 genes (X in Figure 1) was split into clusters supported by high bootstrap values. The highest sequence identity were observed for Glan and Barb. However, divergence within other clades could in theory be overestimated, due to forced assembly of similar transcripts.
Figure 1 Structural relatedness of transcribed rainbow trout Tc1-like transposons. The ML tree is based on sequences encoding for at least 170 amino acids at the C-termini. The TIGR sequences are designated by the accession numbers, transposons Barb and Glan were identified in this study. Tree was produced using Dnaml (Phylip package), nodes with bootstrap values greater than 0.75 are indicated. Accession numbers of TIGR contigs in the clusters are: I-1 – BX884691; I-2 – TC46229; II-1 – TC52875; II-2 – TC46539; III-1 – TC46343; III-2 – TC46498; III-3 – TC46491; III-4 – CB488722; IV-1 – TC46455; IV-2 – CA377451; IV-3 – TC47500; IV-4 – TC47499; IV-5 – CA369142; V-1 – TC46391; V-2 – CA369399; VIII-1 – TC54663; VIII-2 – TC54666; IX-1 – CB488927; IX-2 – TC46493; X-1 – TC46521; X-2 – TC46197; X-3 – TC46383; X-4 – TC46308; XI-1 – CA361855; XI-2 – TC54683; XI-3 – CR368829.
Sequencing of complete reading frames for 3 copies of Glan and 4 copies of Barb allowed for the study of transposons molecular evolution within the rainbow trout genome. All 7 sequences include incapacitating mutations, which prevent translation of transposase. Barb copies have diverged up to 11.4 ± 1.4% (mean ± SD) and accumulation of deletions (Figure 2) impeded reconstruction of the ancestral protein. Low divergence of Glan copies (1.9 ± 0.8%) suggest relatively recent transposition into the rainbow trout genome. The consensus sequence of 3 reading frames was identical to TIGR contig [TGI:TC46394], which encoded a protein with characteristic features of Tc1-like transposase, such as the presence of domains required for nuclear localization, DNA binding, cleavage and joining and DDE motif found in the catalytic units of diverse MGEs and retroviruses (Figure 3). Noteworthy of mention is that all transcripts of Glan contained mutations that prevented translation of transposase, however the consensus contig sequence that was assembled from a large number of EST from different cDNA libraries appeared intact. Given that the rate of spontaneous mutations in vertebrate germ cell lines is ~10-5 [9], transposition of Glan could have taken place as recently as only a few thousand years ago. We also performed PCR screen of this gene in fish from inland reservoirs of Finland, where it was detected in 17 species from different orders (Table 2). Interestingly, three of the four species in which Glan was not found (grayling, whitefish and vendace) are more closely related to rainbow trout than most of those species carrying this gene. Low divergence of copies and discontinuous distribution are evidence for horizontal transmission. We analysed the rates of synonymous (Ks) and non-synonymous (Ka) substitutions in newly identified rainbow trout transposons using a sequence of the nearest Swissprot protein (hypothetical transposase of plaice, with 77% homology [Genbank:CAB51372]) as a reference (Table 3). With respect to this transposase, the Ks/Ka ratio was high and significantly greater in the younger gene (4.85 ± 0.30 in Glan and 3.35 ± 0.04 in Barb). A comparison of copies indicated a probability of divergent evolution in Glan (Ks/Ka = 0.69 ± 0.05). In Barb the rates of synonymous and non-synonymous substitutions approached unity (Ks/Ka = 1.03 ± 0.12), which is consistent with the protracted accumulation of mutations in a solely neutral mode.
Table 2 Presence of Glan in genomic DNA of fish from inland waters of Finland.
Species Result
Arctic charr (Salvelinus alpinus) Found
Brown trout (Salmo trutta) Found
Smelt (Osmerus eperlanus) Found
Grayling (Thymallus thymallus) Not found
Vendace (Coregonus albula) Not found
Whitefish (Coregonus lavaretus) Not found
Pikepearch (Sander lucioperca) Found
Perch (Perca fluviatilis) Found
Ruffe (Gymnocephalus cernuus) Found
Crucian carp (Carassius carassius) Found
Roach (Rutilus rutilus) Found
Bullhead (Cottus gobio) Found
Bream (Abramis brama) Found
Silver bream (Abramis bjoerkna) Found
Bleak (Alburnus alburnus) Found
Dace (Leuciscus leuciscus) Found
Rudd (Scardinius erythrophthalmus) Found
Ide (Leuciscus idus) Found
Burbot (Lota lota) Found
Northern pike (Esox lucius) Found
Eel (Anguilla anguilla) Found
Table 3 Synonymous (Ks) and non-synonymous (Ka) divergences of the rainbow trout transposons Glan and Barb. Plaice transposase was used as a reference.
Genes Ks Ka Ks/Ka
Glan
Rainbow trout (3 copies) 0.016 ± 0.007 0.022 ± 0.009 0.69 ± 0.05
Plaice transposase 0.69 ± 0.02 0.14 ± 0.01 4.85 ± 0.31
Barb
Rainbow trout (4 copies) 0.12 ± 0.02 0.11 ± 0.01 0.99 ± 0.13
Plaice transposase 0.68 ± 0.02 0.20 ± 0.00 3.35 ± 0.04
Figure 2 Alignment of protein coding sequences of new transcribed rainbow trout TC1-like transposases cloned in this study.
Figure 3 Alignment of deduced amino acid sequences of rainbow trout transposon Glan with the Tc1-like transposon of plaice, Pleuronectes platessa (Genbank: CAB51372), TPA of frog, Rana pipiens (Genbank: DAA01561) and tcb1 of the nematode, C. elegans (Genbank: NP_741053). Homeodomain (indicated with box) is involved in the binding of DNA; the DDE/D motif (indicated with arrows) is present in diverse MGE [1].
We did not find sequences of any other known proteins in the salmonid Tc1-like contigs and probably transposons are transcribed from own promoters. Evidence for regulation of transposon transcription rate was produced in microarray analyses. We used a platform designed for studies of responses to environmental stress, toxicity and pathogens in salmonid fish [10,11]. Overall this platform included more than 1300 genes, 7 of which were similar to Tc1-like transposons. Five transposons showed marked differential expression in response to external stimuli, such as handling stress, exposure to toxic compounds and injection of cortisol or bacterial antigens; the microarray results were confirmed with real-time qPCR. A consensus profile of transposons correlated with those of 27 protein coding genes in 35 microarray experiments (Pearson r2 > 0.63); examples are presented in Figure 4. The highest correlation (r2 > 0.8) was shown by classical markers of cellular stress, such as the aryl hydrocarbon receptor, MAP kinase 13 and hypoxia inducible factor. We also searched for enrichment of Gene Ontology [12] categories in this list of genes. Significant over-representation was demonstrated by functional classes that are implicated in protective reactions to acute conditions (i.e., response to stress and oxidative stress, defense and humoral immune response, receptors and regulators of transcription, Table 4).
Figure 4 Differential expression of transposons in rainbow trout. The panel presents profiles of transposons and a group of genes that showed coordinated expression in 35 microarray experiments (Pearson r2 is indicated). Selected experiments are reported: 1–8 – exposure of yolk sac rainbow trout fry to model contaminants [10], β-naphthoflavone, low (1) and high (2) doses; cadmium, low (3) and high (4) doses; carbon tetrachloride, low (5) and high (6) doses; pyrene, low (7) and high (8) doses. Items 9–12 – response to handling stress [11, GEO:GSM22355], kidney, 1 day (9) and 5 days (10); brain, 1 day (11) and 5 days (12).
Table 4 Over-presentation of Gene Ontology classes in a list of genes that showed co-ordinated expression with Tc1-like transposons. The composition of microarray was used as a reference. The gene names and expression profiles are shown in Figure 4.
GO classes P Genes in list Genes on chip
Antimicrobial humoral response 1 × 10-5 7 31
Defense response 2 × 10-4 12 102
Signal transduction 0,008 10 113
Cellular defense response 0,009 3 12
Response to oxidative stress 0,03 3 18
Transcription regulator activity 0,03 5 47
Receptor activity 0,04 5 49
1Exact Fisher's probability
Discussion
Having a large number of transposons and a preponderance of Tc1-like genes is a characteristic feature of salmonid genomes [3-5]. Sequence analysis of the transcribed genes (Figure 1) suggested repeated transpositions at protracted intervals. A wide distribution of Tc1 transposons is believed to account for the limited requirements in the host cellular factors. Sleeping Beauty, an artificially reconstructed salmon transposon [13] is capable of integration into genomes of a wide range of vertebrate species, however different efficiencies observed in various cell lines point to possible involvement of the recipient's proteins in transposition [14]. This is in line with a wide, though limited, distribution of homologs for the transcribed salmonid DNA transposons, which have not been found among EST of warm-blood vertebrates. The variety of salmonid Tc1-like genes is truly remarkable. Phylogenetic analyses of 38 sequences, encoding homologous fragments of C-termini, found 14 distinct types of Tc1-like genes and the real number of different genes is probably much greater. Our search was based on the similarity between proteins that were available from Swissprot, and many transposons could remain unidentified due to the lack of known homologs. Furthermore, the rapid decay of transposons could impede the discovery of ancient transposed genes.
Despite the wide spread occurrence of Tc1-like transposons in vertebrates, not a single active gene has been identified to date [14]. Inactivation of salmonid DNA transposons could take place within a relatively short period of time after transmission. Cloning of 2 transposons having a relatively low divergence rate indicates the rapid accumulation of incapacitating mutations, such as insertions or deletions, shifts of reading frames and premature stop codons (Figure 2). Analysis of synonymous and non-synonymous substitutions suggest that inactivation of younger transposon could be preceded by selective divergence within a limited period of time, whilst evolution of the older gene appeared entirely neutral. Results from a study on recent transpositions in insects from four different orders suggest that selective constraints operate exclusively by horizontal gene transfer [15]. A comparison of rainbow trout genes with Tc1-like transposon from plaice confirm the conservation of functionally important domains in distantly related proteins, which is gradually obscured during the course of neutral evolution (Ks/Ka ratios in the younger Glan and older Barb genes are 4.85 ± 0.31 and 3.35 ± 0.04 respectively).
Silencing of transposons takes place at the transcriptional or post-transcriptional levels [16], and both of these mechanisms could act in salmonid fish. Based on frequency in a 1.2 MB gene fragment, we can assume that Tc1-like genes comprise nearly 5% of the Atlantic salmon genome and only a minor fraction preserved transcription after inactivation of translation. A survey of salmonid EST found untranslated transposons in both sense and anti-sense polarities, which is the main prerequisite for the formation of double-stranded RNA. RNA interference (RNAi) is implicated in the control of transposition in germ cell lines of the nematode C. elegans [17], and existence of an RNAi pathway in rainbow trout was recently demonstrated [18]. Suppression of intact transposases with mutant genes was also reported in insects, and this control mechanism is referred to as dominant-negative complementation [19].
Given efficient protection against transposition in animals, the tenacity and variety of transposons may seem surprising. Sustained persistence of transposons can, in theory, account for their residence in unknown reservoir species; e.g, the role of parasites as potential vectors of horizontal transfer across phylogenetically remote organisms has been hypothesized [20]. However this can hardly explain the remarkable diversity of these genes. The ML tree (Figure 1) suggests that at each transposition event, the rainbow trout genome was invaded with a new transposon, although several genes could have a common ancestor. If expression of translated genes is under control of RNAi, successful recurring transposition of identical or highly similar genes appears unlikely. Hence, the combination of neutral or divergent evolution within a genome with transfer across phylogenetic boundaries can be the most efficient strategy for the survival and diversification of transposons. PCR screen detected Glan in genomes of many fish species from phylogenetically remote taxonomic groups (Table 2). Clades I-VII of the ML tree (Figure 1) can correspond to genes that evolved independently. However it is also possible that descendants of a founder gene has returned several times into the rainbow trout genome, after passage through a chain of co-evolutionary hosts.
Results of our microarray studies suggested that a large fraction of transcribed Tc1-genes can be stimulated under acute conditions, but it remains unclear whether or not the transposon transcripts have any functional importance. In theory, they can be transcribed from cryptic promoters, which are activated by the remodelling of chromatin. However, input from stress-responsive promoters is also plausible. Transcripts can be required for the control of transposition through RNAi, however such explanation appears unlikely for highly mutated genes that were probably silenced long ago in evolutionary time. Currently, there is a growing body of evidence to support the involvement of non-coding RNA into the regulation of gene expression at different levels. The role of small and large RNA in modification of the chromatin structure was reviewed recently [21-23]. Stress-induced transcription of short interspersed repeated sequences (SINE) was reported in human, mouse and silkworm [24-27]; and SINE transcripts were shown to enhance translation of reporter genes [28,29]. Stress also activates the transcription of satellite III repeat [30]. Because this large non-coding RNA is consistently associated with chromatin, it can be required for the protection of sensitive regions from stress-induced damage. Synthetic double-stranded RNA enhances the expression of anti-viral proteins in salmonid fish [31,32] and, in theory, endogenous dsRNA can mimic a viral infection by launching protective reactions.
Tc1-like transposons are co-regulated with a group of genes that are implicated in the defense response, signal transduction and regulation of transcription. In this respect, it is noteworthy to mention that Tc1-like fragments reside in a number of immune and stress-related salmonid genes, such as the non-classical MHC class I antigen [Genbank:AF091779, Genbank:AF091780], immunoglobulin heavy chain, IgD [Genbank:AF141605, Genbank:AF278717], inducible nitric oxide synthase iNOS/NOS2 [Genbank:AJ295231] and aryl hydrocarbon receptor 2b, AhR2 [Genbank:AY463929]. Multiple copies of Glan in sense and anti-sense polarity are found in rainbow trout MHC class Ia [Genbank: AB162342.1] and b [AB162343.1] regions, in the vicinity of genes encoding the complement proteasome subunit and several MHCI loci. Modulation of gene expression that was due to the insertion of transposons has been documented in many studies (reviewed in [33]), and involvement of dispersed repeated sequences into the co-ordination of gene expression with similar functions was hypothesized more than three decades ago [34]. The role of transposon transcripts in the regulation of gene expression was recently discovered in yeast [35], where the induction of an RNAi-dependent silent chromatin configuration resulted in reduced transcription of several meiotic genes. A possible involvement of transposon transcripts in the regulation of gene expression in salmonid fish remains to be studied.
Conclusion
Information produced by the sequencing of salmonid fish cDNA libraries and identification of recently transmitted transposons provide new insights into the structure, diversity and molecular evolution and life cycle of mobile genetic elements. High expression levels in rainbow trout tissues and marked responses to external stimuli indicate potential functional roles of transposon pseudogenes, which requires further investigation. These genes can be used as sensitive molecular biomarkers of acute conditions in salmonid fish.
Methods
Sequence analyses
The expressed transposons were analysed in rainbow trout and Atlantic salmon TIGR Gene indices, and sequence comparison was conduced with stand-alone blast [36]. Multiple sequence alignments were performed with ClustalW [37] and the conserved protein domains were searched in Interpro [38]. Synonymous and non-synonymous substitutions in newly cloned genes were determined by Dnasp [39]. Maximum Likelihood (ML) phylogenetic analyses were performed with Phylip [40].
PCR cloning
The conserved sequence in the untranslated regions of rainbow trout Tc1-like transposases were inferred from EST sequences. RNA was extracted from rainbow trout brain and treated with Rnase-free Dnase (Promega). Reverse transcription with SuperScriptIII (Invitrogen) was primed with oligo(dT). PCR was performed with primer 5'-ATACAGTGCCTTGCGAGAGTATTC-3' using a TripleMaster kit (Eppendorf), and the product was cloned into pcDNA3.1/V5-His-TOPO (Invitrogen). Seven of nine sequenced clones contained complete reading frames.
PCR analyses of genomic DNA
The fish samples were collected from inland reservoirs in Finland, and DNA from fin clips was prepared with salt extraction [41]. In brief, fin samples were digested at 60°C in 440 μl of buffer (1.8 mM EDTA, 9 mM Tris-HCl, pH 8; 1.8% SDS) containing 160 μg of proteinase K. After addition of 300 μl of 6 M NaCl, lysates were centrifuged at 12,000 g for 30 min. DNA in the supernatants was precipitated with isopropanol, washed with 70% aqueous ethanol and dissolved in water. The 654-base fragments of Glan PCR were amplified using the Hot Master Taq kit (Eppendorf). Primers (5'-TGAAGAATCGACAACAAGTGGGACA-3' and 5'-GCTTTCTTCTTGCCACTCTTCCATA-3') were annealed to templates at 68°C.
Microarray analyses
Fish experiments, design of the rainbow trout cDNA microarray, hybridization protocol and data analyses are described in detail elsewhere [10,11]. In brief, the platform included 1,300 genes printed in 6 replicates each. The dye swap design was used; each sample containing RNA from 4 individuals was hybridized to slides with reverse assignment of fluorescent dyes (Cy3- and Cy5-dCTP from Amersham Pharmacia). Labels were incorporated at the stage of cDNA synthesis. The measurements in spots were filtered by criteria I/B ≥ 3 and (I-B)/(SI+SB) ≥ 0.6, where I and B were the mean signal and background intensities, respectively, and SI, SB were the standard deviations. Lowess normalization was performed and differential expression was analysed with the Student's t-test (p < 0.01). The genes were ranked by the log(p-level).
Abbreviations
MGE – mobile genetic element; ML – maximum likelihood.
Authors' contributions
AK carried out sequence analyses and drafted the manuscript, HK conducted the microarray analyses, SA performed the statistical analyses and HM coordinated research. All authors read and approved the final manuscript.
Acknowledgements
This study was supported by the National Agency of Technology, Finland. We wish to thank Rolf Sara for preparation of the microarrays and Miia Antikainen for assistance in cloning the rainbow trout transposons.
==== Refs
Plasterk RH Izsvak Z Ivics Z Resident aliens: the Tc1/mariner superfamily of transposable elements Trends Genet 1999 15 326 332 10431195 10.1016/S0168-9525(99)01777-1
Robertson HM Soto-Adames FN Walden KKO Avancini RMP Lampe DJ Syvanen M, Kado CI The mariner transposons of animals: horizontally jumping genes Horizontal gene transfer 1998 London: Chapman and Hall 268 284
Goodier JL Davidson WS Tc1 transposon-like sequences are widely distributed in salmonids J Mol Biol 1994 241 26 34 8051704 10.1006/jmbi.1994.1470
Radice AD Bugaj B Fitch DH Emmons SW Widespread occurrence of the Tc1 transposon family: Tc1-like transposons from teleost fish Mol Gen Genet 1994 244 606 612 7969029 10.1007/BF00282750
Reed KM Tc1-Like Transposable Elements in the Genome of Lake Trout (Salvelinus namaycush) Mar Biotechnol 1999 1 60 67 10373611
Leaver MJ A family of Tc1-like transposons from the genomes of fishes and frogs: evidence for horizontal transmission Gene 2001 271 203 214 11418241 10.1016/S0378-1119(01)00530-3
Quackenbush J Cho J Lee D Liang F Holt I Karamycheva S Parvizi B Pertea G Sultana R White J The TIGR Gene Indices: analysis of gene transcript sequences in highly sampled eukaryotic species Nucleic Acids Res 2001 29 159 164 11125077 10.1093/nar/29.1.159
Pertea G Huang X Liang F Antonescu V Sultana R Karamycheva S Lee Y White J Cheung F Parvizi B TIGR Gene Indices clustering tools (TGICL): a software system for fast clustering of large EST datasets Bioinformatics 2003 19 651 652 12651724 10.1093/bioinformatics/btg034
Kohler SW Provost GS Fieck A Kretz PL Bullock WO Sorge JA Putman DL Short JM Spectra of spontaneous and mutagen-induced mutations in the lacI gene in transgenic mice Proc Natl Acad Sci USA 1991 88 7958 7962 1832771
Koskinen H Pehkonen P Vehniäinen E Krasnov A Rexroad C Afanasyev S Mölsä H Oikari A Response of rainbow trout transcriptome to model chemical contaminants Biochem Biophys Res Commun 2004 320 745 753 15240111 10.1016/j.bbrc.2004.06.024
Krasnov A Koskinen H Pehkonen P Rexroad CE Afanasyev S Mölsä H Gene expression in the brain and kidney of rainbow trout in response to handling stress BMC Genomics 2005 6 3 15634361 10.1186/1471-2164-6-3
Ashburner M Ball CA Blake JA Botstein D Butler H Cherry JM Davis AP Dolinski K Dwight SS Eppig JT Harris MA Hill DP Issel-Tarver L Kasarskis A Lewis S Matese JC Richardson JE Ringwald M Rubin GM Sherlock G Gene ontology: tool for the unification of biology. The Gene Ontology Consortium Nat Genet 2000 25 25 29 10802651 10.1038/75556
Ivics Z Hackett PB Plasterk RH Izsvak Z Molecular reconstruction of Sleeping Beauty, a Tc1-like transposon from fish, and its transposition in human cells Cell 1997 91 501 510 9390559 10.1016/S0092-8674(00)80436-5
Izsvak Z Ivics Z Plasterk RH Sleeping Beauty, a wide host-range transposon vector for genetic transformation in vertebrates J Mol Biol 2000 302 93 102 10964563 10.1006/jmbi.2000.4047
Lampe DJ Witherspoon DJ Soto-Adames FN Robertson HM Recent horizontal transfer of mellifera subfamily mariner transposons into insect lineages representing four different orders shows that selection acts only during horizontal transfer Mol Biol Evol 2003 20 554 562 12654937 10.1093/molbev/msg069
Rudenko GN Ono A Walbot V Initiation of silencing of maize MuDR/Mu transposable elements Plant J 2003 33 1013 1025 12631326 10.1046/j.1365-313X.2003.01683.x
Sijen T Plasterk RH Transposon silencing in the Caenorhabditis elegans germ line by natural RNAi Nature 2003 426 310 314 14628056 10.1038/nature02107
Boonanuntanasarn S Yoshizaki G Takeuchi T Specific gene silencing using small interfering RNAs in fish embryos Biochem Biophys Res Commun 2003 310 1089 1095 14559227 10.1016/j.bbrc.2003.09.127
Hartl DL Lohe AR Lozovskaya ER Modern thoughts on an ancient marinere: function, evolution, regulation Annu Rev Genet 1997 31 337 358 9442899 10.1146/annurev.genet.31.1.337
Kordis D Gubensek F Molecular evolution of Bov-B LINEs in vertebrates Gene 1999 238 171 178 10570995 10.1016/S0378-1119(99)00260-7
Akhtar A Dosage compensation: an intertwined world of RNA and chromatin remodelling Curr Opin Genet Dev 2003 13 161 169 12672493 10.1016/S0959-437X(03)00016-9
Stevenson DS Jarvis P Chromatin silencing: RNA in the driving seat Curr Biol 2003 13 R13 15 12526757 10.1016/S0960-9822(02)01380-5
Wutz A RNAs templating chromatin structure for dosage compensation in animals Bioessays 2003 25 434 442 12717814 10.1002/bies.10274
Jang KL Collins MK Latchman DS The human immunodeficiency virus tat protein increases the transcription of human Alu repeated sequences by increasing the activity of the cellular transcription factor TFIIIC J Acquired Immune Defic Syndr 1992 5 1142 1147 1403646
Fornace AJ Mitchell JB Induction of B2 RNA polymerase III transcription by heat shock: enrichment for heat shock induced sequences in rodent cells by hybridization subtraction Nucleic Acids Res 1986 14 5793 5811 2426659
Li T-H Spearow J Rubin CM Schmid CW Physiological stresses increase mouse short interspersed element (SINE) RNA expression in vivo Gene 1999 239 367 372 10548739 10.1016/S0378-1119(99)00384-4
Kimura RH Choudary PV Schmid CW Silkworm Bm1 SINE RNA increases following cellular insults Nucleic Acids Res 1999 27 3380 3387 10454647 10.1093/nar/27.16.3380
Chu WM Ballard R Carpick BW Williams BRG Schmid CW Potential Alu Function: Regulation of the Activity of Double-Stranded RNA-Activated Kinase PKR Mol Cell Biol 1998 18 58 68 9418853
Rubin CM Kimura RH Schmid CW Selective stimulation of translational expression by Alu RNA Nucleic Acids Res 2002 30 3253 3261 12136107 10.1093/nar/gkf419
Jolly C Metz A Govin J Vigneron M Turner BM Khochbin S Vourc'h C Stress-induced transcription of satellite III repeats J Cell Biol 2004 164 25 33 14699086 10.1083/jcb.200306104
Trobridge GD Chiou PP Kim CH Leong JC Induction of the Mx protein of rainbow trout Oncorhynchus mykiss in vitro and in vivo with poly I:C dsRNA and infectious haematopoietic necrosis virus Dis Aquat Org 1997 30 91 98
Jensen I Larsen R Robertsen B An antiviral state induced in Chinook salmon embryo cells (CHSE-24) by transfection with the double stranded RNA poly I: C Fish Shellfish Immunol 2002 13 367 378 12458743 10.1006/fsim.2002.0412
Labrador M Corces VG Transposable element-host interactions: regulation of insertion and excision Annu Rev Genet 1997 31 381 404 9442901 10.1146/annurev.genet.31.1.381
Britten RJ Davidson EH Gene regulation for higher cells: a theory Science 1969 165 349 357 5789433
Schramke V Allshire R Hairpin RNAs and retrotransposon LTRs effect RNAi and chromatin-based gene silencing Science 2003 301 1069 1074 12869699 10.1126/science.1086870
Altschul SF Madden TL Schaffer AA Zhang J Zhang Z Miller W Lipman DJ Gapped BLAST and PSI-BLAST: a new generation of protein database search programs Nucleic Acids Res 1997 25 3389 3402 9254694 10.1093/nar/25.17.3389
Thompson JD Higgins DG Gibson TJ CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice Nucleic Acids Res 1994 22 4673 4680 7984417
Mulder NJ Apweiler R Attwood TK Bairoch A Barrell D Bateman A Binns D Biswas M Bradley P Bork P Bucher P Copley RR Courcelle E Das U Durbin R Falquet L Fleischmann W Griffiths-Jones S Haft D Harte N Hulo N Kahn D Kanapin A Krestyaninova M Lopez R Letunic I Lonsdale D Silventoinen V Orchard SE Pagni M Peyruc D Ponting CP Selengut JD Servant F Sigrist CJA Vaughan R Zdobnov EM The InterPro Database, 2003 brings increased coverage and new features Nucleic Acids Res 2003 31 315 318 12520011 10.1093/nar/gkg046
Rozas J Rozas R DnaSP version 3: an integrated program for molecular population genetics and molecular evolution analysis Bioinformatics 1999 15 174 175 10089204 10.1093/bioinformatics/15.2.174
Felsenstein J PHYLIP – Phylogeny Inference Package (Version 3.2) Cladistics 1989 5 164 166
Aljanabi SM Martinez I Universal and rapid salt-extraction of high quality genomic DNA for PCR-based techniques Nucleic Acids Res 1997 25 4692 4693 9358185 10.1093/nar/25.22.4692
|
16095544
|
PMC1192797
|
CC BY
|
2021-01-04 16:32:48
|
no
|
BMC Genomics. 2005 Aug 12; 6:107
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-107
|
oa_comm
|
==== Front
BMC GenomicsBMC Genomics1471-2164BioMed Central London 1471-2164-6-1101610722010.1186/1471-2164-6-110Research ArticleReverse transcriptional profiling: non-correspondence of transcript level variation and proximal promoter polymorphism Brown Rebecca Petersen [email protected] Martin E [email protected] Department of Organismal Biology & Anatomy, The University of Chicago, Chicago, IL 60637, USA2005 17 8 2005 6 110 110 23 6 2005 17 8 2005 Copyright © 2005 Brown and Feder; licensee BioMed Central Ltd.2005Brown and Feder; licensee BioMed Central Ltd.This is an Open Access article distributed under the terms of the Creative Commons Attribution License (), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Background
Variation in gene expression between two Drosophila melanogaster strains, as revealed by transcriptional profiling, seldom corresponded to variation in proximal promoter sequence for 34 genes analyzed. Two sets of protein-coding genes were selected from pre-existing microarray data: (1) those whose expression varied significantly and reproducibly between strains, and (2) those whose transcript levels did not vary. Only genes whose regulation of expression was uncharacterized were chosen. At least one kB of the proximal promoters of 15–19 genes in each set was sequenced and compared between strains (Oregon R and Russian 2b).
Results
Of the many promoter polymorphisms, 89.6% were SNPs and 10.4% were indels, including homopolymer tracts, microsatellite repeats, and putative transposable element footprints. More than half of the SNPs were changes within a nucleotide class. Hypothetically, genes differing in expression between the two strains should have more proximal promoter polymorphisms than those whose expression is similar. The number, frequency, and type of polymorphism, however, were the same in both sets of genes. In fact, the promoters of six genes with significantly different mRNA expression were identical in sequence.
Conclusion
For these genes, sequences external to the proximal promoter, such as enhancers or in trans, must play a greater role than the proximal promoter in transcriptomic variation between D. melanogaster strains.
==== Body
Background
Transcriptional profiling via whole genome arrays has revealed both impressive diversity and unexpected commonalities in gene expression. Some fraction of this variation results in the various functional, developmental, and reproductive phenotypes that are downstream of gene expression [1,2]. The source of this variation, however, is in the nucleotides that regulate gene expression (as reviewed in [3]). Given the relationship between these nucleotides and gene expression, it should in principle be possible to reverse this relationship and retrodict the nucleotide sequences that have given rise to variation in gene expression. Indeed, this is the basis for numerous bioinformatic efforts to identify novel cis-regulatory motifs if not to elucidate the entire cis-regulatory code [4,5] from sequence that is conserved among genes with similar expression patterns. Here we ask a somewhat different but equally fundamental question of sequence that varies among genes with dissimilar expression: Are regulatory regions different for genes differing in expression, and similar for genes not differing in expression?
We ask this question of cis-regulatory sequence immediately upstream of the transcription start site and of similarities or differences in gene expression between members of the same species. Candidate cis-regulatory variation includes polymorphisms in transcription factor binding sites, multiallelic tandem repeat variation, such as microsatellites, single nucleotide polymorphisms (SNPs), and transposable element insertions. These polymorphisms can alter transcription rates and affect mRNA expression levels in vivo [6-10]. Functional cis-regulatory variation is widespread in the human genome [11] and these sites have a high degree of variability (about 0.6% of these sites are polymorphic [12]). Clearly the source of variation in gene expression need not lie in the proximal promoter and could be either in cis but far up- or downstream, or in trans. Lack of correlated variation in proximal promoter sequence and gene expression can be viewed as implicating these alternative sources.
We focus on variation in gene expression among members of the same species because this is the raw material of evolution. For evolutionary change to ensue by ordinary Darwinian mechanisms, variation within a species must exist, be heritable, and be consequential for fitness. Both classical [13,14] and recent [15-17] and others reviewed in [18] works implicate variation in gene expression among members of a population, controlled by regulatory as opposed to coding sequence, as a principal source of such variation. Although simple sequence analysis of regulatory regions from diverse and related members of a species can readily ascertain whether sequence variation is present and heritable, ascertainment of impact on gene expression from sequence alone is less precise. Although differences in gene expression are not necessarily consequential for fitness, if variation in regulatory sequence affects gene expression (or that similarity in sequence reliably corresponds to no variation in gene expression), then intraspecific variability in regulatory sequence could be used as a preliminary screen for evolvability. Indeed, sequence analysis of regulatory regions may potentially assess the evolutionary forces that shape them [15].
To test this possibility, we mined the whole-genome transcriptional profiles of two near-isogenic strains (Oregon R and Russian 2b) of Drosophila melanogaster, here representing variation within species. The original data [1] are for genes whose expression does or does not vary between non-virgin adults of the Oregon R and Russian 2b strains. Of the 12017 genes examined, 527 do not differ between strains and 483 differ between strains regardless of sex. With no a priori knowledge of these genes' sequence variation or function, we have selected 15 and 19 representatives of each class, respectively. We chose genes whose regulation is yet uncharacterized to avoid bias due to over- or under-sampling of genes with known transcriptional regulation, and included both known transcription response elements/conserved motifs and uncharacterized sequence in our analysis. We now report whether or not their proximal promoter sequence varies and, if so, whether or not this variation corresponds to variation in gene expression.
Results
Similar numbers of genes composed the two genesets. 483 candidate genes whose expression varied reproducibly and highly significantly between the strains with no sex or sex by line effects (P < 0.01 for line and P > 0.05 for sex and sex by line) composed the first geneset. The second geneset comprised 527 candidate genes whose expression did not vary at all (P > 0.1 for sex, line, and sex by line). The remaining 11007 genes, whose significance values fell in between these P values, were not analyzed. Within the first geneset, 172 genes were more actively transcribed in Russian 2b females than in Oregon R females and 311 genes in Oregon R than Russian 2b females. The expression of 154 genes was greater for Russian 2b males than for Oregon R males and 329 genes for Oregon R than Russian 2b males. Thus, expression was consistently greater for the Oregon strain than the Russian strain (Table 1).
Table 1 Categorization of microarray data set.
Genes whose expression varies only between strains (P < 0.01) with no sex or sex by line effects (P > 0.05) Genes whose expression does not vary between sex, strain, or sex by strain (P > 0.1)
Strain Female Male
Russian 2b 172 154 527
Oregon R 311 329
Total 483 483
Many genes in the first geneset are false positives. The expected number of candidate genes in the first geneset, 325, is less than 483, the observed number resulting in a False Discovery Rate (FDR) of 67.4%. Thus, of the 483 genes chosen in the first gene set, 67.4% or ~325 genes are called as significantly different when they are not. In other words, the expression of ~158 of the 483 genes in the first geneset presumably differs significantly between the two strains.
Polymorphisms are numerous in both genesets. Of the 34 promoters analyzed, six do not differ between strains; however, they belong to genes whose transcripts differ between strains. The remaining 28 proximal promoters contain at least one polymorphism, with one promoter containing as many as 37 polymorphisms (Figures 1 and 2). The mean, median, and mode of polymorphisms per gene are 8.5, 6, and 0, respectively. Of the 288 total polymorphisms detected in at least 1 kb of the proximal promoters, 258 (89.6%) were SNPs and 30 (10.4%) were indels. Over half of the SNPs were transitions within a nucleotide class, while 43.8% were transversions between nucleotide classes (Figure 3). 239 putative binding sites for transcription factors were created or removed by these proximal promoter polymorphisms; thus 258/288 or 90% of proximal promoter polymorphisms fall within putative transcription factor binding sites (Figures 1 and 2).
Figure 1 Schematics of proximal promoters. At least one kb of the proximal promoters of 34 candidate genes whose transcripts vary (left and center columns) or do not vary (right column) in expression between D. melanogaster strains. Genes whose expression is greater in the Russian 2b (R2b) strain are shown in the left column and those whose expression is greater in the Oregon R (OrR) strain in the center column. Genes with fewer polymorphisms are shown in Figure 1 and those with seven or more are shown in Figure 2. The long horizontal lines for each gene designate the sequence of the proximal promoter with changes in the R2b strain shown above the line and those in the OrR strain below the line. The large, bent arrows indicate transcriptional start sites. Single Nucleotide Polymorphisms (SNPs) are shown as small vertical lines and indels as triangles, with the sequence or length in nucleotides (nt). The numbers of SNPs are listed where there are too many to illustrate clearly. Putative transcription factor binding sites created or removed by the SNP or indel are shown above or below the polymorphisms, with short horizontal lines or arrows designating the included polymorphisms. Because there are too many to illustrate clearly, putative transcription factor binding sites are not shown for CYP9C1 (CG3616), Fkbp13 (CG9847), qkr58E-3 (CG3584), bin (CG18647), tensin (CG9379), Cry (CG16963), KP78a (CG6715), Cng (CG7779), Fer2 (CG5952), Mt2 (CG10692). Shaded boxes designate introns in the 5'untranslated regions (UTRs).
Figure 2 Schematics of proximal promoters. At least one kb of the proximal promoters of 34 candidate genes whose transcripts vary (left and center columns) or do not vary (right column) in expression between D. melanogaster strains. Genes whose expression is greater in the Russian 2b (R2b) strain are shown in the left column and those whose expression is greater in the Oregon R (OrR) strain in the center column. Genes with fewer polymorphisms are shown in Figure 1 and those with seven or more are shown in Figure 2. The long horizontal lines for each gene designate the sequence of the proximal promoter with changes in the R2b strain shown above the line and those in the OrR strain below the line. The large, bent arrows indicate transcriptional start sites. Single Nucleotide Polymorphisms (SNPs) are shown as small vertical lines and indels as triangles, with the sequence or length in nucleotides (nt). The numbers of SNPs are listed where there are too many to illustrate clearly. Putative transcription factor binding sites created or removed by the SNP or indel are shown above or below the polymorphisms, with short horizontal lines or arrows designating the included polymorphisms. Because there are too many to illustrate clearly, putative transcription factor binding sites are not shown for CYP9C1 (CG3616), Fkbp13 (CG9847), qkr58E-3 (CG3584), bin (CG18647), tensin (CG9379), Cry (CG16963), KP78a (CG6715), Cng (CG7779), Fer2 (CG5952), Mt2 (CG10692). Shaded boxes designate introns in the 5'untranslated regions (UTRs).
Figure 3 Categories of SNPs between two strains. We identified 258 single nucleotide polymorphisms (SNPs) in 1–2 kb of the promoter region 5' of the translational start site of genes whose expression does and does not vary on microarray. SNPs are reported here irrespective of the direction of change. Transitions, or changes within a nucleotide class such as A to G or C to T, comprise 56.2% of the number of SNPs, whereas transversions, or changes between nucleotide classes, occur less frequently.
Although few, indels varied between the strains by kind and from 1 to 43 nt in length [see Additional File 1]. Indels were classified as direct repeats (dr), homopolymer repeats (hpr), microsatellites (mcs), or non-repetitive (nr) according to designations of Schaeffer [19], page 165. In most cases, indels in one strain were the same as in the Celera strain used as a reference. However, six repeats in five different genes resulted in differences in sequence among the Russian 2b, Oregon R, and Celera strains. In bin (CG18647), five additional Ts are found in the Oregon R strain in comparison with the poly-T4 tract in the Russian 2b strain. This caused a transition and insertion from T4CT3 in the Celera strain. Distal to this in bin, the ATACCCGTACCCGTACCCAT sequence in the Russian 2b strain was shortened to ATACCCGTACCCAT in the Celera strain but absent altogether in the Oregon R strain. In tacc (CG9765), the poly-A tract varies from A14–17 nt among individuals in the Russian 2b strain, to A23–26 nt among individuals in the Oregon R strain, to A22 in the Celera strain. In qkr58E-3 (CG3584), a SNP and variation in the length of a poly-T tract resulted in T3GT4 in the Oregon R strain, T9 in the Celera strain and T11 in the Russian 2b strain. In KP78a (CG6715), the dinucleotide microsatellite AC was repeated 9X in the Russian 2b strain, 10X in the Celera strain, and 11X in the Oregon R strain. In stan (CG11895), the homopolymer repeat was T10 nt long in the Celera strain, T11 in the Russian 2b strain, and T9 in the Oregon R strain.
Indels in Scab, Cry, Ih, and bin (CG8095, CG16963, CG8585, CG18647) contained small regions (12–15 nt long) that shared sequence similarity with known transposable elements. Despite these matches, the sequences are not long enough to discriminate confidently between a TE footprint and chance occurrence of the same sequence. Flanking the indel in Cry listed in the Supplementary Table [see additional file 1], a 152 bp sequence from -889 to -1041 relative to the translational start site in the Oregon R strain matched a DNA LINE retroelement. In the same location in the Russian 2b strain a 152 bp sequence matched two overlapping DNAREP1_DM LINE elements [20].
The diversity and frequency of polymorphisms did not differ in proximal promoters from genes differing in expression and those with similar expression (P = 0.911 for indels, P = 0.935 for transition SNPs, and P = 0.842 for transversion SNPs) (Figures 1 and 2). For example, we identified 59 transversion SNPs, 76 transition SNPs (summing to 135 total SNPs), and 16 indels in the first geneset, and 54 transversion SNPs, 69 transition SNPs (123 total SNPs), and 14 indels in the second geneset. Also, the average promoter length in the first geneset was 1629 nt and 1620 nt in the second geneset (Table 2). Thus, for the 34 genes examined in this study, the lack of variation in proximal promoter sequence between the two genesets implicates alternative sources for divergent patterns of gene expression.
Table 2 Genes used in this study. From the data of Gibson et al. (2004), 34 candidate genes were chosen for study whose transcripts were expressed to higher levels in either Oregon R (OrR) or Russian 2b (R2b) strain (P < 0.01) (first geneset) or did not vary between strains (P > 0.1) (second geneset) and did not vary between the sexes or have sex by strain interaction (P > 0.05 or P > 0.1, for the two genesets respectively). (CODE, CG number from FlyBase Release 3.0; EXP. DIFF., indicates where gene expression on microarray does or does not vary between strains; N, number of replicate probes representing gene on microarray; CHR., chromosomal location; LEN., length of promoter analyzed in this study; ACC. #, GenBank accession numbers for gene's proximal promoter in the OrR strain followed by the R2b strain)
CODE NAME EXP. DIFF. N CHR. FUNCTION LEN. ACC. #
CG1762 Bintv OrR>R2b 4 2L cell adhesion 1518 DQ017407,6
CG1944 CYP4P2 OrR>R2b 5 2R steroid metabolism 1530 DQ017409,8
CG5105 Plap OrR>R2b 5 2L phospholipase A2 activator 1539 DQ017395,4
CG5952 Fer2 OrR = R2b 3 3R transcription factor 1452 DQ017400,1
CG6547 mRpL2a OrR = R2b 3 3R ribosome structure 1448 DQ017398,9
CG7779 Cng OrR = R2b 5 2R cation channel 1622 DQ017397,6
CG8095 Scab OrR>R2b 5 2R cell adhesion 1652 DQ017403,2
CG9765 tacc OrR = R2b 3 3R microtubule binding 1428 DQ017405,4
CG10488 eyg OrR>R2b 4 3L transcription factor 1494 DQ017410,1
CG14827 mei-P22 OrR = R2b 4 3L meiotic recombination 1392 DQ017412,3
CG1922 onecut OrR = R2b 3 4 transcription factor 2410 DQ017414,5
CG3584 qkr58E-3 R2b>OrR 3 2R RNA binding 1720 DQ017416,7
CG3616 CYP9C1 R2b>OrR 3 2R unknown 1645 DQ017418,9
CG3758 esg OrR = R2b 3 2L transcription factor 1890 DQ017420,1
CG4485 CYP9B1 R2b>OrR 3 2R unknown 1629 DQ017422,3
CG4871 ST6Gal OrR>R2b 3 2R polysaccharide metabolism 1776 DQ017424,5
CG5517 Ide OrR>R2b 3 3L unknown 1746 DQ017426,7
CG6715 KP78a OrR = R2b 3 3R protein kinase 1623 DQ017428,9
CG8585 Ih OrR = R2b 3 2R cation channel 1516 DQ017430,1
CG9847 Fkbp13 R2b>OrR 4 2R protein folding 1676 DQ017432,3
CG10002 fkh OrR = R2b 3 3R transcription factor 1857 DQ017434,5
CG10094 CYP313A2 OrR = R2b 3 3R steroid metabolism 1537 DQ017436,7
CG10692 Mt2 OrR = R2b 3 2L DNA methylation 1577 DQ017438,9
CG11084 pk OrR = R2b 3 2R cell polarity 1523 DQ017440,1
CG11186 toy OrR>R2b 3 4 transcription factor 2061 DQ017442,3
CG11895 stan OrR = R2b 3 2R GPC receptor 1636 DQ017444,5
CG15807 CYP313A5 OrR = R2b 3 3R steroid metabolism 1390 DQ017446,7
CG16963 Cry OrR>R2b 3 2L eye lens structure 1769 DQ017448,9
CG18647 bin R2b>OrR 3 3L transcription factor 1595 DQ017450,1
CG4088 lat R2b>OrR 3 2R olfactory learning 2000 DQ017455,4
CG9379 by/tensin R2b>OrR 3 3R actin binding 1498 DQ017456,7
CG9712 TSG101 R2b>OrR 3 3L ubiquitin-protein ligase 1374 DQ017453,2
CG13432 l(2)05510 R2b>OrR 3 2R unknown 1404 DQ017458,9
CG31908 NA R2b>OrR 2 2L unknown 1328 DQ017460,1
Discussion
The variation in proximal promoter sequence among the Oregon R, Russian 2b, and Celera strains, while extensive, typifies that among individuals and populations of eukaryotes. In a similar survey of 107 transcriptionally active genes in the human genome, Rockman and Wray [11] identified 140 experimentally validated cis-regulatory polymorphisms resulting in two-fold or greater variation in transcription rate and subsequent gene expression. In another survey of regulatory variation, Cowles et al. [21] found a relatively high frequency (6%) of cis-regulatory polymorphism (including SNPs, complex nucleotide repeats, insertions/deletions, and microsatellites) in the 1 kb region 5' of the predicted transcription start site of genes in four inbred mouse strains with allelic differences in expression by 1.5-fold or greater.
The frequencies of SNPs and indels were not independent. In fact, indels, including microsatellites, homopolymer repeats, and possibly TE footprints (see next paragraph), co-occurred with SNPs but SNPs could occur in the absence of indels. This suggests that indels occur infrequently and only after a promoter is poised for mutation via the presence of SNPs [22,23].
Some promoters have diverged remarkably between the two strains. For example, the Cry promoter contains 28 SNPs, 7 indels, and a 152 bp sequence including and flanking one indel that matches two different classes of transposable elements a DNA LINE element in the Oregon R strain and 2 overlapping DNAREP1_DM LINE elements in the Russian 2b strain [20]. This variation suggests that the Cry promoter is susceptible to mutation.
The polymorphisms are in regions within the proximal promoter that ought to affect gene expression. 90% of them fall within putative transcription factor binding sites (Figures 1 and 2). In addition, all regulatory information necessary for transcription in Drosophila is generally present within 1 kb of the basal promoter [24].
Surprisingly, genes differing in expression and genes not differing in expression had the same diversity and frequency of polymorphisms in their proximal promoters. Indeed, the proximal promoters from six genes whose expression differed were identical in sequence between the two strains. Because this region includes core regulatory sequences, polymorphisms between the two genesets (or their absence) ought reasonably be correlated with gene expression. In support of this expectation, 74% of transcription factor binding sites (as proxies for proximal promoter polymorphisms) in yeast genes are between 100 to 500 bases upstream of the translational start site and fewer than expected lie outside this region [25]. Thus, the non-concordance between proximal promoter sequence and gene expression is unexpected.
One possible explanation for this outcome is that the 34 genes examined are anomalous or unrepresentative of Drosophila genes in general. This explanation is unlikely. As discussed above, these genes' proximal promoters resemble other genes' in variation among individuals. Choice of these genes was also unbiased with respect to prior knowledge of transcriptional regulation, known transcriptional response elements, and mode of gene effect (e.g., dominance, underdominance, or additivity). Indeed, 5/34 (15%) of the blindly chosen genes were dominant, underdominant, or additive in their impact on gene expression, which compares favorably with the proportions of such effects for the Drosophila genome in general [1].
A second explanation for non-concordance between proximal promoter sequence and gene expression is that the expression of the genes in the first geneset actually does not differ between strains. To exclude this possibility, the first geneset included only genes whose expression profiles were different between strains (P < 0.01), and not between sexes or by the sex*strain interaction (P > 0.05 for both) [1]. These selection criteria confer a rather high False Discovery Rate of 67.4%. The actual statistical significance of variation between the two strains averages <0.0009 for the 19 genes in the first geneset. An additional 333 genes also meet this criterion. For these 352 genes (total), the FDR becomes ((0.0009)*(0.95)*(0.95)*(12017)*(3))/352 = 29.28/352 = 0.0832 or 8.3%. This low rate indicates 91.7% certainty that the 19 genes in this geneset truly differ in expression.
A third explanation is that some of these promoter variants may affect gene expression but that neutral sites dilute their regulatory activities. Functionally important regulatory sites may make up only a small fraction of the many promoter variants detected in the 1 kb region and the vast majority may be neutral. On the other hand, promoter variants whose expression levels are similar may be functionally equivalent (or neutral) as a result of stabilizing selection or epigenetic forces may even out promoter variation resulting in similar expression levels. Functional assays should be able to distinguish between these outcomes.
Another possibility is that the regulatory elements responsible for the observed variation in transcript abundance occur outside the proximal promoter, including in introns and 5' and 3' untranslated regions, and/or in trans. Several recent localizations of regulatory sequence also support this possibility. In a recent assignment of 142 expression phenotypes in humans to both cis- and trans-acting loci, 27 (19%) had a single cis-acting regulator (defined as regulators that map within 5 megabases of the target gene), 110 (77.5%) had a single trans-acting regulator (those that map elsewhere), and 5 (3.5%) had two or more regulators acting in cis and/or trans [26]. In budding yeast, trans-acting regulators were linked to 365 of 570 (64%) genes whose expression diverged between a laboratory and a wild strain of Saccharomyces cerevisiae whereas only 36% of divergent gene expression was linked to individual cis-acting loci [27]. Of 2294 genes whose expression differs between laboratory and wild yeast strains, 1716 (75%) map to 100–200 distinct trans-acting loci with widespread genomic effects and molecular functions but that do not encode transcription factors [28]. Of 3546 highly heritable transcripts in the same two strains of yeast, 3% map to a single locus, 17–18% to 1–2 loci, and half to more than 5 loci [29]. Using gene expression differences as quantitative traits, Schadt et al. [30] determined that genes involved in determining patterns of obesity in mice and humans map to QTLs other than the genes themselves. Many genes map to the same QTL, suggesting these are trans-regulatory hotspots.
This study leads us to suggest that the relationship of coding sequence conservation and functional similarity may not be true for cis-regulatory sequences. Indeed, recent work as shown this to be the case for enhancers [31]. The problem, however, is that there is not a precise quantitative framework for the interpretation of cis-regulatory variation.
Conclusion
In summary, we have characterized the frequency and diversity of proximal promoter polymorphisms on a genome-wide scale and shown that they do not differ between D. melanogaster strains with divergent gene expression. We conclude that sequences elsewhere, such as in trans, may cause differential gene expression. Indeed, in the very strains examined in the present study, extensive nonadditivity of gene expression among these strains and their reciprocal F1 hybrids indicates that transcription is controlled predominantly by trans-regulatory factors [1]. Linkage analysis might allow for mapping of regulatory regions outside the proximal promoter that cause expression variation within the D. melanogaster transcriptome.
Methods
Microarray data analysis
The data are from the work of Gibson et al. [1] with Agilent 60-mer oligonucleotide microarrays, and are available online [32]. P-values were calculated from the NLP (Negative Log P) values reported by Gibson et al. [1], and used to sort the data into two genesets whose expression reproducibly varied at a highly significant level (P < 0.01) or not at all (P > 0.1) between the two strains. Genes whose significance values fell in between these P values were ignored. Genes for which the effect of sex and/or sex by line interaction was significant (P < 0.05) were also excluded. Also, gene CG4109 was removed from this analysis because it was missing expression and significance values. For each gene in the first geneset, we identified the strain whose Least Squares Mean value was higher and designated it as the strain with greater expression.
From the genesets, we chose autosomal, protein-coding genes whose expression pattern was the same for at least three to five replicate probes, except for gene CG31908, which was probed only twice. To avoid ascertainment bias on the basis of known regulatory variation or lack thereof, only those genes whose regulation had not yet been studied were chosen for this study. Of these, 34 genes (ten whose expression is greater in the Russian 2b strain, nine in the Oregon R strain, fifteen whose expression does not vary) were chosen for inclusion. The expected number of genes whose expression varied between the two strains by chance alone was calculated by multiplying the P values at each selection criterion by the total number of tests (genes in the microarray) and the number of selection criteria: e.g., expected number of genes that vary = (0.01)*(0.95)*(0.95)*(12017)*(3) = 325. The False Discovery Rate (FDR), an indicator of the number of significant features that are truly null [33], was calculated by dividing the expected number by the observed: e.g., FDR = 325.36/483 = 0.6736 = 67.4%.
Flies
The strains from which the foregoing data were collected, Oregon R and Russian 2b, were obtained from Gregory Gibson at North Carolina State University. Before Gibson et al. [1] began their work, these strains were inbred by sib-mating for over a hundred generations and are more isogenic than isofemale lines (G. Gibson, personal communication). Flies were reared on standard cornmeal diet at room temperature without undergoing any experimentally imposed treatments.
Gene amplification and sequencing
Gene-specific primers for amplification and sequencing were designed with Oligo 4.0 (National Biosciences, Inc.) and gene sequences in FlyBase Release 3.0 [34]. During the course of this study, the entry for tensin was annotated and updated. Accordingly, the sequencing and analysis of this gene was modified to accommodate these changes. Primers were synthesized by Integrated DNA Technologies (Iowa City, IA) or at The University of Chicago Oligonucleotide Synthesis Core Facility.
Genomic DNA was extracted en masse from 50–100 male or female flies of either strain according to Lerman et al. [8] or using a DNeasy Tissue Kit (Qiagen). Between 1–2 kb of the proximal promoter for each gene was amplified with Taq DNA Polymerase (Promega), MasterAmp Extra Long DNA Polymerase Mix (Epicentre Technologies), Pfx (Invitrogen), or Pfu (Stratagene) according to the manufacturers' instructions. Promoter polymorphisms found in more than half (9/15) of the sequences amplified with Taq were verified with a proofreading polymerase. Because these nine sequences amplified with Taq were identical to those amplified with a proofreading polymerase, we assumed the remaining six sequences amplified with Taq also do not contain amplification errors. PCR products were cleaned with QIAquick PCR Purification Kits (Qiagen) or a Sephadex G-50 column and sequenced at The University of Chicago Cancer Research Center DNA Sequencing Facility with the original primer pair used for amplification and two pairs of internal primers. Primer sequences and PCR protocols are available upon request.
Bioinformatics
Overlapping sequences were compiled and edited using Sequencher 4.1 (Gene Codes, Corp.). The consensus sequences for each gene in both sexes of either strain were aligned to each other and to the Celera strain [35] from FlyBase Release 3.0, used as a reference, with ClustalW on Biology Workbench 3.2 [36] with default alignment parameters. To reduce PCR error and verify the identity of the promoter polymorphisms, we independently amplified and sequenced both sexes. For 25/34 genes (74%), both forward and reverse strands of the promoter in both sexes of both strains were sequenced multiple times, thus obtaining at least 4X coverage. For the remaining 9 genes (Cng, Scab, eyg, esg, Fkbp13, fkh, pk, toy, CG31908), sequence reads covered the promoter region at least twice in each strain, either by covering both strands in one sex or one strand in both sexes. The additional sequencing beyond 2X coverage did not change the base calls as no ambiguous bases were observed. Numbers of promoter polymorphisms (transition SNPs, transversion SNPs, indels) were tabulated and compared between the two genesets with a binomial model (R version 2.0.0; [37]). The two genesets were logit transformed before applying the following Generalized Linear Model:
glm(formula = Y1 ~ indel + transition_SNP + transversion_SNP, family = binomial(link = 'logit'))
where gene expression (the dependent variable X) varies according to the number of each indel, transition SNP or transversion SNP (each as the independent variable Y).
The MatInspector Tool v2.2 was used to search Genomatix Suite [38] for putative transcription factor binding sites created or abolished by the promoter polymorphisms. This database is based upon TRANSFAC and consensus sequences for putative transcription factor binding sites found in the scientific literature (Cartharius et al., unpublished). All matrices were searched with default settings; however only those putative transcription factor-binding sites with a core similarity of 0.75 or greater are reported here.
Insertions longer than 12 nt and full-length sequences of the promoters of four genes (Scab, Cry, Ih, bin) were BLASTed against a transposable element (TE) database for Drosophila extracted from the NCBI nr database (J.-C. Walser, pers. comm.). Only those hits with 100% similarity are included in this study. The direction of insertion or deletion cannot be determined from the data here as there is no outgroup for comparison.
Authors' contributions
R.P.B. conducted all experimental procedures including mining microarray data and selecting genes of interest, amplifying, sequencing and analyzing the proximal promoters, conducted bioinformatics and statistical analysis, and drafted the manuscript. M.E.F. helped conceive the study and draft the manuscript. Both authors read and approved the manuscript.
Supplementary Material
Additional File 1
Supplementary Table.doc is a Microsoft Word document that lists and categorizes the sequences of the insertions/deletions between two strains found in the proximal promoter regions of 34 genes.
Click here for file
Acknowledgements
We thank Greg Gibson, North Carolina State University, for sharing the microarray data and flies. Martha Alonzo, James Meza, and Teresa Rodgers assisted with gene amplification and sequencing and maintained fly stocks. Justin Borevitz and Geoffrey Morris helped with bioinformatic analysis of the microarray data set. Jake Byrnes and The University of Chicago Department of Statistics Consulting Office helped with statistical analysis of the promoter polymorphism data set. Jean-Claude Walser aided the search for TE footprints. This work is funded by NSF Postdoctoral Fellowship in Interdisciplinary Informatics (Award # DBI-0305982) to R. P. Brown and NSF IBN 03–16627 to M. E. Feder.
==== Refs
Gibson G Riley-Berger R Harshman L Kopp A Vacha S Nuzhdin S Wayne M Extensive sex-specific nonadditivity of gene expression in Drosophila melanogaster Genetics 2004 167 1791 1799 15342517 10.1534/genetics.104.026583
Levine M Davidson EH Gene regulatory networks for development Proc Natl Acad Sci U S A 2005 102 4936 4942 15788537 10.1073/pnas.0408031102
Wray GA Hahn MW Abouheif E Balhoff JP Pizer M Rockman MV Romano LA The evolution of transcriptional regulation in eukaryotes Mol Biol Evol 2003 20 1377 1419 12777501 10.1093/molbev/msg140
Emberly E Rajewsky N Siggia ED Conservation of regulatory elements between two species of Drosophila Bmc Bioinformatics 2003 4 57 14629780 10.1186/1471-2105-4-57
Istrail S Davidson EH Logic functions of the genomic cis-regulatory code Proc Natl Acad Sci U S A 2005 102 4954 4959 15788531 10.1073/pnas.0409624102
Kidwell MG Lisch D Transposable elements as sources of variation in animals and plants Proc Natl Acad Sci U S A 1997 94 7704 7711 9223252 10.1073/pnas.94.15.7704
Stone JR Wray GA Rapid evolution of cis-regulatory sequences via local point mutations Mol Biol Evol 2001 18 1764 1770 11504856
Lerman DN Michalak P Helin AB Bettencourt BR Feder ME Modification of heat-shock gene expression in Drosophila melanogaster populations via transposable elements Mol Biol Evol 2003 20 135 144 12519916 10.1093/molbev/msg015
Martin P Makepeace K Hill SA Hood DW Moxon ER Microsatellite instability regulates transcription factor binding and gene expression Proc Natl Acad Sci U S A 2005 102 3800 3804 15728391 10.1073/pnas.0406805102
Knight JC Regulatory polymorphisms underlying complex disease traits J Mol Med 2005 83 97 109 15592805 10.1007/s00109-004-0603-7
Rockman MV Wray GA Abundant raw material for cis-regulatory evolution in humans Mol Biol Evol 2002 19 1991 2004 12411608
Stephens JC Schneider JA Tanguay DA Choi J Acharya T Stanley SE Jiang RH Messer CJ Chew A Han JH Duan JC Carr JL Lee MS Koshy B Kumar AM Zhang G Newell WR Windemuth A Xu CB Kalbfleisch TS Shaner SL Arnold K Schulz V Drysdale CM Nandabalan K Judson RS Ruano G Vovis GF Haplotype variation and linkage disequilibrium in 313 human genes Science 2001 293 489 493 11452081 10.1126/science.1059431
Britten RJ Davidson EH Gene Regulation For Higher Cells - A Theory Science 1969 165 349 & 5789433
King MC Wilson AC Evolution At 2 Levels In Humans And Chimpanzees Science 1975 188 107 116 1090005
Kohn MH Fang S Wu CI Inference of positive and negative selection on the 5 ' regulatory regions of Drosophila genes Mol Biol Evol 2004 21 374 383 14660692 10.1093/molbev/msh026
Monks SA Leonardson A Zhu H Cundiff P Pietrusiak P Edwards S Phillips JW Sachs A Schadt EE Genetic inheritance of gene expression in human cell lines Am J Hum Genet 2004 75 1094 1105 15514893 10.1086/426461
Wittkopp PJ Haerum BK Clark AG Evolutionary changes in cis and trans gene regulation Nature 2004 430 85 88 15229602 10.1038/nature02698
Stamatoyannopoulos JA The genomics of gene expression Genomics 2004 84 449 457 15498452 10.1016/j.ygeno.2004.05.002
Schaeffer SW Molecular population genetics of sequence length diversity in the Adh region of Drosophila pseudoobscura Genet Res 2002 80 163 175 12688655 10.1017/S0016672302005955
Kapitonov VV Jurka J Molecular paleontology of transposable elements in the Drosophila melanogaster genome Proc Natl Acad Sci U S A 2003 100 6569 6574 12743378 10.1073/pnas.0732024100
Cowles CR Hirschhorn JN Altshuler D Lander ES Detection of regulatory variation in mouse genes Nature Genet 2002 32 432 437 12410233 10.1038/ng992
Bergman CM Kreitman M Analysis of conserved noncoding DNA in Drosophila reveals similar constraints in intergenic and intronic sequences Genome Res 2001 11 1335 1345 11483574 10.1101/gr.178701
Sinha S Siggia ED Sequence turnover and tandem repeats in cis-regulatory modules in Drosophila Mol Biol Evol 2005 22 874 885 15659554 10.1093/molbev/msi090
Arnosti DN Analysis and function of transcriptional regulatory elements: insights from Drosophila Annual Review of Entomology 2003 48 579 602 12359740 10.1146/annurev.ento.48.091801.112749
Harbison CT Gordon DB Lee TI Rinaldi NJ Macisaac KD Danford TW Hannett NM Tagne JB Reynolds DB Yoo J Jennings EG Zeitlinger J Pokholok DK Kellis M Rolfe PA Takusagawa KT Lander ES Gifford DK Fraenkel E Young RA Transcriptional regulatory code of a eukaryotic genome Nature 2004 431 99 104 15343339 10.1038/nature02800
Morley M Molony CM Weber TM Devlin JL Ewens KG Spielman RS Cheung VG Genetic analysis of genome-wide variation in human gene expression Nature 2004 430 743 747 15269782 10.1038/nature02797
Brem RB Yvert G Clinton R Kruglyak L Genetic dissection of transcriptional regulation in budding yeast Science 2002 296 752 755 11923494 10.1126/science.1069516
Yvert G Brem RB Whittle J Akey JM Foss E Smith EN Mackelprang R Kruglyak L Trans-acting regulatory variation in Saccharomyces cerevisiae and the role of transcription factors Nature Genet 2003 35 57 64 12897782 10.1038/ng1222
Brem RB Kruglyak L The landscape of genetic complexity across 5,700 gene expression traits in yeast Proc Natl Acad Sci U S A 2005 102 1572 1577 15659551 10.1073/pnas.0408709102
Schadt EE Monks SA Drake TA Lusis AJ Che N Colinayo V Ruff TG Milligan SB Lamb JR Cavet G Linsley PS Mao M Stoughton RB Friend SH Genetics of gene expression surveyed in maize, mouse and man Nature 2003 422 297 302 12646919 10.1038/nature01434
Ludwig MZ PAAEBCMNJKM Functional evolution of a cis-regulatory module. PLoS Biology 2005 3 e93 15757364 10.1371/journal.pbio.0030093
Gibson G
Storey JD Tibshirani R Statistical significance for genomewide studies Proc Natl Acad Sci U S A 2003 100 9440 9445 12883005 10.1073/pnas.1530509100
FlyBase
Adams MD Celniker SE Holt RA Evans CA Gocayne JD Amanatides PG Scherer SE Li PW Hoskins RA Galle RF George RA Lewis SE Richards S Ashburner M Henderson SN Sutton GG Wortman JR Yandell MD Zhang Q Chen LX Brandon RC Rogers YHC Blazej RG Champe M Pfeiffer BD Wan KH Doyle C Baxter EG Helt G Nelson CR Miklos GLG Abril JF Agbayani A An HJ Andrews-Pfannkoch C Baldwin D Ballew RM Basu A Baxendale J Bayraktaroglu L Beasley EM Beeson KY Benos PV Berman BP Bhandari D Bolshakov S Borkova D Botchan MR Bouck J Brokstein P Brottier P Burtis KC Busam DA Butler H Cadieu E Center A Chandra I Cherry JM Cawley S Dahlke C Davenport LB Davies A de Pablos B Delcher A Deng ZM Mays AD Dew I Dietz SM Dodson K Doup LE Downes M Dugan-Rocha S Dunkov BC Dunn P Durbin KJ Evangelista CC Ferraz C Ferriera S Fleischmann W Fosler C Gabrielian AE Garg NS Gelbart WM Glasser K Glodek A Gong FC Gorrell JH Gu ZP Guan P Harris M Harris NL Harvey D Heiman TJ Hernandez JR Houck J Hostin D Houston DA Howland TJ Wei MH Ibegwam C Jalali M Kalush F Karpen GH Ke ZX Kennison JA Ketchum KA Kimmel BE Kodira CD Kraft C Kravitz S Kulp D Lai ZW Lasko P Lei YD Levitsky AA Li JY Li ZY Liang Y Lin XY Liu XJ Mattei B McIntosh TC McLeod MP McPherson D Merkulov G Milshina NV Mobarry C Morris J Moshrefi A Mount SM Moy M Murphy B Murphy L Muzny DM Nelson DL Nelson DR Nelson KA Nixon K Nusskern DR Pacleb JM Palazzolo M Pittman GS Pan S Pollard J Puri V Reese MG Reinert K Remington K Saunders RDC Scheeler F Shen H Shue BC Siden-Kiamos I Simpson M Skupski MP Smith T Spier E Spradling AC Stapleton M Strong R Sun E Svirskas R Tector C Turner R Venter E Wang AHH Wang X Wang ZY Wassarman DA Weinstock GM Weissenbach J Williams SM Woodage T Worley KC Wu D Yang S Yao QA Ye J Yeh RF Zaveri JS Zhan M Zhang GG Zhao Q Zheng LS Zheng XQH Zhong FN Zhong WY Zhou XJ Zhu SP Zhu XH Smith HO Gibbs RA Myers EW Rubin GM Venter JC The genome sequence of Drosophila melanogaster Science 2000 287 2185 2195 10731132 10.1126/science.287.5461.2185
Workbench B
2.0.0 R R version 2.0.0
Genomatix
|
16107220
|
PMC1192798
|
CC BY
|
2021-01-04 16:32:49
|
no
|
BMC Genomics. 2005 Aug 17; 6:110
|
utf-8
|
BMC Genomics
| 2,005 |
10.1186/1471-2164-6-110
|
oa_comm
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.